Disclosure of Invention
Based on the problems existing in the prior art, the application provides an imaging method, an imaging system and a thermal imaging device of an external thermal imager of a mobile phone. The specific scheme is as follows:
the application provides an imaging method of an external thermal imager of a mobile phone, which comprises the following steps:
Determining a region to be detected and detecting requirements for the region to be detected;
Scanning the temperature distribution condition and the illumination condition of the surrounding environment of the region to be detected by a preset thermal imager arranged outside the mobile phone to respectively obtain environment temperature data and environment illumination data;
If yes, determining a type selection strategy of the band-pass filter according to the environmental illumination data and the detection requirement, selecting the band-pass filter according to the type selection strategy, and configuring the band-pass filter on the thermal imager;
synchronizing shooting parameters of a mobile phone lens to a thermal imager, determining a spatial mapping relation between the thermal imager and the mobile phone lens on an imaging area, and adjusting imaging parameters of the thermal imager according to the spatial mapping relation and the shooting parameters so that the thermal imager realizes parameter change along with the mobile phone lens synchronization;
And selecting a data fusion algorithm according to the detection requirement, and enabling the mobile phone to carry out superposition fusion on optical image data shot by a lens of the mobile phone and thermal imaging data shot by the thermal imaging instrument according to the data fusion algorithm to obtain a final thermal imaging image.
In some embodiments, the method further comprises:
Selecting pose information of a preset target object according to the optical image data;
and carrying out visual calibration on the target object in the thermal imaging image according to the pose information and the spatial mapping relation, superposing a temperature value on the target object according to the thermal imaging data, and distinguishing the target object from the surrounding environment in the thermal imaging image.
In some embodiments, the method further comprises:
comparing the optical image data with the thermal imaging data, and searching a region with difference in effect in the optical image data and the thermal imaging data to obtain a difference region;
Determining edge information and texture information of the difference region according to the optical image data to obtain optical characteristic information, and determining temperature distribution and temperature values of the difference region according to the thermal imaging data to obtain thermal characteristics;
and constructing a cross-modal feature vector of the difference region based on the optical features and the thermal features, and carrying out target identification on the difference region according to the cross-modal feature vector.
In some embodiments, evaluating whether a bandpass filter is to be externally positioned on a thermal imager specifically includes:
Calculating the average temperature of the environment of the area to be detected according to the environment temperature data;
If the average ambient temperature exceeds the preset high-temperature average value, a bandpass filter is considered to be needed;
if the average ambient temperature does not exceed the preset high-temperature average value, calculating the area ratio exceeding the preset high-temperature average value in the area to be detected according to the temperature distribution condition in the ambient temperature data, and analyzing whether the area ratio exceeds the preset ratio:
If the difference is not exceeded, analyzing the difference between the average temperature of the preset target area and the average temperature of the surrounding environment, and if the difference is lower than the preset difference, determining that the bandpass filter is required.
In some embodiments, the selection strategy comprises:
Analyzing illumination characteristic parameters including environmental illumination distribution, interference spectrum bands and dynamic illumination change frequency based on the environmental illumination data;
Determining at least one of characteristic radiation wave bands, imaging resolution and sensitivity requirements, anti-interference levels and imaging scene types of the target object as constraint conditions based on the detection requirements;
Constructing a spectrum matching model about the matching relation between the band-pass filter and the spectrum characteristic according to the illumination characteristic parameters and the constraint conditions;
a selection policy rule base related to guiding the selection of the band-pass filter is generated based on the spectrum matching model.
In some embodiments, the spectral matching model includes at least the following constraints:
The passband range constraint is matched with the characteristic radiation wave band of the target object, and the environment light intensity peak wave band is avoided;
Cut-off depth constraint, the light intensity attenuation rate of the interference wave band to the environment is not lower than a preset threshold value;
Bandwidth adaptation, selecting a fixed bandwidth or a tunable bandwidth filter according to the dynamic illumination change frequency.
In some embodiments, the mapping relationships included in the selection policy rule base at least include:
selecting a bandpass filter with corresponding center wavelength and bandwidth according to the ambient light intensity and the target radiation wave band;
when periodic strong light interference exists, a tunable filter is selected and configured for dynamic switching;
When the detection requirement is high sensitivity, a narrow-band filter is selected and the cut-off depth is increased.
In some embodiments, when the real-time requirement is involved in the detection requirement, directly performing pixel value weighted calculation according to a preset weight after registering the optical image data and the thermal imaging data, and performing pseudo-color mapping on the thermal imaging part to quickly obtain a thermal imaging image;
And/or when the content requirement is related to the detection requirement, decomposing the optical image and the thermal imaging data into low-frequency and high-frequency information with different scales, adopting weighted fusion for the low frequency, selecting fusion for the high frequency according to the absolute value of the gradient, and reconstructing the image to obtain a thermal imaging image containing rich details;
and/or when the detection requirement relates to the precision requirement, based on an end-to-end fusion algorithm of the deep learning, training by constructing a coding and decoding neural network model and utilizing a large amount of pairing data, inputting the optical image and the thermal imaging data into a trained model, and acquiring the thermal imaging image subjected to intelligent feature fusion.
The application provides an imaging system of an external thermal imager of a mobile phone, which comprises:
The input unit is used for determining a region to be detected and detecting requirements of the region to be detected;
The data acquisition unit is used for scanning the temperature distribution condition and the illumination condition of the surrounding environment of the area to be detected through a preset thermal imager arranged outside the mobile phone to respectively obtain environment temperature data and environment illumination data;
the instrument configuration unit is used for evaluating whether a bandpass filter is needed to be externally arranged on the thermal imager or not based on the environmental temperature data, if so, determining a type selection strategy of the bandpass filter according to the environmental illumination data and the detection requirement, and selecting the bandpass filter according to the type selection strategy and configuring the bandpass filter on the thermal imager;
The device comprises an instrument synchronization unit, a camera synchronization unit and a camera synchronization unit, wherein the instrument synchronization unit is used for synchronizing shooting parameters of a mobile phone lens to a thermal imager, determining a spatial mapping relation between the thermal imager and the mobile phone lens on an imaging area, and adjusting imaging parameters of the thermal imager according to the spatial mapping relation and the shooting parameters so that the thermal imager realizes parameter change along with the synchronization of the mobile phone lens;
And the output unit is used for selecting a data fusion algorithm according to the detection requirement, so that the mobile phone can superimpose and fuse the optical image data shot by the mobile phone lens and the thermal imaging data shot by the thermal imaging instrument according to the data fusion algorithm to obtain a final thermal imaging image.
The third part, the application provides a thermal imaging device, which comprises a mobile phone, a thermal imager and a band-pass filter, and is used for realizing the imaging method of the external thermal imager of the mobile phone in any one of the first parts.
The imaging method, the imaging system and the thermal imaging device of the external thermal imager of the mobile phone have the beneficial effects that the adaptability of the thermal imager in a complex environment is improved through the selection of the environment adaptive optical filter, the synchronization of equipment parameters, the spatial mapping and the intelligent data fusion, the precise coordination of optics and thermal imaging is ensured, and the imaging quality, the detection efficiency and the scene universality are obviously improved. The bandpass filter is dynamically evaluated and selected by scanning the ambient temperature and the illumination condition, and accurate adaptation can be performed on complex environments such as high temperature, strong light, smoke and the like. The mobile phone lens and the shooting parameters of the thermal imager are synchronized, and an accurate spatial mapping relation is established, so that real-time linkage of the imaging parameters of the mobile phone lens and the imaging parameters of the thermal imager is realized. And selecting an adaptive data fusion algorithm according to different detection requirements, so that the fused thermal imaging image achieves the best effect in terms of instantaneity, detail reservation, temperature information accuracy and the like. Through the modularized and intelligent design, the system is compatible with various application scenes such as industrial detection, security monitoring, outdoor search and rescue and the like.
In order to make the above objects, features and advantages of the present application more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
According to the imaging method of the external thermal imager of the mobile phone, through the selection of the environment self-adaptive optical filter, the synchronization of equipment parameters, the spatial mapping and the intelligent data fusion, the adaptability of the thermal imager in a complex environment is improved, the accurate coordination of optical and thermal imaging is ensured, and the imaging quality, the detection efficiency and the scene universality are remarkably improved. The imaging method is schematically shown in fig. 1, the principle is shown in fig. 2, and the specific scheme is as follows:
an imaging method of an external thermal imager of a mobile phone comprises the following steps:
101. Determining a region to be detected and the detection requirement of the region to be detected;
102. scanning the temperature distribution condition and the illumination condition of the surrounding environment of the area to be detected by a preset thermal imager arranged outside the mobile phone to respectively obtain environment temperature data and environment illumination data;
103. If yes, determining a type selection strategy of the band-pass filter according to the ambient illumination data and the detection requirement, selecting the band-pass filter according to the type selection strategy, and configuring the band-pass filter on the thermal imager;
104. synchronizing shooting parameters of the mobile phone lens to the thermal imager, determining a spatial mapping relation between the thermal imager and the mobile phone lens on an imaging area, and adjusting imaging parameters of the thermal imager according to the spatial mapping relation and the shooting parameters so that the thermal imager realizes parameter change along with the mobile phone lens synchronization;
105. And selecting a data fusion algorithm according to the detection requirement, and enabling the mobile phone to carry out superposition fusion on optical image data shot by a lens of the mobile phone and thermal imaging data shot by the thermal imaging instrument according to the data fusion algorithm to obtain a final thermal imaging image.
In step 101, the area to be detected is a specific spatial range indicating that scanning detection is required by using the external thermal imager of the mobile phone. The area may be a specific physical space, such as a building, an industrial facility, or a geographic area, such as a forest, a campus. For example, in the power inspection scenario, the area to be detected may be a transformer and a power transmission line in a certain substation, and in the outdoor search and rescue, the area to be detected may be a certain mountain area or jungle.
The detection requirement is the target or acquired information which is wanted to be achieved in the region by using a thermal imager, and comprises the type of the detected target (such as a human body and high-temperature equipment), the detection precision requirement (such as the temperature resolution is required to reach +/-0.1 ℃), the imaging real-time requirement (such as 15 frames of images are required to be output per second) and the like. For example, in the fault detection of industrial equipment, the detection requirement is to find abnormal temperature change of the surface of the equipment above 0.5 ℃, and in the security monitoring scene, the detection requirement is to detect and track personnel targets entering an area in real time. The detection requirement relates to real-time requirements, precision requirements, detail requirements and the like.
Real-time requirement means that the thermal imager is required to quickly acquire and transmit temperature data and image information when detecting, so as to realize instant imaging. For example, in a security monitoring scene, the thermal activity of a person or an object needs to be monitored in real time, and an abnormal situation is found in time, so that the frame rate of the thermal imager is required to be high, for example, 30 frames or more images can be output per second, so that smooth pictures and no blocking are ensured, the scene used for real-time inspection of power equipment is like some scenes, the temperature change of the equipment during running needs to be captured rapidly, and the real-time requirement is met.
Precision requirements relate to the requirements of temperature measurement accuracy and image resolution. The temperature measurement accuracy is generally expressed in terms of an error range, such as a common ±2 ℃ or ±2% of a reading, and in the fault detection of industrial equipment, it may be necessary to accurately detect a temperature change within 1 ℃ of the surface of the equipment to accurately determine the hidden trouble. In terms of image resolution, for example, different specifications such as 256×192, 384×288 and the like, the higher the resolution is, the clearer the thermal imaging image is, the more detail temperature distribution of an object can be presented, and the high-precision temperature measurement and high-resolution imaging are of great importance in scenes with extremely high precision requirements such as scientific research detection, medical auxiliary diagnosis and the like.
Details need to emphasize the ability of thermal imaging images to present subtle features and temperature differences to objects. In the inspection scene of the circuit board, the temperature condition of each element on the circuit board needs to be clearly distinguished to find out a tiny temperature abnormal area caused by short circuit and other problems, and in the building heating and ventilation detection, the tiny heat conduction difference of the wall body, the pipeline and other parts can be found out through a thermal imaging image to judge whether the heat preservation defect or the leakage problem exists. This requires a thermal imager with high sensitivity that can capture small temperature changes, e.g., a thermal sensitivity NETD of 40mK or less, and thus presents rich detailed information
Step 102 is a core link of environmental perception, which provides data support for subsequent filter type selection and parameter adjustment. In the application, the preset thermal imager refers to a thermal imaging device which is independent of the mobile phone body and is matched with the mobile phone through a physical interface (such as USB-C, lightning) or a wireless connection (such as Bluetooth and Wi-Fi). The infrared detector, the optical lens and other components are integrated in the infrared detector, and the infrared detector has the capability of sensing infrared radiation and converting the infrared radiation into an electric signal. The device has finished driving adaptation and function debugging in advance for mobile phone systems (such as Android and iOS), and can be controlled directly through mobile phone APP.
In the application, the thermal imager is required to be provided with an infrared detection module (such as an uncooled micro-bolometer) and a visible light sensing module (or linked with a mobile phone lens) at the same time, so that synchronous acquisition of temperature and illumination data is ensured. The thermal imager scans the target area in the horizontal or vertical direction through the lens, and acquires the infrared radiation intensity (corresponding to temperature) and the visible light intensity (corresponding to illumination) of each point in the space. In practical applications, pixel data may be acquired row by row through a microelectromechanical system (MEMS) driven lens, or a Focal Plane Array (FPA) detector may be used to capture radiation signals across the field of view at once. The temperature distribution refers to the temperature of the surface of each object in the region to be detected and the space distribution state, and the environmental temperature data is the temperature value (unit: ° C or K) of each point recorded in a numerical value or matrix form, and generally corresponds to each pixel point of the pixel matrix of the thermal imager. Illumination refers to the intensity, direction and spectral composition of visible light in the environment (e.g., sunlight, lamp light, infrared interference light), and ambient illumination data is typically expressed in terms of illumination intensity values (units: lux, lux) or spectral distribution curves. A visible light sensor (such as a CMOS camera) is arranged in the thermal imaging instrument or visible light information is synchronously acquired through a lens of the mobile phone.
Step 103 is a key link for optimizing thermal imaging quality, and targeted suppression of environmental interference is realized through intelligent evaluation and type selection.
The environmental temperature data reflects the heat radiation intensity of each point in the scene, and the evaluation core is used for judging whether the environmental heat interference can influence the target detection. Specific evaluation indexes include:
temperature dynamic range if the environmental temperature fluctuation range exceeds the target temperature change range (for example, the ambient temperature change of the industrial furnace reaches +/-50 ℃ and the target abnormal temperature change is only +/-2 ℃), the optical filter is needed to restrain the environmental radiation.
The background and target temperature difference is that when the background temperature is close to the target temperature (such as the difference between the human body temperature and the 30 ℃ environment temperature is small), the contrast of the signals in the specific wave band needs to be enhanced through the optical filter.
Analyzing the radiation peak wave band of environmental heat source (such as high temperature metal radiation concentrated at 3-5 μm and human body radiation concentrated at 8-14 μm), and judging whether overlapping with the target signal wave band
The ambient light data provides visible light interference intensity and spectrum distribution information, and guides the design of a light transmission wave band of the optical filter. Specific evaluation indexes include:
strong light scene (illumination intensity >10000 Lux):
Selecting an optical filter with a cut-off depth OD of more than or equal to 6, and inhibiting near infrared interference (such as near infrared radiation in sunlight) of 0.7-3 mu m.
An example is that when the outdoor inspection is carried out, the light transmittance of the optical filter to 780nm wavelength is less than 0.0001%, so that thermal imaging is not interfered by sunlight.
Low light scene (illumination intensity <10 Lux):
a broadband pass filter (such as 8-14 mu m full-band transmission) is selected, so that the luminous flux of weak signals is improved, and stray light reflection is reduced through optical coating.
In practical application, the optical filter can be quickly fixed at the front end of the lens of the thermal imaging instrument through the magnetic attraction structure, so that the coaxiality of the optical path is ensured to be less than 0.1 degrees. The rotary filter wheel can be adopted, a plurality of filters with different parameters are integrated, and the switching is driven by a stepping motor. After the filter is installed, the spectral response function of the detector is adjusted according to the actual light transmission curve of the filter, and attenuation introduced by the filter is compensated. If the optical filter has reduced light transmittance (e.g., a narrow band pass filter), the detector gain is increased to maintain the signal-to-noise ratio.
In the application, the core function of the band-pass filter in the thermal imaging system is to selectively filter infrared radiation in a specific wave band, and the definition and detection accuracy of thermal imaging are obviously improved by inhibiting environmental interference and enhancing the contrast of a target signal. The bandpass filter is designed through an interference film layer (such as a multi-layer dielectric film stack), only allows infrared light in a specific wavelength range to pass through, and blocks radiation interference of other wave bands. When the ambient temperature is too high (e.g., industrial high Wen Changjing), the infrared energy radiated by the high-temperature object is concentrated in the short wavelength band (e.g., 3-5 μm). At this time, a narrow band-pass filter (for example, only allowing 8-14 μm wave band to pass) is selected to block short-band energy (for example, strong interference signal of 3-5 μm) of high-temperature radiation of environment, and long-band signal (for example, 8-14 μm characteristic radiation of human body and normal temperature equipment) of self-radiation of the target is reserved. If multiple heat sources exist in the environment (such as a high-temperature area and a normal-temperature area are mixed), a broadband pass filter (such as an optimized 7-15 mu m wave band) can be adopted, and the differential radiation signals of the target and the background are reserved while the main interference wave band is restrained.
In some embodiments, the evaluation of whether the external bandpass filter of the thermal imager is needed is specifically shown in fig. 3, and the evaluation comprises calculating the average ambient temperature of a region to be detected according to ambient temperature data, determining that the bandpass filter is needed if the average ambient temperature exceeds a preset high-temperature average value, calculating the area ratio of the region to be detected exceeding the preset high-temperature average value according to the temperature distribution condition in the ambient temperature data if the average ambient temperature does not exceed the preset high-temperature average value, analyzing whether the area ratio exceeds a preset ratio, determining that the bandpass filter is needed if the area ratio exceeds the preset ratio, and analyzing the difference between the average temperature of a preset target region and the average ambient temperature if the area ratio does not exceed the preset high-temperature average value, and determining that the bandpass filter is needed if the difference is lower than the preset difference. The infrared energy radiated by the high-temperature environment can be superimposed on the target signal, so that the contrast ratio of the target and the background is reduced. Also, the radiation band of the high temperature object (e.g., 3-5 μm band of the industrial furnace) may overlap with the target characteristic band, resulting in difficulty in distinguishing the target signal by the detector. Therefore, the heat radiation interference intensity can be predicted by evaluating the ambient temperature, and a basis is provided for the optical filter type selection.
If the ambient average temperature approaches the target temperature, the natural contrast decreases. For example, when detecting a 38 ℃ human body, if the ambient temperature is 35 ℃, the temperature difference is only 3 ℃, which is lower than the ideal detection threshold of the thermal imager. When the high temperature region duty ratio P exceeds a threshold, a large area of high temperature background may produce a "halo effect" that blurs the target boundary. After the band-pass filter filters the strong interference wave band of the environmental radiation, the background noise energy received by the detector is reduced, and the relative intensity of the target signal is improved. For example, high temperatures in areas above 30% of the furnace periphery can lead to weak temperature anomalies in remote pipes that are difficult to identify. Even if the overall temperature of the environment is not high, too small a temperature difference between the target and the local background can result in signals being masked by noise. Through quantitative analysis of the three dimensions, the detectability of the target signal can be scientifically evaluated, and whether the optical filter is required to enhance the contrast ratio is determined. The real-time change of the ambient temperature can lead to baseline drift of thermal imaging, and the bandpass filter can reduce the interference of temperature fluctuation on detection by stabilizing the incident spectrum.
In some specific embodiments, the type selection strategy comprises analyzing illumination characteristic parameters including environmental illumination distribution, interference spectrum wave bands and dynamic illumination change frequency based on environmental illumination data, determining at least one of characteristic radiation wave bands of a target object, imaging resolution and sensitivity requirements, anti-interference level and imaging scene type as constraint conditions based on detection requirements, constructing a spectrum matching model related to a matching relation between a band-pass filter and spectral characteristics according to the illumination characteristic parameters and the constraint conditions, and generating a selection strategy rule base related to guiding band-pass filter selection based on the spectrum matching model. After the filter is deployed, the model selection effect is evaluated through the contrast (such as temperature gradient > delta T) between the contrast target and the background, if the contrast target and the background do not meet the requirements, the parameters are adjusted for reselection, and the filter parameters are adjusted through real-time environmental illumination data, so that the problem that the static filter cannot cope with complex scenes is solved. Meanwhile, the target characteristics, the environmental interference and the imaging performance are considered, and the model selection deviation caused by a single parameter is avoided. And a rule base is constructed based on a mathematical model and scene experience, so that the complexity of system design is reduced, and the instantaneity is improved.
Based on the ambient illumination data, at least one parameter of the ambient light intensity distribution (such as the intensity ratio of ultraviolet light/visible light/infrared light), the dominant interference spectrum band (such as the radiation band of an ambient heat source, the strong reflection band in sunlight) and the dynamic illumination change frequency (such as the time interval of periodic strong light interference) is extracted. The main interference wave bands (such as visible light in sunlight and near infrared radiation of industrial heat sources) are identified through the ambient light data, so that imaging noise caused by overlapping of the ambient light and target heat radiation is avoided. For dynamic scenes (such as day-night alternation and strong light flickering), an optical filter (such as an electrically-controlled tunable optical filter) with matched response speed is required to be selected.
Illustratively, the ambient illumination data is decomposed into different frequency components by fourier transform or wavelet transform, and the center wavelength, bandwidth, and intensity peak of the interference band are extracted. For example, in the sunlight environment, the near infrared band (0.7-3 μm) usually has Jiang Fanshe peaks, and in the industrial furnace scene, the mid-wave infrared (3-5 μm) may be a major source of interference due to high temperature equipment radiation. And calculating the time-varying rate (such as delta I/delta t) of the illumination intensity, judging whether the environment has fast-varying strong light interference (such as car light flickering and arc welding), and providing a basis for selecting an optical filter with matched response speed.
Based on the detection requirement, at least one constraint condition of characteristic radiation wave bands (such as 8-14 mu m wave bands of human body radiation) of a target object, imaging resolution and sensitivity requirements (such as minimum temperature resolution and spatial resolution), anti-interference level (such as threshold requirement for suppressing ambient light noise) and imaging scene types (such as night detection, high-temperature industrial scene and medical detection) is determined. The difference of heat radiation wave bands of different targets (such as human body, mechanical parts and high-temperature fluid) is obvious, and the pass band of the optical filter needs to be matched (such as 8-14 mu m wave band for human body temperature measurement). The narrow band filter can improve spectral purity, but can reduce luminous flux, and needs to balance sensitivity and noise, and the high cut-off depth filter can inhibit ambient light, but can increase cost and complexity of light path.
According to the blackbody radiation law, the radiation spectrum of the target at a temperature T is calculated. The energy concentration of the characteristic radiation band (e.g., the 8-14 μm band contains 98% of the radiation energy) is calculated by presetting the target temperature range (e.g., 37 ℃ C. Of the human body to a peak wavelength of about 9.5 μm). The detection requirement is converted into a multidimensional constraint vector, such as a spatial resolution requirement, temperature sensitivity, noise equivalent temperature difference and the like.
The core parameters of the filter include center wavelength, half-width, cut-off depth, peak transmittance, etc. The center wavelength is the center position of the passband, determines the main band of the system response, the half-width is the measurement of the passband width, influences the spectral resolution, the cut-off depth is the attenuation capability of light outside the passband, which is usually expressed by logarithm (for example, OD4 represents attenuation ratio is 10-4), and the peak transmittance is the maximum light transmission efficiency in the passband.
The spectrum matching model is a model constructed based on a mathematical algorithm or machine learning and is used for describing the matching relation between the target spectrum characteristics (such as reflectivity, transmissivity, emissivity and the like) and the spectrum parameters (such as center wavelength, bandwidth, cut-off depth and the like) of the band-pass filter. And judging whether the specific optical element meets the target spectrum requirement or not through quantitatively analyzing indexes such as spectrum overlapping degree, characteristic peak matching degree and the like. And when the optical filter is selected, calculating the matching degree fraction of the target spectrum and the transmission spectrum of the optical filter, and predicting optimal optical filter parameters (such as center wavelength and bandwidth) based on spectrum data.
The spectrum matching model realizes the accurate adaptation of the band-pass filter and the complex scene through a multi-dimensional constraint condition. In some specific embodiments, the spectrum matching model at least comprises the following constraint conditions of passband range constraint, cut-off depth constraint and bandwidth adaptability, wherein the passband range constraint is used for matching a characteristic radiation wave band of a target object and avoiding an environment light intensity peak wave band, the cut-off depth constraint is used for limiting the light intensity attenuation rate of an environment interference wave band to be not lower than a preset threshold value, and the bandwidth adaptability is used for selecting a fixed bandwidth or tunable bandwidth filter according to dynamic illumination change frequency. The bandpass filter is precisely standardized from three key dimensions of passband, cutoff depth and bandwidth, so that the bandpass filter is ensured to adapt to complex environments and detection requirements.
The object can generate stronger radiation in a specific wave band due to the temperature of the object, for example, the radiation of a human body is obvious in the wave band of 8-14 mu m, and the ambient light has strong interference in certain wave bands (for example, the light intensity of sunlight is high in the range of 0.5-1.2 mu m). The passband range constraint requires that the passband of the optical filter coincides with the characteristic radiation wave band of the target object, and simultaneously avoids the peak wave band of the environmental light intensity, thereby ensuring the effective passing of the target signal and reducing the environmental interference from entering the thermal imaging system. The method comprises the steps of establishing a characteristic radiation wave band database of a target object in advance, determining characteristic wave bands according to the target type during detection, monitoring the spectrum distribution of ambient light in real time through a spectrometer and other equipment, and identifying a light intensity peak wave band. For example, when detecting human body outdoors, a filter with a passband of 8-14 μm is selected to avoid interference wave bands of 0.5-1.2 μm of sunlight.
The cut-off depth determines the attenuation capability of the filter to light of a specific wave band and is measured by the light intensity attenuation rate. Interference wave bands exist in the environment, such as strong light interference generated by welding in an industrial scene, and the light intensity attenuation rate of the optical filter to the environment interference wave bands is required to be not lower than a preset threshold value by setting cut-off depth constraint, so that interference can be effectively restrained, and the thermal imaging image quality is improved. And setting corresponding attenuation rate thresholds according to the interference intensities of different application scenes, wherein the attenuation rate is not lower than 99% when the attenuation rate is required by the strong interference scenes.
The dynamic illumination variation frequency affects the selection of the filter bandwidth. The fixed bandwidth filter has a simple structure and low cost, but cannot adapt to a scene with rapid change of illumination, and the tunable bandwidth filter can dynamically adjust the passband bandwidth according to the environment. The proper bandwidth type is selected according to the dynamic illumination change frequency, so that the cost and the performance can be balanced while the imaging effect is ensured.
The model selection strategy rule base is a rule set based on industry experience, design specification or experimental verification and is used for guiding the model selection process of the optical element. The rules in the rule base are typically in the form of "condition-action" (e.g. "if the target spectral peak is Xnm, a filter with a center wavelength of x±ynm is preferred"). And converting the standardized flow into executable rules, and quickly narrowing the selection range or directly giving a recommended scheme through logic judgment. Combining the output result of the spectrum matching model, and performing secondary screening (such as excluding options which do not meet the cost or process requirements) through a rule base. The spectrum matching model provides data support for the rule base, and the matching rule obtained by analyzing a large amount of spectrum data can be converted into a specific rule in the rule base. For example, the model finds that "when the half-peak width of the target spectrum is smaller than 50nm, the bandwidth of the optical filter is smaller than or equal to 30nm to ensure the signal to noise ratio", and the conclusion can be used as a rule in a rule base. The spectrum matching model is a data brain and is responsible for excavating rules from spectrum characteristics, the model selection strategy rule library is a decision engine and is responsible for actually combining data results with engineering, the data and the engineering are combined to form a complete model selection system from spectrum analysis to engineering landing, the technical feasibility is ensured, and the scheme practicability is ensured.
In some embodiments, the mapping relation included in the policy rule base at least comprises a band-pass filter for selecting corresponding center wavelength and bandwidth according to the environmental light intensity and the target radiation wave band, a tunable filter is selected and configured to perform dynamic switching when periodic strong light interference exists, and a narrow-band filter is selected and cut-off depth is increased when the detection requirement is high sensitivity. The selection strategy rule base directly relates the environmental characteristics and the detection requirements to the filter core parameters through the scene mapping relation.
Illustratively, a decision tree is constructed based on the environmental parameters and the probe needs. And adopting a three-layer decision tree structure, and sequentially evaluating environmental conditions, interference types and detection requirements according to the priority. The input environment light intensity level (low/medium/high), the target characteristic radiation wave band, and the selection strategy rule base outputs the center wavelength and the bandwidth of the optical filter.
The illumination intensity is acquired in real time through a mobile phone lens or a built-in photosensitive sensor of the thermal imager and is compared with a preset threshold (100 Lux, 1000 Lux) to judge the light intensity level. For example, when the target spectrum peak is Xnm, a filter having a center wavelength of x±ynm is preferably selected. When the energy density of the environment light intensity in the X wave band exceeds the threshold A and the characteristic radiation wave band of the target object is Y, a band-pass filter with the central wavelength being Y central value and the bandwidth being B is selected, and meanwhile, the cut-off band is ensured to cover the X wave band.
The periodic fluctuation of the ambient light is detected, and the interference band overlaps with the target band. For example, if there is periodic fluctuation in the ambient light intensity, the frequency exceeds 1Hz, the overlap ratio of the interference band and the target band exceeds 20%, and a tunable filter (such as an LCTF or MEMS filter) is activated. The signal is enhanced using narrowband pass (bandwidth=b/2) during the interference trough period, and the interference peak period expands the bandwidth (bandwidth=b×1.5) to increase the luminous flux.
The detection requirement comprises a high sensitivity index or requires detection of a small temperature difference, and a narrow-band filter is preferably selected to improve the cut-off depth and reduce the ripple degree in the passband.
For example, a decision tree is constructed based on environmental parameters and probe needs as follows:
input ambient light data (I (lambda)), target characteristics (T, epsilon), performance requirements (C)
The flow is as follows:
1. Judging the environmental interference level:
-marking as a medium wave interference scenario if I (λ) exceeds a threshold in the 3-5 μm band
-Marking near infrared interference scene if I (λ) exceeds a threshold in the 0.7-3 μm band →
2. Determining a characteristic band range R from a target temperature T
3. Matching the type of the filter:
-if R ⊂ -14 μm and there is medium wave interference → select long pass filter (λ0 =10μm)
-If R ⊂ -5 μm and target temperature >500 ℃ →select medium pass filter (λ0 =4.2 μm)
4. Adjusting FWHM and OD according to performance requirement C
Output filter parameter combination (lambda0,FWHM,OD,Tp)
Verification of evaluation of shape selection Effect by contrast of contrast target and background
In some embodiments, adaptive adjustment rules are designed for complex scenarios. When the ambient light intensity is detected to exceed the threshold value, the cut-off depth is automatically increased (such as OD is increased from 4 to 6), the signal-to-noise ratio (SNR) under different filter parameter combinations is calculated, and the parameter combination with the largest SNR is selected as the optimal solution. When the target temperature range is wide (e.g., -20 ℃ -100 ℃), a wider bandwidth (e.g., fwhm=8 μm) is selected, and when high resolution temperature measurement is required (e.g., ±0.1 ℃), a narrower bandwidth (e.g., fwhm=2 μm) is selected.
The method has the advantages that the limitations of the traditional fixed optical filters are broken through by dynamically adjusting the parameters of the optical filters through real-time environment perception, the optimal compromise is realized among signal enhancement, noise suppression and system response speed, the complex spectrum matching problem is converted into an executable decision rule, and the system design difficulty is reduced.
In some embodiments, the method further comprises the steps of selecting pose information of a preset target object according to the optical image data, performing visual calibration on the target object in the thermal imaging image according to the pose information and the spatial mapping relation, superposing a temperature value on the target object according to the thermal imaging data, and distinguishing the target object from surrounding environments in the thermal imaging image. The optical image data is preprocessed, wherein the preprocessing comprises image enhancement, denoising and illumination normalization, so that the accuracy of a target detection algorithm is improved. The preset target object is identified by a target detection algorithm (such as YOLO, SSD), and the position coordinates (such as bounding box coordinates (x1,y1,x2,y2)) and the contour features (such as polygon vertex coordinate sequences) of the target are output. And establishing a spatial mapping relation between the optical image and the thermal imaging image, acquiring transformation matrixes (including translation, rotation and scaling parameters) of the optical image and the thermal imaging image through a camera calibration technology, and mapping target positions and contour coordinates in the optical image to a thermal imaging image coordinate system. And carrying out visual calibration on the target object in the thermal imaging image according to the mapped coordinates, wherein the calibration method comprises the steps of, but is not limited to, superposing a boundary box at a target position, drawing highlight lines along a target contour, and filling semitransparent color blocks in a target area.
Wherein the pose information includes a position and a pose. The position comprises two-dimensional coordinates or three-dimensional space coordinates of the target in the optical image, and the gesture is the direction information such as the direction, the angle and the like of the target. The model of YOLO, fasterR-CNN, etc. is used to identify the target (e.g. human body, equipment), output the bounding box coordinates, and estimate the pose by key point detection. The three-dimensional model of the target can be obtained through a built-in 3D sensor of the mobile phone, and a pose matrix is generated. For example, the left shoulder coordinates (200,150) and the right shoulder coordinates (300,150) of the human body in the optical image are detected, and the calculated posture angle is deflected by 10 ° to the right. And (3) mapping the target pose in the optical image to a thermal imaging image coordinate system by using a spatial mapping model (such as a rotation matrix R, a translation vector T and a homography matrix H) of the thermal imager and the mobile phone lens, which is established in the step 104. And drawing graphs such as a target boundary box, a center point, a gesture arrow and the like in the thermal imaging image. For example, the target area is marked with a red rectangular box, and the arrow points from the center point to the gesture direction (length represents confidence). An independent pseudo-color map (e.g., high temperature region red, low temperature region blue) is applied to the target region, as opposed to a default pseudo-color of the background (e.g., iron-red palette).
Illustratively, temperature information of a target area is extracted from raw data (such as a 16-bit gray scale map, each pixel corresponds to a temperature value) acquired from a thermal imager. And applying a Canny operator to the target boundary to improve the definition of the contour. Highlighting the target detail by Gaussian blur processing the background region, expanding the gray dynamic range of the target region to [0,255], and compressing the background to [50,200].
The space mapping relation establishment comprises the steps of collecting synchronous images of an optical camera and a thermal imager by using a checkerboard calibration plate, calculating an internal reference matrix, an external reference matrix and a distortion coefficient of the synchronous images, extracting homonymous characteristic points in the optical image and the thermal image based on characteristic point matching, and fitting a transformation matrix by using a RANSAC algorithm.
Establishing a target feature library, storing optical features (such as shapes and textures) and thermal features (such as temperature distribution modes) of a preset target object, and verifying consistency of a target area in a thermal imaging image based on the feature library to eliminate false calibration of non-target objects.
In some embodiments, the method further comprises the steps of comparing the optical image data with the thermal imaging data, searching a region with difference in effect in the optical image data and the thermal imaging data to obtain a difference region, determining edge information and texture information of the difference region according to the optical image data to obtain optical characteristic information, determining temperature distribution and temperature values of the difference region according to the thermal imaging data to obtain thermal characteristics, constructing a cross-modal characteristic vector of the difference region based on the optical characteristics and the thermal characteristics, and carrying out target identification on the difference region according to the cross-modal characteristic vector.
And establishing a pixel mapping relation between the optical lens and the thermal imager through homography conversion or camera calibration. For example, the internal reference matrix and the external reference matrix (rotation matrix R and translation vector T) of the two are obtained by using a Zhang calibration method, so that the coordinate alignment of the same physical point in two images is ensured, and the error is controlled within 1 pixel. The time difference between the optical and thermal imaging data acquisition is less than 50ms by adopting hardware triggering (synchronous pulse signals) or software interpolation (linear interpolation based on time stamps, for example), so that the misjudgment of the difference of a dynamic target (such as a moving vehicle) caused by displacement is avoided. The difference value for each pixel is calculated for the registered optical image and the thermographic image. Continuous regions with a difference value above a threshold (e.g., 0.4) are marked as difference regions by thresholding (e.g., otsu adaptive thresholding) or clustering algorithms (e.g., K-means). For example, the optical image is displayed as a wall (gray value 180), the corresponding region in thermal imaging is at a high temperature (gray value 220), the difference value d=0.22, and the difference region is marked if the threshold t=0.2.
The contours of the difference regions are extracted by using a Canny edge detection algorithm, and edge densities (the number of edge pixels/the total number of pixels in the region) and edge direction histograms (the main direction duty ratios of 0 °/45 °/90 °/135 °/and the like are counted) are calculated. For example, the edge density of the mechanical parts is high (> 0.3) and the main direction is concentrated (e.g. 90 ° 60%) while the vegetation edge density is low (< 0.1) and the direction is cluttered. Texture features are extracted using either Local Binary Pattern (LBP) or gray level co-occurrence matrix (GLCM). For example, LBP calculates the gray difference between each pixel and its neighborhood to generate statistics such as texture pattern histogram, statistical mean, variance, etc., GLCM analyzes the gray correlation of pixel pairs to extract features such as contrast, entropy, angular second moment, etc. for distinguishing smooth surface from rough surface.
And calculating a temperature gradient matrix of the difference region, and judging a heat flow mode (such as uniform heat dissipation or local heat source) through gradient direction consistency. For example, the temperature gradient in the region of the pipe leakage is radial, while the normal region gradient is gentle. And extracting the average temperature, the temperature standard deviation, the temperature extreme value and the temperature entropy (measuring the uniformity of the temperature distribution) of the difference region. For example, the temperature entropy of a human body target is low (< 2.0), the temperature distribution is relatively uniform, the temperature entropy of a high-temperature fault point is high (> 3.0), and remarkable temperature fluctuation exists.
The optical features and thermal features are concatenated into a multi-dimensional vector that fuses the geometry (edges), surface properties (texture), thermal state (temperature values), and thermal distribution pattern (gradient/entropy) of the target to form a comprehensive representation of the target. The optical image provides visual clues such as the shape, texture and the like of the target, the thermal imaging reveals the temperature distribution and the thermal characteristics, and the combination of the two can solve the problem of insufficient single-mode information (such as that the thermal imaging can not distinguish objects with different materials and similar temperatures). The cross-modal feature fusion improves the target recognition accuracy by 30% -50% compared with a single mode, and is particularly suitable for camouflage target detection (such as high-temperature objects hidden in a grass).
After which object recognition is performed. Illustratively, Z-score normalization is performed on numerical features (e.g., temperature, texture mean), dimensional effects are eliminated, and feature dimensions are reduced, e.g., 20-dimensional original features are compressed to 8 dimensions, preserving more than 90% of the information entropy. Voting decisions or deep learning networks through a plurality of decision trees. The spatio-temporal errors are controlled within an acceptable range using hardware timestamp synchronization (e.g., GPS time synchronization) and dynamic registration algorithms (e.g., optical flow based real-time registration). Key features may be automatically screened by an attention mechanism (e.g., SENet) to suppress extraneous noise (e.g., interference of thermal features by illumination noise in an optical image). Through systematic difference analysis and feature fusion, the technology span from 'passive imaging' to 'active understanding' is realized, and a more accurate target analysis means is provided for the fields of industrial detection, intelligent security, medical diagnosis and the like.
In some embodiments, when the real-time requirement is involved in the detection requirement, the pixel value weighting calculation is directly performed according to the preset weight after the registration of the optical image data and the thermal imaging data, and the pseudo-color mapping is performed on the thermal imaging part, so that the thermal imaging image can be obtained quickly. The method comprises the steps of registering an optical image shot by a mobile phone lens with thermal imaging data shot by a thermal imager, calculating a homography matrix through a feature point matching algorithm, mapping the thermal imaging data to an optical image coordinate system, carrying out resolution adjustment on the thermal imaging data by a bicubic interpolation method if resolution of the optical image and the thermal imaging data are inconsistent, and determining optical image weight and thermal imaging weight according to a detection scene. The registered optical image and the thermal imaging image are fused, so that the thermal imaging temperature is normalized to the gray value after [0,255 ]. And performing pseudo-color coding on the fused thermal imaging part by adopting a Jet color chart and overlapping the pseudo-color coding with the optical image to obtain a final thermal imaging image. Only simple algebraic operation and color mapping are needed, the calculated amount is extremely small, and the method is suitable for real-time monitoring (such as security protection and unmanned aerial vehicle inspection). The pseudo color makes the temperature difference clear at a glance, and the weighted fusion retains the contour information of the optical image, thereby facilitating the rapid positioning of the target. Only realizing the fusion of pixel level, not deeply extracting the characteristics, and limited details and precision, and being suitable for scenes with high requirements on real-time performance and lower requirements on details.
And/or when the content requirement (the requirement for fusion of image details and multi-scale information) is related in the detection requirement, the optical image and the thermal imaging data are decomposed into low-frequency and high-frequency information with different scales, the low-frequency is subjected to weighted fusion, the high-frequency is subjected to selective fusion according to the absolute value of the gradient, and the image is reconstructed, so that the thermal imaging image containing rich details is obtained. The method comprises the steps of respectively constructing a Gaussian pyramid and a Laplace pyramid for optical images shot by a mobile phone lens and thermal imaging data shot by a thermal imager, wherein the Gaussian pyramid is obtained by carrying out Gaussian convolution and downsampling on images, the Laplace pyramid is obtained by subtracting adjacent layers of the Gaussian pyramid, the low-frequency information of the bottommost layer of the Gaussian pyramid is fused in a weighted average mode, a weight coefficient is set according to detection requirements, the high-frequency information of each layer of the Laplace pyramid is compared with the gradient absolute value of corresponding pixels of the optical images and the thermal imaging images, the pixel value of the image where the pixel with the large gradient absolute value is selected as the fused pixel value, and the image is reconstructed layer by layer from the fused low-frequency image through upsampling and adding the fused high-frequency image of the corresponding layer until the final thermal imaging image is obtained. The low-frequency fusion keeps the whole structure, and the high-frequency selection keeps the edges and textures of two types of images (such as texture details of optical images and temperature edges of thermal imaging), so that the method is suitable for scenes (such as industrial detection and medical images) needing fine analysis. Through multi-scale decomposition, images with different resolutions or noise levels can be flexibly processed, and useful information is enhanced while noise is suppressed.
And/or when the precision requirement is related in the detection requirement, based on an end-to-end fusion algorithm of the deep learning, training by building a coding and decoding neural network model and utilizing a large amount of pairing data, inputting the optical image and the thermal imaging data into a trained model, and acquiring the thermal imaging image subjected to intelligent feature fusion. Specifically, a coding and decoding neural network model based on U-Net variants is built, a convolutional layer sharing weight is adopted by an encoder part to conduct feature extraction on an optical image and a Thermal imaging image, feature information of different scales is fused through jump connection by a decoder part, paired data of the optical image and the Thermal imaging image are collected, normalization processing is conducted on the paired data of the optical image and the Thermal imaging image, the paired data are divided into a training set, a verification set and a test set, if a public data set is adopted, data sets such as FLIR, MSCOCO-Thermal and the like are selected, and mean square error MSE and structural similarity SSIM are used as joint loss functions. Inputting the optical image shot by the lens of the mobile phone and the thermal imaging data shot by the thermal imager into a trained model, outputting the model to obtain a final thermal imaging fusion image, quantifying the model into FP16 precision when the mobile phone is deployed, and carrying out acceleration reasoning by utilizing the NPU of the mobile phone. The deep learning can capture complex nonlinear relations, and the fusion result is more accurate in detail (such as tiny temperature difference) and semantic level (such as target recognition), and is suitable for high-precision detection (such as space remote sensing and medical diagnosis). The trained model can adapt to different scenes without manually adjusting parameters, and has good robustness to noise, illumination change and the like. A large amount of labeling data and high-performance hardware (such as GPU) training are needed, the real-time performance is poor, but the speed of the reasoning stage can be optimized.
The fusion method under different requirements reflects the trade-off of speed, precision and complexity, wherein the real-time preference selects simple weighting, the content richness preference adopts multi-scale decomposition, and the high-precision scene depends on the strong characterization capability of deep learning. In practical application, the method can be flexibly selected or used in combination according to hardware conditions, data scale and task targets.
The application also provides an imaging system of the external thermal imager of the mobile phone, as shown in fig. 4, the imaging system comprises:
an input unit A1, configured to determine a region to be detected and a detection requirement of the region to be detected;
The data acquisition unit A2 is used for respectively acquiring environmental temperature data and environmental illumination data by scanning the temperature distribution condition and the illumination condition of the surrounding environment of the area to be detected through a preset thermal imager arranged outside the mobile phone;
The instrument configuration unit A3 is used for evaluating whether the bandpass filter is needed to be externally arranged on the thermal imager based on the environmental temperature data, if so, determining a type selection strategy of the bandpass filter according to the environmental illumination data and the detection requirement, and selecting the bandpass filter according to the type selection strategy and configuring the bandpass filter on the thermal imager;
the instrument synchronization unit A4 is used for synchronizing shooting parameters of the mobile phone lens to the thermal imager, determining a spatial mapping relation between the thermal imager and the mobile phone lens on an imaging area, and adjusting imaging parameters of the thermal imager according to the spatial mapping relation and the shooting parameters so that the thermal imager realizes parameter change along with the synchronization of the mobile phone lens;
and the output unit A5 is used for selecting a data fusion algorithm according to the detection requirement, so that the mobile phone can carry out superposition fusion on optical image data shot by a lens of the mobile phone and thermal imaging data shot by the thermal imaging instrument according to the data fusion algorithm, and a final thermal imaging image is obtained.
The application also provides a thermal imaging device which comprises a mobile phone, a thermal imager and a band-pass filter, and is used for realizing the imaging method of the external thermal imager of the mobile phone.
The thermal imaging device is shown in fig. 5, and the mobile phone 1 is used as a core carrier to integrate components such as a main board, a processor, a display screen, a storage and the like. The optical lens 14 is positioned on the back of the body of the mobile phone 1. The thermal imager 2 is fixed outside the mobile phone 1 (such as back and side) through an interface (such as USB-C, lightning) or a buckle, and is responsible for collecting thermal radiation data (thermal imaging data), and a bandpass filter 3 can be arranged in front of the lens. The bandpass filter 3 is directly installed in front of the lens of the thermal imager 2 (and is tightly attached to or fixed by a bracket) and is used for filtering light in a specific wave band (for example, only allowing the characteristic radiation wave band of a target object to pass through), and can be installed in front of the lens of the thermal imager 2 by a thread, a buckle and other structures, and the filters with different specifications can be replaced according to environmental requirements.
The present application provides a computer program product comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes an imaging method of the external thermal imager of the mobile phone. The imaging method of the external thermal imager of the mobile phone is applied to a computer program product and is convenient to execute.
The application also provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the imaging method steps of a mobile phone external thermal imager as described above.
The computer storage media of the present application may take the form of any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The imaging method of the external thermal imager of the mobile phone is applied to a computer readable storage medium, and a computer program is stored on the computer readable storage medium, and when the program is executed by a processor, the steps of the garment simulation method provided by the application are realized, so that the method is simple, convenient and quick, is easy to store and is not easy to lose.
The application provides an imaging method, an imaging system and a thermal imaging device of an external thermal imager of a mobile phone, which improve the adaptability of the thermal imager in a complex environment, ensure the accurate coordination of optics and thermal imaging and remarkably improve the imaging quality, the detection efficiency and the scene universality through the selection of an environment self-adaptive optical filter, the synchronization of equipment parameters, the spatial mapping and the intelligent data fusion. The bandpass filter is dynamically evaluated and selected by scanning the ambient temperature and the illumination condition, and accurate adaptation can be performed on complex environments such as high temperature, strong light, smoke and the like. The mobile phone lens and the shooting parameters of the thermal imager are synchronized, and an accurate spatial mapping relation is established, so that real-time linkage of the imaging parameters of the mobile phone lens and the imaging parameters of the thermal imager is realized. And selecting an adaptive data fusion algorithm according to different detection requirements, so that the fused thermal imaging image achieves the best effect in terms of instantaneity, detail reservation, temperature information accuracy and the like. Through the modularized and intelligent design, the system is compatible with various application scenes such as industrial detection, security monitoring, outdoor search and rescue and the like.
It will be appreciated by those skilled in the art that the modules of the application described above may be implemented in a general purpose computing system, concentrated on a single computing system, or distributed across a network of computing systems, and that they may alternatively be implemented in program code executable by a computer system, such that they are stored in a memory system and executed by the computing system, or individually fabricated into individual integrated circuit modules, or multiple modules or steps within them are fabricated into a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.
Note that the above is only a preferred embodiment of the present application and the technical principle applied. It will be understood by those skilled in the art that the present application is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the application. Therefore, while the application has been described in connection with the above embodiments, the application is not limited to the embodiments, but may be embodied in many other equivalent forms without departing from the spirit or scope of the application, which is set forth in the following claims.
The foregoing disclosure is merely illustrative of some embodiments of the application, and the application is not limited thereto, as modifications may be made by those skilled in the art without departing from the scope of the application.