Traffic image processing method, system, storage medium and computer

Abstract

The present invention provides a traffic image processing method, system, storage medium and computer. The processing method includes: obtaining an original traffic image, preprocessing the original traffic image to reduce the noise of the original traffic image; sparse processing the preprocessed traffic image, selecting pixel points in the sparsely processed traffic image, selecting image blocks with a size of preset pixels with the pixel points as the center, and constructing an image matrix; solving the image matrix for the maximum value based on a compressed sensing algorithm to reconstruct the traffic image according to the image matrix and identify the target object in the traffic image; inputting the reconstructed traffic image into a user interface for display, and providing decision support for users to perform traffic management. The traffic image processing method provided by the present invention can reconstruct traffic images in real time and accurately, and improve the timeliness of traffic data feedback.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic management, and in particular to a traffic image processing method, system, storage medium and computer. Background Art

[0002] Nowadays, intelligent transportation has become an important means to alleviate traffic pressure and improve traffic efficiency. However, the acceleration of urbanization has made traffic problems more and more serious. Therefore, in order to solve such problems, it is necessary to obtain traffic data accurately and in real time and understand traffic information. After collecting traffic information data, it is particularly necessary to process traffic images.

[0003] Image processing technology is a technology that uses computers, cameras and other digital processing technologies to calculate and process images, and can easily extract information from images; existing image processing methods mainly include the following steps: image enhancement, image segmentation, feature extraction, target detection and recognition, background modeling and moving target detection, but the existing methods include at least the following problems in the image processing process: image enhancement may cause image distortion, image segmentation is sensitive to interference factors, which may lead to inaccurate segmentation results, feature extraction and target recognition have high computational complexity, are difficult to process in real time, feedback is delayed, and cannot provide users with timely decision support for traffic management. Summary of the invention

[0004] Based on this, the purpose of the present invention is to provide a traffic image processing method, system, storage medium and computer to solve the technical problems existing in the prior art.

[0005] The present invention provides a traffic image processing method, comprising:

[0006] Acquire an original traffic image, and preprocess the original traffic image to reduce noise of the original traffic image;

[0007] Performing sparse processing on the preprocessed traffic image, selecting pixel points in the sparsely processed traffic image, selecting an image block with a size of preset pixels with the pixel point as the center, and constructing an image matrix;

[0008] Solving the maximum value of the image matrix based on a compressed sensing algorithm to reconstruct the traffic image according to the image matrix and identify the target object in the traffic image;

[0009] The reconstructed traffic image is input into the user interface for display, providing decision support for users in traffic management.

[0010] Optionally, the step of preprocessing the original traffic image includes:

[0011] Performing multi-scale decomposition on the original traffic image by using two-dimensional wavelet transform to obtain a first image signal wavelet coefficient and a first noise signal wavelet coefficient of the original traffic image;

[0012] Comparing the frequency of the wavelet coefficients of the first image signal with a preset frequency;

[0013] If it is less than the preset frequency, reconstructing the first noise signal wavelet coefficients, obtaining and removing the noise signal, and reconstructing the first image signal wavelet coefficients to complete the preprocessing;

[0014] If it is not less than the preset frequency, continue to perform two-dimensional wavelet transform on the first image signal to obtain the second image signal wavelet coefficient and the second noise signal wavelet coefficient, compare the frequency of the second image signal wavelet coefficient with the preset frequency, and so on;

[0015] The wavelet coefficients of the noise signal obtained after several two-dimensional wavelet transforms are reconstructed to obtain and remove the noise signal, and the wavelet coefficients of the image signal obtained after the last two-dimensional wavelet transform are reconstructed to complete the preprocessing.

[0016] Optionally, the step of performing maximum value solving on the image matrix based on a compressed sensing algorithm to reconstruct the traffic image according to the image matrix includes:

[0017] constructing an optimized coefficient matrix based on an overcomplete dictionary, wherein the optimized coefficient matrix is ​​used to represent a coefficient matrix of a reconstructed image;

[0018] Reconstruct and capture the optimization coefficient matrix based on the deep learning model to obtain image quality feedback during the image reconstruction process to capture features of interest in the image;

[0019] Performing total variation processing on the optimization coefficient matrix based on wavelet sparse regularization, wherein the total variation processing is used to improve image quality in the process of reconstructing images;

[0020] The optimized coefficient matrix after reconstruction capture and total variation processing is processed based on the basis pursuit algorithm to construct an objective function for solving the maximum value of the image matrix;

[0021] Iterate the objective function until the difference between the two iteration results of the objective function is less than the reduction threshold or the number of iterations is greater than the maximum number of iterations, stop iterating, and obtain the target optimization coefficient matrix;

[0022] A reconstructed image matrix is ​​obtained according to the target optimization coefficient matrix and the overcomplete dictionary, thereby completing the reconstruction of the traffic image.

[0023] Optionally, the objective function is expressed as:

[0024]

[0025] In the formula, is an m×n image matrix, n is the number of pixels, m is the pixel size of the image block, is an overcomplete dictionary consisting of p basis vectors of size m, p>m, is a p×n optimization coefficient matrix, for The L1 norm of for The L2 norm of , is the regularization parameter, is the result of total variation processing for the optimization coefficient matrix, To adjust the parameters of the loss function weights of the deep learning model, The constraint term for the reconstruction capture process to optimize the coefficient matrix is ​​used to measure the difference between A and the output of the deep learning model;

[0026] in,

[0027]

[0028] In the formula, is the maximum dimension of the gradient operator, represents the t-th dimension gradient operator, Represents wavelet transform.

[0029] Optionally, after constructing the objective function for solving the maximum value of the image matrix, the traffic image processing method further includes:

[0030] Performing change separation on the optimization coefficient matrix to obtain a positive change part and a negative change part of the optimization coefficient matrix;

[0031] After the change separation, the objective function is expressed as:

[0032]

[0033] In the formula, To optimize the coefficient matrix The positive change part of To optimize the coefficient matrix The negative change part.

[0034] Optionally, the iterating the objective function comprises: iterating the optimization coefficient matrix The positive change of , the optimization coefficient matrix The negative change , pixel weights, and deep learning models are iterated;

[0035] in, The expression for iteration is:

[0036]

[0037] The expression for iteration is:

[0038]

[0039] In the formula, For the k+1th iteration The value of For the k+1th iteration The value of For the kth iteration The value of For the kth iteration The value of

[0040] The expression for iterating pixel weight is:

[0041]

[0042] In the formula, is the weight of pixel (i, j) at the k+1th iteration, is the gradient at the kth iteration, is a preset number greater than zero;

[0043] The expression for the iteration of the deep learning model is:

[0044]

[0045] In the formula, is the updated parameter value of the deep learning model, is the parameter value of the deep learning model before updating, are the deep learning model parameters, is the learning rate, for Relative to gradient.

[0046] The present invention also provides a traffic image processing system, comprising:

[0047] A processing module, used for acquiring an original traffic image and preprocessing the original traffic image to reduce noise of the original traffic image;

[0048] A construction module is used to perform sparse processing on the pre-processed traffic image, select pixel points in the sparsely processed traffic image, select image blocks with a size of preset pixels with the pixel points as the center, and construct an image matrix;

[0049] A solution module, used for performing an extremum solution on the image matrix based on a compressed sensing algorithm, so as to reconstruct the traffic image according to the image matrix and identify the target object in the traffic image;

[0050] The input module is used to input the reconstructed traffic image into the user interface for display, providing decision support for users in traffic management.

[0051] The present invention also provides a storage medium on which a computer program is stored, and when the program is executed by a processor, the above-mentioned traffic image processing method is implemented.

[0052] The present invention also provides a computer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned traffic image processing method when executing the computer program.

[0053] Compared with the prior art, the present invention has the following beneficial effects: the traffic image processing method provided by the present application first obtains the original traffic image, and pre-processes the original traffic image to reduce the noise of the original traffic image; before reconstructing the traffic image, the original traffic image is pre-processed to remove part of the noise in the original traffic image to reduce the influence on the image reconstruction; the traffic image obtained after the pre-processing is sparsely represented, pixel points in the traffic image are selected, and image blocks with a size of preset pixels are selected with the pixel points as the center to construct an image matrix; the image matrix is ​​solved for the maximum value based on the compressed sensing algorithm to reconstruct the traffic image according to the image matrix and identify the target object in the traffic image; the image matrix constructed after the sparse transformation is solved for the maximum value based on the compressed sensing algorithm to reconstruct the traffic image according to the image matrix, and the compressed sensing algorithm can effectively restore high-quality images from noisy data, and has the advantages of reducing computational complexity and improving image edge detail smoothing and reconstruction accuracy. Finally, the reconstructed traffic image is input into the user interface to provide decision support for users to perform traffic management, thereby improving the timeliness of traffic data feedback and being suitable for large-scale promotion.

[0054] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a flow chart of a traffic image processing method in a first embodiment of the present invention;

[0056] Figure 2 FIG. 4 is a structural block diagram of a computer in a fourth embodiment of the present invention.

[0057] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0058] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0060] Embodiment 1

[0061] See also Figure 1 , which shows a traffic image processing method in a first embodiment of the present invention, and the traffic image processing method specifically includes steps S10 to S40:

[0062] S10, acquiring an original traffic image, and preprocessing the original traffic image to reduce noise of the original traffic image;

[0063] In specific implementation, the process of traffic image processing mainly includes: first, completing the initialization configuration of the camera to ensure that it can work normally to capture traffic image data in real time; in this embodiment, the camera realizes the splicing of 16-bit pixel data, packs the spliced ​​data into frames, and sends the data collected by the camera to the image preprocessing module at a speed of 30fps per second through a dedicated interface. The image preprocessing module uses wavelet transform to suppress noise more easily and quickly, and transmits the data to compressed sensing after image preprocessing. Compressed sensing uses compression algorithms to compress the preprocessed image data, reduce the bandwidth required for storage and transmission, and extract key features from the compressed data for subsequent status analysis and decision support. Finally, the images collected by the camera and the processing results are displayed in real time on the user interface. When an abnormal event is detected, an alarm prompt is issued to the user through the user interface.

[0064] Optionally, the step of preprocessing the original traffic image includes:

[0065] Performing multi-scale decomposition on the original traffic image by using two-dimensional wavelet transform to obtain a first image signal wavelet coefficient and a first noise signal wavelet coefficient of the original traffic image;

[0066] Comparing the frequency of the wavelet coefficients of the first image signal with a preset frequency;

[0067] If it is less than the preset frequency, reconstructing the first noise signal wavelet coefficients, obtaining and removing the noise signal, and reconstructing the first image signal wavelet coefficients to complete the preprocessing;

[0068] If it is not less than the preset frequency, continue to perform two-dimensional wavelet transform on the first image signal to obtain the second image signal wavelet coefficient and the second noise signal wavelet coefficient, compare the frequency of the second image signal wavelet coefficient with the preset frequency, and so on;

[0069] The wavelet coefficients of the noise signal obtained after several two-dimensional wavelet transforms are reconstructed to obtain and remove the noise signal, and the wavelet coefficients of the image signal obtained after the last two-dimensional wavelet transform are reconstructed to complete the preprocessing.

[0070] The subband frequency can be obtained through two-dimensional wavelet transform. Each time a wavelet transform is performed, the spatial resolution of the image is reduced to half of the original value. After multiple transformations, the efficiency of obtaining noise signals in the image can be improved. The subband frequency of the noise signal in the image is generally high. Therefore, by reconstructing the high-frequency subband coefficients after the wavelet transform, the fault phase after noise interference can be obtained. Furthermore, by selecting a suitable wavelet basis for multi-scale decomposition as needed, the impact of noise interference can be reduced.

[0071] S20, performing sparse processing on the pre-processed traffic image, selecting pixel points in the sparsely processed traffic image, selecting image blocks with a size of preset pixels with the pixel points as the center, and constructing an image matrix;

[0072] First, the traffic image is sparsely represented. In the subsequent compressed sensing processing, the sparseness of the signal is a necessary prerequisite.

[0073] S30, performing an extremum solution on the image matrix based on a compressed sensing algorithm to reconstruct the traffic image according to the image matrix and identify a target object in the traffic image;

[0074] Compressed sensing technology can perform random sampling under conditions far below the Nyquist sampling theorem, and then a reconstruction algorithm can be used to restore the original sampled signal. In compressed sensing, signal reconstruction is the key to signal quality, and the basis pursuit algorithm is a representative algorithm for signal reconstruction. Compressed sensing technology can use a number of linear measurement vectors in a set of unrelated bases to accurately reconstruct the observed signal. In compressed sensing technology, signal reconstruction is particularly important, and it is the process of finding the optimal solution for the observed signal. Optionally, constructing an image matrix is ​​the observation data in the compressed sensing calculation process of this embodiment.

[0075] The step of solving the maximum value of the image matrix based on the compressed sensing algorithm to reconstruct the traffic image according to the image matrix includes:

[0076] constructing an optimized coefficient matrix based on an overcomplete dictionary, wherein the optimized coefficient matrix is ​​used to represent a coefficient matrix of a reconstructed image;

[0077] Reconstruct and capture the optimization coefficient matrix based on the deep learning model to obtain image quality feedback during the image reconstruction process to capture features of interest in the image;

[0078] Performing total variation processing on the optimization coefficient matrix based on wavelet sparse regularization, wherein the total variation processing is used to improve image quality in the process of reconstructing images;

[0079] The optimized coefficient matrix after reconstruction capture and total variation processing is processed based on the basis pursuit algorithm to construct an objective function for solving the maximum value of the image matrix;

[0080] Iterate the objective function until the difference between the two iteration results of the objective function is less than the reduction threshold or the number of iterations is greater than the maximum number of iterations, stop iterating, and obtain the target optimization coefficient matrix;

[0081] A reconstructed image matrix is ​​obtained according to the target optimization coefficient matrix and the overcomplete dictionary, thereby completing the reconstruction of the traffic image.

[0082] The objective function is expressed as:

[0083]

[0084] In the formula, is an m×n image matrix, n is the number of pixels, m is the pixel size of the image block, is an overcomplete dictionary consisting of p basis vectors of size m, p>m, is a p×n optimization coefficient matrix, for The L1 norm of for The L2 norm of , is the regularization parameter, is the result of total variation processing for the optimization coefficient matrix, To adjust the parameters of the loss function weights of the deep learning model, The constraint term for the reconstruction capture process to optimize the coefficient matrix is ​​used to measure the difference between A and the output of the deep learning model;

[0085] in,

[0086]

[0087] In the formula, is the maximum dimension of the gradient operator, represents the t-th dimension gradient operator, Represents wavelet transform.

[0088] In this embodiment, due to the high complexity of the traffic image itself, it may be affected by factors such as noise during the transmission process, making it difficult for traditional image processing technology to effectively process it. Therefore, in this embodiment, based on the basis pursuit algorithm, combined with deep learning and wavelet sparse regularization, the problem is split into multiple sub-problems for solution, which has the advantages of reducing computational complexity, improving image edge detail smoothing and improving reconstruction accuracy.

[0089] The basis pursuit algorithm is very effective in filtering out noise interference in images. The main role of deep learning models in image reconstruction and denoising is to use their powerful nonlinear feature extraction capabilities to learn useful information and patterns from large amounts of data, help suppress noise and improve the signal-to-noise ratio of reconstructed images, provide feedback on the quality of reconstructed images, and capture important features of images; wavelet sparse regularization combines the advantages of TV regularization and wavelet sparse representation, and can improve the quality of reconstructed images; the purpose of performing total variation processing on the wavelet sparse regularization term is to further enhance the spatial continuity of the image while retaining edge information and avoiding artifacts; this embodiment combines the sparse optimization problem with the local detail feature texture structure information of the image, and replaces the defined optimization problem with a quadratic programming problem, so that a clearer detail image can be obtained.

[0090] Optional, when processing multiple traffic images consecutively;

[0091]

[0092] In the formula, For the The optimization coefficient matrix of the image, is the number of deep learning model outputs, is the loss function of the deep learning model, used to represent the reconstructed image With deep model The difference between the outputs.

[0093] After constructing the objective function for solving the maximum value of the image matrix, the traffic image processing method further includes:

[0094] Performing change separation on the optimization coefficient matrix to obtain a positive change part and a negative change part of the optimization coefficient matrix;

[0095] After the change separation, the objective function is expressed as:

[0096]

[0097] In the formula, To optimize the coefficient matrix The positive change part of To optimize the coefficient matrix The negative change part.

[0098] The iterating of the objective function comprises: The positive change of , the optimization coefficient matrix The negative change , pixel weights, and deep learning models are iterated;

[0099] in, The expression for iteration is:

[0100]

[0101] The expression for iteration is:

[0102]

[0103] In the formula, For the k+1th iteration The value of For the k+1th iteration The value of For the kth iteration The value of For the kth iteration The value of

[0104] The expression for iterating pixel weight is:

[0105]

[0106] In the formula, is the weight of pixel (i, j) at the k+1th iteration, is the gradient at the kth iteration, is a preset number greater than zero;

[0107] The expression for the iteration of the deep learning model is:

[0108]

[0109] In the formula, is the updated parameter value of the deep learning model, is the parameter value of the deep learning model before updating, are the deep learning model parameters, is the learning rate, for Relative to gradient.

[0110] In this embodiment, the optimization coefficient matrix Separation, and integration of deep learning and wavelet sparse regularization optimization model, has shown significant advantages in the field of image processing. This method strengthens the sparse expression of the image through wavelet transform and retains the key features. At the same time, with the help of the complex mapping relationship of the deep learning model, it improves the quality of the reconstructed image. After variation separation, the objective function is the total variation Newton smooth model. The application of this model not only speeds up the convergence speed, but also effectively avoids the problem of excessive image smoothing and protects the edge information. In addition, by adjusting the regularization parameters, the algorithm can flexibly balance the relationship between data fidelity and sparsity, as well as deep learning loss; pixel adaptive weights are introduced in the iterative process, and the iterative reweighting algorithm is used to fit the data points with smaller errors more strongly, and reduce the fitting degree of the data points with larger errors, so that the areas with larger errors are focused on and corrected, thereby improving the convergence speed of the algorithm. During the iteration process, , , , You can choose appropriately, the matrix ,matrix The initial value can also be selected as needed.

[0111] In some optional embodiments, pixel weights and gradients can also be used to optimize the coefficient matrix The positive change of and negative change Iterate

[0112] in, The expression for iteration is:

[0113]

[0114] The expression for iteration is:

[0115]

[0116] In the formula, For the k+1th iteration The value of For the k+1th iteration The value of For the kth iteration The value of For the kth iteration The value of , They are and The iterative learning rate, , They are respectively the objective function relative to and The gradient of

[0117]

[0118]

[0119] Using gradient and The iterative update can help dynamically adjust the optimization path, especially when dealing with non-convex optimization problems, which can effectively avoid falling into the local minimum during solution.

[0120] Optionally, in this embodiment, the expression for stopping iteration of the objective function is:

[0121]

[0122] is the objective function value at the k+1th iteration, is the objective function value at the kth iteration, To reduce the threshold, is the maximum number of iterations; the iterative process includes an effective convergence judgment mechanism, which terminates the iterative process in time by monitoring the reduction of the objective function and the number of iterations to improve the computational efficiency.

[0123] In each iteration, u and v are updated by minimizing the objective function O, and A is gradually optimized in the process of updating the values ​​of u and v, thereby achieving the purpose of denoising and reconstructing the image. After that, pass The reconstructed image can be obtained.

[0124] Furthermore, the evaluation index of image quality is a key step in determining whether an image meets the standard. In this embodiment, the mean square error (MSE) is used for judgment: it reflects the difference between the original image and the reconstructed image. Among them, MSE measures the mean of the difference between the corresponding pixels of the two frames of images. As the MSE decreases, the quality of the image becomes closer and closer to the original image, that is, the degree of distortion is lower. The calculation expression of the mean square error is:

[0125]

[0126] In the formula, are the number of rows and columns of pixels of the reconstructed image, is the pixel value of the pixel at the i-th row and j-th column of the preprocessed traffic image, is the pixel value of the pixel at the i-th row and j-th column of the reconstructed traffic image.

[0127] S40, inputting the reconstructed traffic image into a user interface for display, so as to provide decision support for users in traffic management.

[0128] After reconstructing the traffic image, the target object in the image is identified for further analysis; then the image data, processing results, and recognition results are stored through the storage module, and finally the data is uploaded to the cloud monitoring center and user interface. The feedback is timely, which facilitates the intuitive display of traffic image information to users and provides decision-making support for users in traffic management.

[0129] In summary, the traffic image processing method provided by the present application first obtains the original traffic image, and preprocesses the original traffic image to reduce the noise of the original traffic image; before reconstructing the traffic image, the original traffic image is preprocessed to remove part of the noise in the original traffic image to reduce the influence on the image reconstruction; the traffic image obtained after the preprocessing is sparsely represented, the pixel points in the traffic image are selected, and the image blocks with a size of preset pixels are selected with the pixel points as the center to construct an image matrix; the image matrix is ​​solved for the maximum value based on the compressed sensing algorithm to reconstruct the traffic image according to the image matrix and identify the target object in the traffic image; the image matrix constructed after the sparse transformation is solved for the maximum value based on the compressed sensing algorithm to reconstruct the traffic image according to the image matrix. The compressed sensing algorithm can effectively restore high-quality images from noisy data, and has the advantages of reducing computational complexity and improving image edge detail smoothing and reconstruction accuracy. Finally, the reconstructed traffic image is input into the user interface to provide decision support for users in traffic management, improve the timeliness of traffic data feedback, and is suitable for large-scale promotion.

[0130] Embodiment 2

[0131] This embodiment provides a traffic image processing system, including:

[0132] A processing module, used for acquiring an original traffic image and preprocessing the original traffic image to reduce noise of the original traffic image;

[0133] A construction module is used to perform sparse processing on the pre-processed traffic image, select pixel points in the sparsely processed traffic image, select image blocks with a size of preset pixels with the pixel points as the center, and construct an image matrix;

[0134] A solution module, used for performing an extremum solution on the image matrix based on a compressed sensing algorithm, so as to reconstruct the traffic image according to the image matrix and identify the target object in the traffic image;

[0135] The input module is used to input the reconstructed traffic image into the user interface for display, providing decision support for users in traffic management.

[0136] The step of preprocessing the original traffic image comprises:

[0137] Performing multi-scale decomposition on the original traffic image by using two-dimensional wavelet transform to obtain a first image signal wavelet coefficient and a first noise signal wavelet coefficient of the original traffic image;

[0138] Comparing the frequency of the wavelet coefficients of the first image signal with a preset frequency;

[0139] If it is less than the preset frequency, reconstructing the first noise signal wavelet coefficients, obtaining and removing the noise signal, and reconstructing the first image signal wavelet coefficients to complete the preprocessing;

[0140] If it is not less than the preset frequency, continue to perform two-dimensional wavelet transform on the first image signal to obtain the second image signal wavelet coefficient and the second noise signal wavelet coefficient, compare the frequency of the second image signal wavelet coefficient with the preset frequency, and so on;

[0141] The wavelet coefficients of the noise signal obtained after several two-dimensional wavelet transforms are reconstructed to obtain and remove the noise signal, and the wavelet coefficients of the image signal obtained after the last two-dimensional wavelet transform are reconstructed to complete the preprocessing.

[0142] The step of solving the maximum value of the image matrix based on the compressed sensing algorithm to reconstruct the traffic image according to the image matrix includes:

[0143] constructing an optimized coefficient matrix based on an overcomplete dictionary, wherein the optimized coefficient matrix is ​​used to represent a coefficient matrix of a reconstructed image;

[0144] Reconstruct and capture the optimization coefficient matrix based on the deep learning model to obtain image quality feedback during the image reconstruction process to capture features of interest in the image;

[0145] Performing total variation processing on the optimization coefficient matrix based on wavelet sparse regularization, wherein the total variation processing is used to improve image quality in the process of reconstructing images;

[0146] The optimized coefficient matrix after reconstruction capture and total variation processing is processed based on the basis pursuit algorithm to construct an objective function for solving the maximum value of the image matrix;

[0147] Iterate the objective function until the difference between the two iteration results of the objective function is less than the reduction threshold or the number of iterations is greater than the maximum number of iterations, stop iterating, and obtain the target optimization coefficient matrix;

[0148] A reconstructed image matrix is ​​obtained according to the target optimization coefficient matrix and the overcomplete dictionary, thereby completing the reconstruction of the traffic image.

[0149] The objective function is expressed as:

[0150]

[0151] In the formula, is an m×n image matrix, n is the number of pixels, m is the pixel size of the image block, is an overcomplete dictionary consisting of p basis vectors of size m, p>m, is a p×n optimization coefficient matrix, for The L1 norm of for The L2 norm of , is the regularization parameter, is the result of total variation processing for the optimization coefficient matrix, To adjust the parameters of the loss function weights of the deep learning model, The constraint term for the reconstruction capture process to optimize the coefficient matrix is ​​used to measure the difference between A and the output of the deep learning model;

[0152] in,

[0153]

[0154] In the formula, is the maximum dimension of the gradient operator, represents the t-th dimension gradient operator, Represents wavelet transform.

[0155] After constructing the objective function for solving the maximum value of the image matrix, the traffic image processing method further includes:

[0156] Performing change separation on the optimization coefficient matrix to obtain a positive change part and a negative change part of the optimization coefficient matrix;

[0157] After the change separation, the objective function is expressed as:

[0158]

[0159] In the formula, To optimize the coefficient matrix The positive change part of To optimize the coefficient matrix The negative change part.

[0160] The iterating of the objective function comprises: The positive change of , the optimization coefficient matrix The negative change , pixel weights, and deep learning models are iterated;

[0161] in, The expression for iteration is:

[0162]

[0163] The expression for iteration is:

[0164]

[0165] In the formula, For the k+1th iteration The value of For the k+1th iteration The value of For the kth iteration The value of For the kth iteration The value of

[0166] The expression for iterating pixel weight is:

[0167]

[0168] In the formula, is the weight of pixel (i, j) at the k+1th iteration, is the gradient at the kth iteration, is a preset number greater than zero;

[0169] The expression for the iteration of the deep learning model is:

[0170]

[0171] In the formula, is the updated parameter value of the deep learning model, is the parameter value of the deep learning model before updating, are the deep learning model parameters, is the learning rate, for Relative to gradient.

[0172] Embodiment 3

[0173] This embodiment provides a storage medium on which a computer program is stored. When the program is executed by a processor, the above-mentioned traffic image processing method is implemented.

[0174] Embodiment 4

[0175] The present invention also provides a computer, see Figure 2 , shown is a computer in an embodiment of the present invention, including a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20. When the processor 20 executes the computer program 30, the above-mentioned traffic image processing method is implemented.

[0176] The memory 10 includes at least one type of storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory, etc.), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10 may be an internal storage unit of a computer, such as a hard disk of the computer. In other embodiments, the memory 10 may also be an external storage device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card, etc. Further, the memory 10 may also include both an internal storage unit of the computer and an external storage device. The memory 10 may be used not only to store application software and various types of data installed in the computer, but also to temporarily store data that has been output or is to be output.

[0177] Among them, in some embodiments, the processor 20 can be an electronic control unit (Electronic Control Unit, abbreviated as ECU, also known as a vehicle computer), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor or other data processing chip, used to run the program code stored in the memory 10 or process data, such as executing access restriction programs, etc.

[0178] It should be pointed out that Figure 2 The structure shown does not constitute a limitation on the computer. In other embodiments, the computer may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0179] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0180] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0181] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0182] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0183] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A traffic image processing method, characterized in that: include: Acquire an original traffic image, and preprocess the original traffic image to reduce noise of the original traffic image; Performing sparse processing on the preprocessed traffic image, selecting pixel points in the sparsely processed traffic image, selecting an image block with a size of preset pixels with the pixel point as the center, and constructing an image matrix; Solving the maximum value of the image matrix based on a compressed sensing algorithm to reconstruct the traffic image according to the image matrix and identify the target object in the traffic image; The reconstructed traffic image is input into the user interface for display, providing decision support for users in traffic management; The step of solving the maximum value of the image matrix based on the compressed sensing algorithm to reconstruct the traffic image according to the image matrix includes: constructing an optimized coefficient matrix based on an overcomplete dictionary, wherein the optimized coefficient matrix is ​​used to represent a coefficient matrix of a reconstructed image; Reconstruct and capture the optimization coefficient matrix based on the deep learning model to obtain image quality feedback during the image reconstruction process to capture features of interest in the image; Performing total variation processing on the optimization coefficient matrix based on wavelet sparse regularization, wherein the total variation processing is used to improve image quality in the process of reconstructing images; The optimized coefficient matrix after reconstruction capture and total variation processing is processed based on the basis pursuit algorithm to construct an objective function for solving the maximum value of the image matrix; Iterate the objective function until the difference between the two iteration results of the objective function is less than the reduction threshold or the number of iterations is greater than the maximum number of iterations, stop iterating, and obtain the target optimization coefficient matrix; Acquire a reconstructed image matrix according to the target optimization coefficient matrix and the overcomplete dictionary, thereby completing the reconstruction of the traffic image; The objective function is expressed as: In the formula, is an m×n image matrix, n is the number of pixels, m is the pixel size of the image block, is an overcomplete dictionary consisting of p basis vectors of size m, p>m, is a p×n optimization coefficient matrix, for The L1 norm of for The L2 norm of , is the regularization parameter, is the result of total variation processing for the optimization coefficient matrix, To adjust the parameters of the loss function weights of the deep learning model, The constraint term for the reconstruction capture process to optimize the coefficient matrix is ​​used to measure the difference between A and the output of the deep learning model; in, In the formula, is the maximum dimension of the gradient operator, represents the t-th dimension gradient operator, represents wavelet transform; After constructing the objective function for solving the maximum value of the image matrix, the traffic image processing method further includes: Performing change separation on the optimization coefficient matrix to obtain a positive change part and a negative change part of the optimization coefficient matrix; After the change separation, the objective function is expressed as: In the formula, To optimize the coefficient matrix The positive change part of To optimize the coefficient matrix The negative change part.

2. The traffic image processing method according to claim 1, characterized in that: The step of preprocessing the original traffic image comprises: Performing multi-scale decomposition on the original traffic image by using two-dimensional wavelet transform to obtain a first image signal wavelet coefficient and a first noise signal wavelet coefficient of the original traffic image; Comparing the frequency of the wavelet coefficients of the first image signal with a preset frequency; If it is less than the preset frequency, reconstructing the first noise signal wavelet coefficients, obtaining and removing the noise signal, and reconstructing the first image signal wavelet coefficients to complete the preprocessing; If it is not less than the preset frequency, continue to perform two-dimensional wavelet transform on the first image signal to obtain the second image signal wavelet coefficient and the second noise signal wavelet coefficient, compare the frequency of the second image signal wavelet coefficient with the preset frequency, and so on; The wavelet coefficients of the noise signal obtained after several two-dimensional wavelet transforms are reconstructed to obtain and remove the noise signal, and the wavelet coefficients of the image signal obtained after the last two-dimensional wavelet transform are reconstructed to complete the preprocessing.

3. The traffic image processing method according to claim 1, characterized in that: The iterating of the objective function comprises: The positive change of , the optimization coefficient matrix The negative change , pixel weights, and deep learning models are iterated; in, The expression for iteration is: The expression for iteration is: In the formula, For the k+1th iteration The value of For the k+1th iteration The value of For the kth iteration The value of For the kth iteration The value of The expression for iterating pixel weight is: In the formula, is the weight of pixel (i, j) at the k+1th iteration, is the gradient at the kth iteration, is a preset number greater than zero; The expression for the iteration of the deep learning model is: In the formula, is the updated parameter value of the deep learning model, is the parameter value of the deep learning model before updating, are the deep learning model parameters, is the learning rate, for Relative to gradient.

4. A traffic image processing system, characterized in that: include: A processing module, used for acquiring an original traffic image and preprocessing the original traffic image to reduce noise of the original traffic image; A construction module is used to perform sparse processing on the pre-processed traffic image, select pixel points in the sparsely processed traffic image, select image blocks with a size of preset pixels with the pixel points as the center, and construct an image matrix; A solution module, used for performing an extremum solution on the image matrix based on a compressed sensing algorithm, so as to reconstruct the traffic image according to the image matrix and identify the target object in the traffic image; An input module is used to input the reconstructed traffic image into the user interface for display, so as to provide decision support for users in traffic management; The step of solving the maximum value of the image matrix based on the compressed sensing algorithm to reconstruct the traffic image according to the image matrix includes: constructing an optimized coefficient matrix based on an overcomplete dictionary, wherein the optimized coefficient matrix is ​​used to represent a coefficient matrix of a reconstructed image; Reconstruct and capture the optimization coefficient matrix based on the deep learning model to obtain image quality feedback during the image reconstruction process to capture features of interest in the image; Performing total variation processing on the optimization coefficient matrix based on wavelet sparse regularization, wherein the total variation processing is used to improve image quality in the process of reconstructing images; The optimized coefficient matrix after reconstruction capture and total variation processing is processed based on the basis pursuit algorithm to construct an objective function for solving the maximum value of the image matrix; Iterate the objective function until the difference between the two iteration results of the objective function is less than the reduction threshold or the number of iterations is greater than the maximum number of iterations, stop iterating, and obtain the target optimization coefficient matrix; Acquire a reconstructed image matrix according to the target optimization coefficient matrix and the overcomplete dictionary, thereby completing the reconstruction of the traffic image; The objective function is expressed as: In the formula, is an m×n image matrix, n is the number of pixels, m is the pixel size of the image block, is an overcomplete dictionary consisting of p basis vectors of size m, p>m, is a p×n optimization coefficient matrix, for The L1 norm of for The L2 norm of , is the regularization parameter, is the result of total variation processing for the optimization coefficient matrix, To adjust the parameters of the loss function weights of the deep learning model, The constraint term for the reconstruction capture process to optimize the coefficient matrix is ​​used to measure the difference between A and the output of the deep learning model; in, In the formula, is the maximum dimension of the gradient operator, represents the t-th dimension gradient operator, represents wavelet transform; After constructing the objective function for solving the maximum value of the image matrix, the traffic image processing method further includes: Performing change separation on the optimization coefficient matrix to obtain a positive change part and a negative change part of the optimization coefficient matrix; After the change separation, the objective function is expressed as: In the formula, To optimize the coefficient matrix The positive change part of To optimize the coefficient matrix The negative change part.

5. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the traffic image processing method as described in any one of claims 1 to 3 is implemented.

6. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the traffic image processing method according to any one of claims 1 to 3 is implemented.

Images