Disclosure of Invention
The application provides an image data management method and system based on an engineering supervision platform, which are used for solving the problem that the accuracy and efficiency are difficult to balance in the automatic marking process of image data.
In a first aspect, the present application provides an image data management method based on an engineering supervision platform, applied to an image data management system, the method comprising:
collecting original image data of a construction site through a monitoring camera and cleaning the data to obtain preliminary image data;
Determining a target deep learning model from a plurality of pre-trained deep learning models with different image recognition granularities according to the accuracy level requirements and the efficiency level requirements of the model detection;
inputting the preliminary image data into the target deep learning model for target detection, and outputting a recognition result, wherein the recognition result comprises a target object and characteristic data of the target object;
checking and checking the characteristic data of the target object based on a preset rule engine to obtain a checking result;
and generating a final marking result of the original image data according to the verification result.
According to the embodiment, the image data management system collects the original image data of the construction site, performs data cleaning, selects the target deep learning model based on the accuracy level requirement and the efficiency level requirement of the model detection by the user to perform target detection and verification on the recognition result, reduces the influence of unilateral pursuit of model recognition accuracy or efficiency on model recognition, and realizes automatic balancing of the accuracy and the efficiency of model recognition in the automatic marking process of the image data according to the user requirement.
In some embodiments, before the step of determining the target deep learning model from the pre-trained deep learning models of a plurality of different image recognition granularities, the accuracy level requirements and the efficiency level requirements of the model-dependent detection further comprises:
collecting sample image data at different construction stages and marking to obtain a training data set;
model training is carried out under different network levels according to the training data set, the pre-trained deep learning model under different image recognition granularities is obtained, and the network levels corresponding to different deep learning models are different;
And dividing the accuracy grade and the efficiency grade respectively corresponding to the deep learning model under different image recognition granularities according to the test result of the deep learning model on the test set.
Through the above embodiment, the image data management system collects sample image data at different stages of construction to construct a training dataset. And training the deep learning model under different network layers to obtain the pre-training deep learning model under different image recognition granularities. These deep learning models are then classified into different levels of accuracy and efficiency based on their test results on the test set. According to the method, the deep learning models with different performances under different image recognition granularities can be obtained, more choices are provided for the subsequent flexible selection of the proper target model according to the requirements, and the adaptability and the practicability of the whole image data treatment process are further improved.
In some embodiments, the training data set is used for model training under different network levels to obtain the pre-trained deep learning model under different image recognition granularities, which specifically includes:
constructing an initial model of the deep learning model based on a convolutional neural network, a regional proposal network and a characteristic pyramid network;
training the initial model by using the training data set, and determining model parameters corresponding to different image recognition granularities;
The pre-trained deep learning model at different image recognition granularities is determined from the model parameters.
Through the embodiment, the image data management system can improve the target detection capability and the accuracy of the pre-training model by combining various deep learning network architectures and utilizing the training data set to optimize the model parameters. This provides a more reliable and efficient basis for subsequent model selection, helping to further increase the overall level of image data governance.
In some embodiments, before the step of determining the target deep learning model from the pre-trained deep learning models of a plurality of different image recognition granularities, the accuracy level requirements and the efficiency level requirements of the model-dependent detection further comprises:
acquiring a construction stage corresponding to a current project;
And matching marked key areas from the database according to the construction stage, and the accuracy grade requirement and the efficiency grade requirement corresponding to the key areas.
Through the embodiment, the image data management system matches corresponding marking key areas and corresponding accuracy and efficiency requirements according to the current construction stage of the project. The method can more accurately determine the emphasis points of image recognition and data management according to the characteristics of different construction stages, and avoid the problems that the processing efficiency is influenced due to overhigh requirements or the recognition accuracy is influenced due to overlow requirements. Meanwhile, the construction stage is matched with the accuracy and efficiency requirements, so that the whole image data treatment process is more intelligent and adaptive, the manual intervention is reduced, and the engineering supervision efficiency and reliability are improved.
In some embodiments, before the step of determining the target deep learning model from the pre-trained deep learning models of a plurality of different image recognition granularities, the accuracy level requirements and the efficiency level requirements of the model-dependent detection further comprises:
Acquiring text description information uploaded by a user;
analyzing the text description information by using a natural language processing model, and determining a marked key area;
and matching the corresponding accuracy grade requirement and efficiency grade requirement from the database according to the key region.
Through the embodiment, the image data management system performs natural language analysis processing on the text description information uploaded by the user to determine the marking key area, so as to match corresponding accuracy and efficiency requirements. The emphasis point of image data management can be more flexibly specified, and the personalized requirements of different users or projects can be met. Meanwhile, the user description is automatically analyzed by using a natural language processing technology, so that the workload of the user is reduced, and the man-machine interaction is more efficient and intelligent. The image recognition requirement is combined with the language description of the user, so that the engineering supervision system can better understand and adapt to the actual situation of the site.
In some embodiments, after the step of inputting the preliminary image data into the target deep learning model for target detection and outputting the recognition result, the method further comprises:
estimating the motion state of pixel points between continuous frames in the preliminary image data by using an optical flow algorithm to obtain optical flow information;
tracking the moving track of each target object in the video frame according to the optical flow information.
Through the embodiment, the image data management system estimates the motion state of the pixel points between the continuous frames in the preliminary image data by using a light flow algorithm, and tracks the moving track of each target object in the video frame. The method can further realize tracking and analysis of the dynamic target on the basis of static target identification, obtain more comprehensive and accurate target state information, provide reliable basis for safety and quality control, and greatly improve the automation level of engineering supervision.
In some embodiments, the step of acquiring the original image data of the construction site by the monitoring camera and performing data cleaning to obtain the preliminary image data specifically includes:
collecting original image data of a construction site through a monitoring camera;
Acquiring parameter information corresponding to the original image data, wherein the parameter information is picture frame parameters influencing target detection of the deep learning model;
and scoring the parameter information corresponding to the original image data based on a preset scoring system, and performing data cleaning on the original image data with the score higher than a preset threshold value to obtain preliminary image data.
Through the embodiment, the image data management system analyzes the parameter information of the original image data collected at the construction site and scores the parameter information of the original image data based on the preset scoring system, so that the automatic screening high-quality original image data with the score higher than the preset threshold value is screened out for data cleaning, and the preliminary image data is obtained. Low-quality or useless data can be filtered, the calculation cost of subsequent processing is reduced, and the overall processing efficiency is improved.
In a second aspect, the present application provides an image data abatement system comprising one or more processors and memory;
The memory is coupled to the one or more processors, and the memory is configured to store computer program code, where the computer program code includes computer instructions that the one or more processors call to enable the image data management system to implement an image data management method based on the engineering supervision platform provided in the foregoing embodiments, which is not described herein.
In a third aspect, the present application provides a computer readable storage medium, including instructions, which when executed on an image data management system, enable the image data management system to implement an image data management method based on an engineering supervision platform provided in the foregoing embodiment, which is not described herein.
In a fourth aspect, the present application provides a computer program product, which when executed on an image data management system, enables the image data management system to implement an image data management method based on an engineering supervision platform provided in the foregoing embodiment, which is not described herein again.
One or more technical solutions provided in the embodiments of the present application at least have the following technical effects or advantages:
1. And the image data management system automatically selects an optimal model from the pre-trained deep learning models with different granularities according to the accuracy and efficiency requirements of the user on model detection to perform target detection and identification. The mechanism can balance accuracy and efficiency under different application scenes, and improves adaptability and practicability of the system. Meanwhile, the model selection strategy is dynamically adjusted in different construction stages and under the demands of users, so that the intelligent and self-adaptive capacity of the system is further enhanced, and the manual intervention is reduced.
2. The image data management system collects sample data at different stages of construction, builds a multi-stage training data set, trains a deep learning model at different network levels, and obtains pre-training models with different granularities. By combining multiple deep learning architectures and model parameter optimization, the detection capability and the accuracy of the pre-training model are improved.
3. The image data management system comprehensively utilizes various information to analyze, including image data parameters, user text description, target motion trail and the like. The method comprises the steps of selecting image parameters, grading and screening, optimizing a data cleaning process, analyzing user description by using natural language, realizing flexible and intelligent man-machine interaction, tracking target movement by combining an optical flow algorithm, and expanding dynamic analysis capability of a system. The fusion analysis of the multi-mode information enables the system to comprehensively understand and adapt to complex engineering supervision scenes, provides reliable analysis basis, and improves automation and intelligent level of engineering supervision.
Detailed Description
The terminology used in the following embodiments of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the specification of the present application and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms as well, unless the context clearly indicates to the contrary. It should also be understood that the term "and/or" as used in this disclosure is intended to encompass any or all possible combinations of one or more of the listed items.
The terms "first," "second," and the like, are used below for descriptive purposes only and are not to be construed as implying or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature, and in the description of embodiments of the application, unless otherwise indicated, the meaning of "a plurality" is two or more.
For ease of understanding, the method provided in this embodiment is described in the following. Fig. 1 is a schematic flow chart of an image data management method based on an engineering supervision platform according to an embodiment of the application.
S101, acquiring original image data of a construction site through a monitoring camera and cleaning the data to obtain preliminary image data.
The image data management system continuously collects the original image data of the construction site through a monitoring camera deployed at the construction site. These raw image data may include still pictures and dynamic video, recording real-time status of various areas of the job site, for different time periods.
After the original image data is acquired, in order to ensure the efficiency and accuracy of subsequent processing and analysis, the image data management system performs data cleaning on the acquired original image data, so that the image data with poor quality or irrelevant to engineering supervision tasks is filtered out, and a relatively standard, accurate and available primary image data set is obtained.
S102, determining a target deep learning model from a plurality of pre-trained deep learning models with different image recognition granularities according to accuracy level requirements and efficiency level requirements of model detection.
Considering the complex variability of engineering supervision scenes and different emphasis on recognition accuracy and processing efficiency in practical application, a single deep learning model often has difficulty in meeting all requirements. Therefore, the key of the step is to flexibly select the most suitable target detection model according to the current task demand.
Specifically, the image data management system prepares in advance a plurality of pre-trained deep learning models, each model differing in granularity of image recognition (i.e., the degree of refinement in characterizing the target class and detail). The model with higher granularity can identify more types of targets and give more detailed description, the accuracy is high, but the calculation amount is usually larger, and the model with lower granularity has lower accuracy but higher processing speed although the detail description is coarser.
Further, the image data management system respectively classifies the accuracy and the efficiency of each model, such as high (the accuracy exceeds 98% or the efficiency exceeds 100 frames/second), and the like, so as to obtain a model-accuracy efficiency relation table.
When a user selects a deep learning model, the image data management system matches a corresponding target deep learning model or a closest target deep learning model from a model-accuracy efficiency relation table according to accuracy grade requirements and efficiency grade requirements input by the user.
S103, inputting the preliminary image data into a target deep learning model for target detection, and outputting a recognition result.
Specifically, the target deep learning model performs comprehensive scanning and analysis on each input image. The multi-level features of the image are extracted by means of the convolutional neural network, and the target area in the image is gradually found out through a series of operations such as area candidate, bounding box regression, classification and discrimination. This may include various types of engineering supervision related identification targets for workers, machinery, building components, safety protection facilities, etc.
S104, checking and checking the characteristic data of the target object based on a preset rule engine to obtain a checking result.
After the preliminary recognition results output by the deep learning model are obtained, the image data management system further carries out auditing and verification on the results according to a preset rule engine to obtain verification results so as to ensure the accuracy and usability of the verification results.
The rule engine is a decision system based on expert knowledge and experience, and comprises various specifications, standards, regulations and the like in the field of engineering supervision. These rules are expressed in terms of condition-action, for example, "IF workers work AND at a height of 2 meters or more without a helmet THEN to determine a violation AND pre-warn". The system substitutes the identified target characteristic data into the rules, automatically judges whether the target characteristic data meets the conditions, and makes corresponding check decisions.
For example, in one embodiment, the field monitoring captures an electrode welder in construction. Through target detection, the image data management system recognizes that the worker is not wearing a protective mask, and no obvious isolation fence is arranged in the electric welding area. The built-in safety production rule of the image data management system immediately inspects the situation, AND finds that conditions such as 'IF welding operation AND wearing no protective mask THEN for judging serious violations' are met, so that a 'serious violation' verification result is generated AND an early warning prompt is sent out.
S105, generating a final marking result of the original image data according to the verification result.
Through the auditing and verification of the rule engine, the image data management system has determined the compliance state and risk hidden danger information of the target object identified in each original image. To further refine and clarify this information, the system generates a clear, canonical annotation on the original image, forming the final marking result.
Specifically, the image data management system adds text labels near the boundary boxes of each identified target object, and expresses key attributes such as category, state and hidden danger of the target object by using compact semantics. The style and content of these tags are typically uniformly planned, for example, with green borders and "normal" tags for compliance targets, red borders and specific types of violations ("unworn helmets", "aloft work") for violations targets, orange borders and risk levels ("severe", "larger") for risk hazards, etc.
In the above embodiment, the image data management system collects the original image data of the construction site, performs data cleaning, and then selects the target deep learning model to perform target detection and verification on the basis of the accuracy level requirement and the efficiency level requirement of the user on the model detection, thereby reducing the influence of unilateral pursuit of model recognition accuracy or efficiency on model recognition, and realizing automatic balancing of the accuracy and efficiency in the automatic image data marking process according to the user requirement.
The method provided in this embodiment will be described in more detail. Fig. 2 is a schematic flow chart of an image data management method based on an engineering supervision platform according to an embodiment of the application.
S201, acquiring original image data of a construction site through a monitoring camera and acquiring parameter information corresponding to the original image data.
The image data management system acquires original image data in real time through monitoring cameras deployed in various areas of a construction site. The cameras can generally cover key areas of construction sites, such as large equipment operation areas, high-altitude operation areas, edge hole areas and the like, so that personnel behaviors and material equipment states in the construction process can be recorded in an omnibearing and multi-angle mode. The original image data collected by the camera includes video stream and periodically snap-shot pictures, and the data format can be common mp4, avi, jpg and the like, which is not limited herein.
And acquiring the parameter information corresponding to each frame of image by the image data management system while acquiring the original image data. These parameters typically include the time of capture of the image, camera number, image resolution, image brightness, contrast, etc., which affect the quality of the image and the effect of subsequent analysis. By acquiring these parameters, the image data management system may initially evaluate the availability of raw image data, ready for subsequent data cleaning.
S202, scoring the parameter information corresponding to the original image data based on a preset scoring system, and performing data cleaning on the original image data with the score higher than a preset threshold value to obtain preliminary image data.
After the original image data and parameters thereof are obtained, the image data management system further identifies the quality of each frame of image. Specifically, the image data management system is internally provided with a set of preset image quality scoring system, and a score ranging from 0 to 100 is provided mainly according to indexes such as contrast, sharpness, color fidelity, undistorted degree and the like of the image. Optionally, determining the image with the score of more than 80 as qualified high-quality data, and obtaining preliminary image data after data cleaning. And the image with the score of more than 80 is directly judged to be unqualified, and is discarded or directly submitted to manual processing.
S203, collecting sample image data at different construction stages and marking to obtain a training data set.
This step is used to construct a training dataset of the pre-trained deep learning model used in step S102. Specifically, the image data management system acquires representative pictures from different construction stages, such as a supporting structure image of a foundation pit excavation stage, a template engineering image of a main body construction stage, an adjacent line image of a decoration stage, and the like. The pictures cover aspects of each construction scene as much as possible, and contain enough target categories and postures. After a sufficient number of pictures are acquired, key targets in the pictures are marked in an image data management system by organization professionals, an accurate boundary box is drawn for each target, attribute information such as category, position and state is marked, and after marking is finished, the pictures and marking information thereof are arranged into a standard training data set by the image data management system.
S204, constructing an initial model of the deep learning model based on the convolutional neural network, the regional proposal network and the characteristic pyramid network.
When the training data set is ready, the image data management system enters a construction stage of the deep learning model. Specifically, the image data management system uses a Convolutional Neural Network (CNN) as a backbone network for extracting multi-level and multi-scale features in an image, and can be better adapted to complex and changeable construction environment images through further fine adjustment and optimization. Next, based on the CNN, the image data management system introduces a Region Proposal Network (RPN) for extracting candidate regions that may contain targets from the image feature map. The RPN slides on the feature map by using a group of rectangular frames with fixed size and aspect ratio, and the target area with higher probability is rapidly screened out by binary classification and bounding box regression, so that the calculation amount of subsequent detection is greatly reduced. In addition, the image data management system can also add a Feature Pyramid Network (FPN) on the basis of the CNN. The FPN builds a feature pyramid on the multi-scale feature map extracted by the CNN, and fuses the high-level semantic features and the bottom-level detail features to generate a series of feature maps with different resolutions. In the detection process, targets with different sizes can be effectively covered on feature maps with different levels, so that self-adaptive detection is realized.
S205, training an initial model by using a training data set, and determining model parameters corresponding to different image recognition granularities to obtain a plurality of pre-trained deep learning models.
Specifically, the image data management system acquires the corresponding relation between different parameter combinations and model performances through parameter tests and statistical analysis such as the number of layers of a test adjusting network, the size of convolution kernels, the number of channels of a feature map, the number of candidate areas, a non-maximum suppression threshold value and the like. For example, increasing the number of network layers and the number of feature channels can increase the feature extraction and classification capabilities of the model, but also significantly increase the computation time, while decreasing the number of candidate regions and increasing the NMS threshold can increase the target screening speed, but may miss some small targets, etc. Of course, the corresponding relation between the different parameter combinations and the model performance can also be directly uploaded by the related technicians to the image data management system, which is not limited herein.
Then, the image data management system pertinently generates pre-training models under different image recognition granularities according to the actual requirements of engineering supervision. For some occasions with strict requirements on efficiency, if highway vehicle monitoring of each frame of picture is required to be detected in real time, the system can properly reduce the network scale, cut out some redundant characteristic channels, and maximize the detection speed at the cost of a certain accuracy. For some occasions where accuracy is critical, such as detail quality detection of critical large-scale engineering, the system can adopt the maximum-scale network parameters, and the aim is to realize extremely high defect identification accuracy at the cost of long processing time delay.
In the above embodiment, the image data management system may improve the target detection capability and accuracy of the pre-training model by combining multiple deep learning network architectures and performing model parameter optimization using the training data set. This provides a more reliable and efficient basis for subsequent model selection, helping to further increase the overall level of image data governance.
S206, dividing the accuracy level and the efficiency level respectively corresponding to the deep learning model under different image recognition granularities according to the test result of the deep learning model on the test set.
Specifically, for each picture in the test set, the image data management system uses the current deep learning model to be evaluated to perform reasoning and prediction to obtain a group of detection frame results. Then, the predicted detection frame is compared with the manually labeled truth frame, and the intersection ratio is calculated (IoU). If IoU of a certain prediction frame and a truth frame exceeds a preset threshold (e.g., 0.5), the prediction frame is determined to be True Positive (TP), otherwise, the prediction frame is determined to be False Positive (FP). After traversing all the prediction frames, the system can obtain the TP number and the FP number of the current picture. And repeating the process on the whole test set to obtain the accuracy of the model, and grading to obtain the accuracy grade corresponding to the model.
Meanwhile, the image data management system can also count time expenditure of the model in the reasoning process, including time consumption of various stages such as data reading, feature extraction, candidate frame generation, bounding box regression, non-maximum suppression and the like. And (3) overlapping the time consumption of each stage to obtain the average time of processing the single picture by the model, namely the model efficiency, and grading to obtain the accuracy grade corresponding to the model.
In the above embodiment, the image data management system collects sample image data at different stages of construction to construct a training dataset. And training the deep learning model under different network layers to obtain the pre-training deep learning model under different image recognition granularities. These deep learning models are then classified into different levels of accuracy and efficiency based on their test results on the test set. According to the method, the deep learning models with different performances under different image recognition granularities can be obtained, more choices are provided for the subsequent flexible selection of the proper target model according to the requirements, and the adaptability and the practicability of the whole image data treatment process are further improved.
S207, analyzing the text description information uploaded by the user by using a natural language processing model, and determining a marked key area.
Specifically, a user may enter a short text in a system interface of the image data management system, expressing objects and areas that they want to focus on and analyze. These texts may use daily language and industry terminology without requiring strict formatting and grammar. For example, whether a tower crane driver is fastened with a safety belt or not is checked, and a foundation pit earth excavation slope is marked.
When a user submits text, the image data management system uses a pre-trained Chinese word segmentation model, such as jieba, THULAC, etc., to segment the text into a plurality of semantically complete words. And judging and marking the part of speech of each word, and distinguishing nouns, verbs, adjectives and the like. And mapping the words to a plurality of predefined engineering supervision concepts, such as 'mechanical equipment', 'personnel behaviors', and the like by combining a professional dictionary and a knowledge map in the field of building construction.
After the processing, the image data management system can extract a plurality of key information from the text description of the user, and map the key information to specific parameters in the image understanding and analyzing task. As in the example of "check if the tower crane driver is belted", the system positions the important object of interest as "tower crane driver" and determines the preset range area of the tower crane as the important area.
S208, matching the corresponding accuracy grade requirement and efficiency grade requirement from the database according to the key region.
Specifically, the image data management system maintains a demand-performance correspondence database in the background. The database is based on the real cases accumulated in each engineering project in the past, and summarizes the corresponding relation between the image analysis requirements of different construction areas and links and the actual performance of the adopted algorithm model. The corresponding relations are stored in a standardized rule form, so that the quick searching and matching are convenient.
For example, in one embodiment, there may be the following rules in the database:
IF area= "tower cab" AND target= "driver person" THEN accuracy= "high" AND efficiency= "medium", etc.
In the above embodiment, the image data management system performs natural language analysis processing on the text description information uploaded by the user to determine the marking key area, so as to match the corresponding accuracy and efficiency requirements. The emphasis point of image data management can be more flexibly specified, and the personalized requirements of different users or projects can be met. Meanwhile, the user description is automatically analyzed by using a natural language processing technology, so that the workload of the user is reduced, and the man-machine interaction is more efficient and intelligent. The image recognition requirement is combined with the language description of the user, so that the engineering supervision system can better understand and adapt to the actual situation of the site.
S209, matching marked key areas from the database according to the construction stage corresponding to the current project, and accuracy grade requirements and efficiency grade requirements corresponding to the key areas.
In particular, a related technician may also build a database based on historical project experience summaries on the image data management system. The database abstracts various links of building construction into a plurality of standardized stages, such as 'field leveling', 'foundation and foundation, main structure', and the like. In each construction stage, the database further defines the difficulty and difficulty of image monitoring and analysis and the corresponding model performance requirement, and forms rules for matching. For example, IF phase= "field leveling" THEN emphasis= "large earthmoving machinery" AND accuracy= "medium" AND efficiency= "high", etc.
In the above embodiment, the image data management system matches the corresponding marking key area and the corresponding accuracy and efficiency requirements according to the current construction stage of the project. The method can more accurately determine the emphasis points of image recognition and data management according to the characteristics of different construction stages, and avoid the problems that the processing efficiency is influenced due to overhigh requirements or the recognition accuracy is influenced due to overlow requirements. Meanwhile, the construction stage is matched with the accuracy and efficiency requirements, so that the whole image data treatment process is more intelligent and adaptive, the manual intervention is reduced, and the engineering supervision efficiency and reliability are improved.
S210, determining a target deep learning model from a plurality of pre-trained deep learning models with different image recognition granularities according to the accuracy level requirements and the efficiency level requirements of model detection.
The step is the same as step S102, and will not be described here again.
S211, inputting the preliminary image data into a target deep learning model for target detection, and outputting a recognition result.
This step is the same as step S103 and will not be described here again.
S212, estimating the motion state of pixel points between continuous frames in the preliminary image data by using an optical flow algorithm to obtain optical flow information.
Specifically, the image data management system calculates the corresponding points of two continuous frames of images pixel by pixel based on a sparse optical flow or dense optical flow method, and finds their motion vectors. The set of these vectors forms an optical flow field that reflects the speed and direction of movement of the object in the scene.
Taking the Horn-Schunck optical flow method as an example, the image data management system firstly assumes that the brightness of the pixel points on the surface of the same object between continuous frames is kept unchanged, namely the optical flow field is smooth and continuous. And then solving the velocity component of each pixel by minimizing the global energy function, thereby estimating the displacement of the pixel and obtaining optical flow information.
S213, tracking the moving track of each target object in the video frame according to the optical flow information.
Based on the optical flow field information obtained in step S212, the image data management system further realizes continuous tracking of the target in the video. Specifically, the image data management system first detects individual target objects in a first frame of video and calibrates their initial positions with a bounding box. Next, the system superimposes the optical flow field on the current frame, predicting the approximate location of the object in the next frame based on the motion trend of the pixels within the bounding box. Centering on this, the system searches for the exact bounding box of the target in the neighborhood of the next frame, and repeats the above prediction and correction process, updating the position and trajectory of the target from frame to frame.
In the above embodiment, the image data management system estimates the motion state of the pixel points between the continuous frames in the preliminary image data using the optical flow algorithm, and tracks the moving track of each target object in the video frame. The method can further realize tracking and analysis of the dynamic target on the basis of static target identification, obtain more comprehensive and accurate target state information, provide reliable basis for safety and quality control, and greatly improve the automation level of engineering supervision.
The image data management system according to the embodiment of the present invention is applied to an electronic device, and fig. 3 shows a schematic diagram of an architecture of an electronic device suitable for implementing the embodiment of the present invention.
It should be noted that the electronic device shown in fig. 3 is only an example, and should not impose any limitation on the functions and application scope of the embodiments of the present invention.
Those of ordinary skill in the art will appreciate that all or a portion of the steps of the various methods of the above embodiments may be performed by instructions (computer programs) or by control of associated hardware by instructions (computer programs), which may be stored in a computer-readable storage medium and loaded and executed by a processor. The electronic device of the present embodiment includes a storage medium and a processor, where the storage medium stores a plurality of instructions that can be loaded by the processor to perform any of the steps of the methods provided by the embodiments of the present invention.
In particular, the storage medium and the processor are electrically connected, either directly or indirectly, to enable transmission or interaction of data. For example, the elements may be electrically connected to each other by one or more signal lines. The storage medium has stored therein computer-executable instructions for implementing the data access control method, including at least one software functional module that may be stored in the storage medium in the form of software or firmware, and the processor executes the software programs and modules stored in the storage medium to perform various functional applications and data processing. The storage medium may be, but is not limited to, random Access Memory (RAM), read Only Memory (ROM), programmable read only memory (ProgrammableRead-only memory, PROM), erasable read only memory (ErasableProgrammableRead-only memory, EPROM), electrically erasable read only memory (ElectricErasableProgrammableRead-only memory, EEPROM), etc. The storage medium is used for storing a program, and the processor executes the program after receiving the execution instruction.
Further, the software programs and modules within the storage media described above may also include an operating system, which may include various software components and/or drivers for managing system tasks (e.g., memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide an operating environment for other software components. The processor may be an integrated circuit chip with signal processing capabilities. The processor may be a general-purpose processor, including a Central Processing Unit (CPU), a network processor (NetworkProcessor NP), etc., which may implement or execute the methods, steps, and logic flow diagrams disclosed in the embodiments. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
Because the instructions stored in the storage medium may perform steps in any of the methods provided in the embodiments of the present invention, the beneficial effects of any of the methods provided in the embodiments of the present invention may be achieved, and detailed descriptions of the foregoing embodiments are omitted herein.
The present invention is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.