Linear polarized light-based passive reflection type target recognition method and recognition system thereofTechnical Field
The invention belongs to the technical field of complex environment target recognition, and particularly relates to a passive reflection type target recognition method and system based on linearly polarized light.
Background
In strong light settings, target imaging faces a number of complex and tricky problems, which make target recognition a very challenging task.
Intense light can cause image sensors to generate significant amounts of noise, which can be diverse in types, including thermal noise, shot noise, etc., that are interleaved with useful target signals. The traditional filtering method is used for removing noise and simultaneously blurring edge and detail information of an image, so that target characteristics become unclear, and the follow-up recognition accuracy is affected. For example, mean filtering may blur the edges of the target during smoothing noise, making it difficult to accurately extract its shape features. Meanwhile, large-area clutter, such as reflected light, light spots and the like, can be generated under the strong light background, and the clutter has similarity with a target in terms of color, texture and the like and is easy to confuse. Edge detection-based methods may misjudge the edges of clutter as target edges, thereby creating a large number of false detection areas. For the machine learning method, the complexity of the sample is increased due to the presence of clutter, so that the model is difficult to accurately learn the characteristic mode of the target, and the classification accuracy and reliability are reduced. For example, in training using SVM or CNN, clutter regions may be wrongly learned as features of the target, resulting in a misclassified situation in the actual recognition. Key features of the target such as specific shapes, color marks, textures, etc. are difficult to extract accurately due to noise and clutter. Under strong light irradiation, the color of the target may be distorted, the original clear texture may be covered by noise, so that the recognition method based on the color and texture features is invalid, and the target may be partially shielded, deformed and the like under the strong light background, so that the difficulty of feature extraction is further increased, and stable and distinguishable features are difficult to find to perform effective target recognition. Therefore, in order to overcome the defects of the prior art in noise and clutter processing under the strong light background, it is urgent and necessary to find a passive reflection type target recognition method and a passive reflection type target recognition system based on linearly polarized light so as to realize accurate recognition of the target position and gesture.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a passive reflection type target recognition method based on linearly polarized light. The method comprises the steps of collecting a plurality of target images, carrying out preprocessing operation on the target images to obtain target feature images, obtaining target edge images by means of edge detection and circle fitting on the target feature images, carrying out extraction and coding of the passive reflection type targets on the target edge images, and providing an identification system based on the passive reflection type targets, wherein the identification system comprises the passive reflection type targets, a stroboscopic incandescent lamp, a color camera and a data processing unit. According to the invention, the target light source is actively strobed by a high-frequency asynchronous signal coding technology, a passive reflection target is designed based on an optical reflection principle, strong light and electromagnetic interference are avoided, a measurement result is stable and reliable, the design requirements of miniaturization and passivity are met, and the method has a relatively strong practical application value.
The invention provides a passive reflection type target recognition method based on linearly polarized light, which comprises the following steps:
s1, collecting a plurality of target images, namely collecting a plurality of target images which are irradiated by a stroboscopic incandescent lamp to enable a passive reflection type target to be stroboscopic passively by using a color camera;
S2, preprocessing the target image to obtain a target characteristic image;
s21, converting the target image into a target gray image by a gray conversion formula by means of a weighted average method, wherein the gray conversion formula is as follows:
G=0.299R+0.587G+0.114B (1)
R, G, B represent the pixel value of the corresponding position of red channel, green channel and blue channel in the said target image separately;
s22, converting the target gray level image into a target denoising image by means of self-adaptive median filtering;
s23, acquiring a target characteristic image with the background removed from the target denoising image at the adjacent moment by a frame difference method;
s3, aiming at the target characteristic image, obtaining a target edge image by means of edge detection and circle fitting;
s31, extracting the edge and detail information of the passive reflection type target in the target characteristic image by means of a gradient-based edge detection algorithm to obtain a target edge image;
S32, determining the center coordinates and the radius of the passive reflection target in the target edge image by means of a circle fitting equation;
S4, extracting and encoding the passive reflection type target aiming at the target edge image;
S41, acquiring the target image only containing the passive reflection type target area based on the target edge image and the target image and taking the target image as a target image to be coded;
S42, identifying the color of the passive reflection type target in the target image to be encoded and taking the color as a first encoding basic element, wherein the first encoding basic element related to the color of the passive reflection type target in the target image to be encoded is extracted, and comprises a main tone, a key color occupation and a color distribution characteristic;
S43, recording time domain information of the passive reflection target in the target image to be encoded, and performing quantization processing, wherein the time domain information is used as a second coding basic element, and the second coding basic element comprises a time point, duration and a position in a time sequence;
s44, based on the first coding basic element and the second coding basic element, according to preset coding logic, combining to generate a unique code special for the passive reflection target as a passive reflection target code.
Further, the step S1 specifically includes the following steps:
s11, emitting two beams of overlapped and twinkled transverse polarized light and longitudinal polarized light by a stroboscopic incandescent lamp, irradiating the transverse polarized light and the longitudinal polarized light to the surface of a target to be detected, and exciting passive reflection type target passive stroboscopic light on the surface of the target to be detected;
s12, shooting the passive reflection type target area on the surface of the target to be detected by using a color camera, and collecting a plurality of target images.
Preferably, in the step S44, a predefined color code table is adopted for the main tone coding in the first coding base element, red, blue and yellow in the predefined color code table are respectively coded into 01, 02 and 03, the key color duty ratio coding is quantized into a [0,1] interval value and then Huffman compressed by Huffman coding, the color distribution characteristics are coded in a simplified topological graph, and the connectivity and the position relationship of the color region are marked to generate a color coding string.
Preferably, the window size of the adaptive median filter in the step S22 is dynamically adjusted based on noise intensity, and the window size ranges from 3×3 to 9×9 pixels.
Preferably, after the passive reflection target is encoded in step S44, the passive reflection target code is stored in a distributed database in a structured data format, and is transmitted to a monitoring center and associated equipment in real time by using a 422 protocol.
Preferably, the frequency of the photographing of the color camera in the step S1 and the flicker frequency of the strobe incandescent lamp are consistent and synchronous, and the field of view of the photographing of the color camera covers all passive reflective targets.
The invention provides a recognition system based on a linear polarized light passive reflection target recognition method, which comprises a passive reflection target, a stroboscopic incandescent lamp, a color camera and a data processing unit, wherein the passive reflection target comprises a first red reflection target, a second red reflection target, a first blue reflection target, a second blue reflection target, a first yellow reflection target and a second yellow reflection target, the first red reflection target, the second red reflection target, the first blue reflection target, the second blue reflection target, the first yellow reflection target and the second yellow reflection target are respectively glued to the surface of a target to be detected, the stroboscopic incandescent lamp is provided with 2 stroboscopic incandescent lamps which are alternately flashed, 1 polarizing plate is arranged in front of each stroboscopic incandescent lamp and respectively emit two beams of overlapped and flashing transverse polarized light and longitudinal polarized light through the polarizing plates, the transverse polarized light and the longitudinal polarized light are irradiated to the surface of the target to be detected and excite the passive reflection target to be passive, the color camera is used for collecting images of the area where the passive reflection target is located, the images are input to the processing unit, and the stroboscopic incandescent lamp is provided with 1 polarizing plate and the polarizing plate to emit two beams to the surface of the target to be detected, and the passive reflection target data processing unit to encode the target images.
Preferably, the first red light reflecting target and the second red light reflecting target comprise 1 red light reflecting film and 1 polaroid, the first blue light reflecting target and the second blue light reflecting target comprise 1 blue light reflecting film and 1 polaroid, the first yellow light reflecting target and the second yellow light reflecting target comprise 1 yellow light reflecting film and 1 polaroid, the polarization directions corresponding to the first red light reflecting target and the second red light reflecting target are perpendicular to each other, the polarization directions corresponding to the first blue light reflecting target and the second blue light reflecting target are perpendicular to each other, and the polarization directions corresponding to the first yellow light reflecting target and the second yellow light reflecting target are perpendicular to each other.
Preferably, the first red light reflecting target, the second red light reflecting target, the first blue light reflecting target, the second blue light reflecting target, the first yellow light reflecting target and the second yellow light reflecting target are all round.
Preferably, the red reflective film, the blue reflective film and the yellow reflective film are made of flexible reflective material reflective films, and the day visible light reflectivity and the night visible light reflectivity are respectively greater than or equal to 60% and 65%.
Compared with the prior art, the invention has the technical effects that:
1. According to the passive reflection type target recognition method based on linearly polarized light, which is designed by the invention, the target light source is actively stroboscopic through a high-frequency asynchronous signal coding technology, a target image is separated and recognized according to the characteristics of a background, a target and a noise point by utilizing a frame difference method, the passive reflection type target is designed based on an optical reflection principle, and light rays are reflected by a special reflection surface, so that the accurate recognition of the position and the gesture of the target is realized, and the practicability is higher.
2. According to the passive reflection type target recognition method based on linearly polarized light, when the environment is a complex air strong light interference environment, the method does not need to modulate an energy output process, the complexity and cost of a system are reduced, the method is not influenced by factors such as electromagnetic interference and the like, the method is suitable for various application scenes including severe weather such as atmospheric turbulence or rain and fog, the measurement result is stable and reliable, and the requirements of miniaturization and passive design are met.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the detailed description of non-limiting embodiments made with reference to the following drawings.
FIG. 1 is a flow chart of a passive reflection target recognition method based on linearly polarized light of the present invention;
FIG. 2 is a schematic illustration of the principle of reflection of polarized light in accordance with the present invention;
FIG. 3 is a schematic representation of edge extraction of a target image in one embodiment of the invention;
fig. 4 is a schematic diagram of a passive-reflex target arrangement in one particular embodiment of the present invention.
Detailed Description
The application is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the application and are not limiting of the application. It should be noted that, for convenience of description, only the portions related to the present application are shown in the drawings. It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other. The application will be described in detail below with reference to the drawings in connection with embodiments.
Fig. 1 shows a linearly polarized light based passive reflection target recognition method of the present invention, comprising the steps of:
S1, collecting a plurality of target images, namely collecting a plurality of target images which are irradiated by the stroboscopic incandescent lamp to enable the passive reflection type target to be stroboscopic passively by using the color camera.
S11, emitting two beams of transverse polarized light and longitudinal polarized light which are overlapped and flash through a stroboscopic incandescent lamp, irradiating the transverse polarized light and the longitudinal polarized light to the surface of the target to be detected, and exciting passive reflection type target passive stroboscopic effect on the surface of the target to be detected, wherein as shown in fig. 2, the polarized light irradiates the surface of a target reflection film to reflect back, the target passively flashes at a specific frequency, and background stray light is constantly bright or irregularly changed so as to cut out noise.
S12, shooting a passive reflection type target area on the surface of the target to be detected by using a color camera, and collecting a plurality of target images. The frequency of the color camera shots and the flicker frequency of the stroboscopic incandescent lamp are kept consistent and synchronous, and the field of view of the color camera shots covers all passive reflective targets.
And an optical imaging device with high resolution and wide dynamic range, such as an industrial CCD high-speed camera, is used for acquiring images of a scene containing the passive reflection target from a specific observation angle, so that the target area is ensured to be completely included in the field of view. The acquisition frame rate is flexibly set according to the dynamic characteristics of the target and the real-time requirements of the application scene, and is generally between 100 frames per second and 200 frames per second.
S2, preprocessing the target image to obtain a target characteristic image.
S21, converting the target image into a target gray image by a gray conversion formula by means of a weighted average method, so that the data dimension is reduced, and the subsequent processing efficiency is improved, wherein the gray conversion formula is as follows:
G=0.299R+0.587G+0.114B (1)
Wherein G represents gray pixel values, R, G, B respectively represent pixel values of corresponding positions of a red channel, a green channel and a blue channel in the target image.
S22, converting the target gray level image into a target denoising image by means of self-adaptive median filtering, accurately retaining target edge and detail information, and avoiding excessive smoothing, wherein the window size of the self-adaptive median filtering is dynamically adjusted based on noise intensity, and the window size ranges from 3X 3 to 9X 9 pixels.
S23, acquiring target characteristic images with background removed from target denoising images at adjacent moments through a frame difference method.
S3, as shown in FIG. 3, aiming at the target characteristic image, obtaining a target edge image by means of edge detection and circle fitting.
S31, extracting the edge and detail information of the passive reflection type target in the target characteristic image by means of a gradient-based edge detection algorithm to obtain a target edge image, wherein the error is controlled within +/-0.1 mm.
S32, determining the center coordinates and the radius of the passive reflection type target in the target edge image by means of a circle fitting equation.
S4, extracting and encoding the passive reflection type target aiming at the target edge image.
And S41, acquiring a target image only containing the passive reflection type target area based on the target edge image and the target image, and taking the target image as a target image to be encoded.
And S42, identifying the colors of the passive reflection targets in the target image to be encoded and taking the colors as first encoding basic elements, namely extracting first encoding basic elements related to the colors of the passive reflection targets in the target image to be encoded, wherein the first encoding basic elements comprise main tone, key color occupation and color distribution characteristics.
S43, recording time domain information of the passive reflection type target in the target image to be encoded, and quantifying, wherein the time domain information is used as a second encoding basic element, and the second encoding basic element comprises a time point, duration and positions in a time sequence.
S44, based on the first coding basic element and the second coding basic element, according to preset coding logic, combining to generate a unique code special for the passive reflection target as a passive reflection target code.
The method comprises the steps of adopting a predefined color code table for main tone coding in a first coding basic element, respectively coding red, blue and yellow in the predefined color code table into 01, 02 and 03, adopting key color duty ratio coding to be quantized into [0,1] interval values, then carrying out Huffman coding and Huffman compression, and marking connectivity and position relation of color areas by color distribution characteristics to simplify topological graph coding, so as to generate a color coding string.
After the passive reflection type target codes are encoded, the passive reflection type target codes are stored in a distributed database in a structured data format, high availability and expansibility of the data are guaranteed, the data are transmitted to a monitoring center and associated equipment in real time by using a 422 protocol, transmission delay is controlled within 20ms, and application scenes with strict real-time requirements are met.
In another aspect, the invention provides an identification system utilizing the linear polarized light-based passive reflection type target identification method, which comprises a passive reflection type target, a stroboscopic incandescent lamp, a color camera and a data processing unit, wherein the passive reflection type target comprises a first red reflection type target, a second red reflection type target, a first blue reflection type target, a second blue reflection type target, a first yellow reflection type target and a second yellow reflection type target, the first red reflection type target, the second red reflection type target, the first blue reflection type target, the second blue reflection type target, the first yellow reflection type target and the second yellow reflection type target are glued to the surface of a target to be detected respectively, as shown in fig. 4, the first red reflection type target and the second red reflection type target comprise 1 red reflection film and 1 polaroid, the first blue reflection type target and the second blue reflection type target comprise 1 blue reflection film and 1 polaroid, the first red reflection type target and the second yellow reflection type target are perpendicular to each other, the first red reflection type target and the second red reflection type target, the first blue reflection type target and the second yellow reflection type target, the first red reflection type target and the second red reflection type target are perpendicular to each other, the first blue reflection type target and the second red reflection type target are perpendicular to each to the target.
The first red light reflecting target, the second red light reflecting target, the first blue light reflecting target, the second blue light reflecting target, the first yellow light reflecting target and the second yellow light reflecting target are all round, the red light reflecting film, the blue light reflecting film and the yellow light reflecting film are all made of flexible light reflecting materials, and the day visible light reflectivity and the night visible light reflectivity are respectively larger than or equal to 60% and 65%.
In a specific embodiment, the target is mainly designed by integrating a flexible reflecting material and a polaroid, the accumulated thickness of the component parts is not more than 1mm, the size is 25mm in diameter, and the high-transmittance transverse and longitudinal polaroid, red, yellow and Lan Sanse diamond-grade reflecting films are adopted.
The stroboscopic incandescent lamps are provided with 2 stroboscopic incandescent lamps and alternately flash, each stroboscopic incandescent lamp is provided with 1 polaroid in front and respectively emits two beams of transverse polarized light and longitudinal polarized light which overlap and flash through the polaroids, the transverse polarized light and the longitudinal polarized light irradiate the surface of a target to be detected and excite passive reflection type target passive stroboscopic, a color camera collects target images of areas where the passive reflection type targets are located and inputs a plurality of target images into a data processing unit, and the data processing unit identifies and codes the passive reflection type targets in the target images.
If the polarization plate and the stroboscopic incandescent lamp have the same polarized light direction, the target is lightened. If the direction of polarized light of the stroboscopic incandescent lamp is perpendicular, all light cannot be transmitted, the light becomes dark, three red and blue Huang Babiao are bright at the same time, three red and blue Huang Babiao are dark, and target serial numbers are distinguished through three primary colors. Time 0 was defined, no. ①③⑤ target was light, no. ②④⑥ target was dark, and time 1 was defined, no. ①③⑤ target was dark, no. ②④⑥ target was light. As shown in table 1, the first bit in the code represents either light (0) or dark (1), and the second two bits represent color, where 00 is red, 01 is blue, and 10 is yellow.
TABLE 1
The passive reflection type target recognition method based on linearly polarized light, which is designed by the invention, enables a target light source to actively strobe through a high-frequency asynchronous signal coding technology, separates and recognizes target images according to the characteristics of the background, the target and noise points by utilizing a frame difference method, designs a passive reflection type target based on an optical reflection principle, reflects light rays through a special reflection surface so as to realize accurate recognition of the position and the gesture of the target, has stronger practicability, does not need to modulate an energy output process when the environment is a complex air strong light interference environment, reduces the complexity and the cost of a system, is not influenced by factors such as electromagnetic interference, is suitable for various application scenes including severe weather such as atmospheric turbulence or rain and fog, has stable and reliable measurement results, and meets the requirements of miniaturization and passive design.
It should finally be noted that the above-mentioned embodiments illustrate rather than limit the technical solution of the present invention, and although the invention has been described in detail with reference to the above-mentioned embodiments, it should be understood by those skilled in the art that the present invention may be modified or equivalently replaced without departing from the spirit and scope of the present invention.