A data visualization processing method for intelligent transportation

Abstract

The present invention discloses a data visualization processing method for intelligent transportation, which relates to the technical field of data processing and includes the following steps: S1, collecting panoramic traffic images of a target area; S2, performing denoising processing on the panoramic traffic images to obtain high-definition traffic images; S3, obtaining key texture locations of each street image block; S4, determining the number of existing vehicles; S5, generating a traffic flow heat map for the target area. By extracting the feature probabilities of each street image block in the high-definition traffic images, the method of the present invention can automatically identify key texture locations; based on the key texture locations, the model is used to extract the number of contours as the number of vehicles in the street image block; finally, according to the number of vehicles in each street image block, a heat map reflecting traffic flow is generated in real time, providing intuitive and accurate visualization support for urban traffic management and planning.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly relates to a data visualization processing method for intelligent transportation. Background Art

[0002] With the acceleration of urbanization and the increase in traffic flow, the problem of traffic congestion has become increasingly serious. In order to effectively manage and optimize urban traffic, real-time and accurate traffic data analysis and visualization have become particularly important. As an intuitive data display method, a traffic heat map can clearly present the distribution and changes of traffic flow, providing important references for traffic planning and management. The generation of a traffic heat map relies on the accurate acquisition of vehicle density or the number of vehicles. However, most of the existing methods for obtaining vehicle numbers or densities rely on traffic images obtained by surveillance cameras, but traffic images are easily affected by weather, and the extracted vehicle numbers are inaccurate and cannot reflect the traffic density of a certain area.

[0003] In summary, accurately obtaining traffic density is of great significance for generating a traffic flow heat map and then realizing the visualization effect of traffic data. Summary of the Invention

[0004] In order to solve the above problems, the present invention proposes a data visualization processing method for intelligent transportation.

[0005] The technical solution of the present invention is: A data visualization processing method for intelligent transportation includes the following steps:

[0006] S1. Collect panoramic traffic images of the target area;

[0007] S2. Denoise the panoramic traffic images to obtain high-definition traffic images;

[0008] S3. Extract the feature probabilities of each street image block in the high-definition traffic image to obtain the key texture locations of each street image block;

[0009] S4. Based on the key texture locations of each street image block, determine the number of existing vehicles;

[0010] S5. Generate a traffic flow heat map for the target area according to the number of existing vehicles in each street image block.

[0011] Further, S3 includes the following sub-steps:

[0012] S31. Divide the high-definition traffic image into several street image blocks;

[0013] S32. Use a convolution kernel to extract multi-layer feature matrices of each street image block;

[0014] S33. Take the sum of the maximum singular values of all multi-layer feature matrices as the total sample value, and take the ratio between the maximum singular value of the multi-layer feature matrix of the street image patch and the total sample value as the feature probability of the street image patch;

[0015] S34. Calculate the background observation degree of the street image patch according to the feature probability of the street image patch;

[0016] S35. Based on the background observation degrees of all street image patches, obtain the key texture locations.

[0017] The beneficial effects of the above further solution are as follows: In the present invention, in the convolutional neural network, the input image is represented as a multi-layer feature matrix. Using the convolutional kernel to extract the multi-layer feature matrices of each street image patch can capture various features from low level to high level; by calculating the sum of the maximum singular values of the multi-layer feature matrix as the total sample value, and taking the ratio between the maximum singular value of the multi-layer feature matrix of the street image patch and the total sample value as the feature probability, the feature saliency of each street image patch can be quantified. Based on the background observation degrees of all street image patches, the background texture locations can be obtained.

[0018] In S33, the calculation formula for the feature probability of the th street image patch is: ; where represents the maximum singular value of the multi-layer feature matrix of the th street image patch, and represents the number of street image patches in the high-definition traffic image.

[0019] Further, in S34, the calculation formula for the background observation degree of the th street image patch is: ; where represents the feature probability of the th street image patch, represents the weight of the street image patch, represents the logarithmic function with base 2, and represents taking a random value between 0 and 1.

[0020] The weight of the street image patch can be determined by the ratio of the area occupied by the street image patch in the high-definition traffic image.

[0021] Further, S35 includes the following sub-steps:

[0022] S351. Generate local background constraint conditions;

[0023] S352. Input each street image patch and its adjacent street image patches into the local background constraint condition, retain the pixel points in the street image patch that satisfy the local background constraint, and eliminate the pixel points in the street image patch that do not satisfy the local background constraint;

[0024] S353. Take the retained pixel points in the street image patch as the key texture areas.

[0025] The beneficial effect of the above further solution is that in the present invention, by inputting the street image patch and its adjacent image patches into the local background constraint condition, this method can retain the pixel points that meet the constraint conditions and eliminate the background pixel points at the same time, obtaining the area that can be used for vehicle counting.

[0026] Further, in S352, the expression of the local background constraint condition is: ; where represents the background observation degree of the th street image patch, represents the background observation degree of the th adjacent street image patch to the th street image patch, represents the background observation degree of the th pixel point in the th street image patch, represents the sum of the background observation degrees of all pixel points in the th street image patch, represents the total number of adjacent street image patches to the th street image patch.

[0027] Further, S4 includes the following sub-steps:

[0028] S41. Extract the pixel values of all pixel points in the key texture areas of the street image patch as the texture vector;

[0029] S42. Obtain the texture regularization term according to the texture vector of the street image patch;

[0030] S43. Obtain the number of vehicles existing in the street image patch according to the texture regularization term.

[0031] The beneficial effect of the above further solution is that in the present invention, by extracting the pixel values of all pixel points in the key texture areas of the street image patch as the texture vector, the key texture features in the street image can be accurately captured; using the texture regularization term as the parameter of the Snake model can reduce the risk of overfitting to a certain extent and improve the accuracy of vehicle number recognition.

[0032] Further, in S42, the expression of the texture regularization term is:

[0033] ;

[0034] In the formula, represents the regularization coefficient, represents the number of pixel points at the key texture, represents a constant, represents the texture vector of the street image block, represents the second norm of the vector.

[0035] The regularization coefficient is a parameter that adds constraints (penalties) to the objective function in machine learning.

[0036] Further, in S43, the processed texture regularization term is used as the elastic coefficient of the Snake model, and the number of contours of the street image block is extracted by using the Snake model as the number of existing vehicles.

[0037] The beneficial effect of the above further solution is that in the present invention, the SNAKE model is a deformable parametric curve and the corresponding energy function, which is used to extract the target contour. By using the processed texture regularization term as the elastic coefficient of the Snake model, the texture features of the street image block can be accurately incorporated into the contour extraction process.

[0038] Further, in S43, the processed texture regularization term has the following expression:

[0039] ;

[0040] In the formula, represents the texture regularization term, represents the rounding function.

[0041] The beneficial effect of the present invention is that by extracting the feature probabilities of each street image block in the high-definition traffic image, this method can automatically identify the key texture locations; based on the key texture locations, the model is used to extract the number of contours as the number of vehicles in the street image block; finally, according to the number of vehicles in each street image block, a heat map reflecting the traffic flow is generated in real time, providing intuitive and accurate visual support for urban traffic management and planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flowchart of a data visualization processing method for intelligent transportation. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0044] As Figure 1As shown in the figure, the present invention provides a data visualization processing method for intelligent transportation, including the following steps:

[0045] S1. Collect panoramic traffic images of the target area;

[0046] S2. Denoise the panoramic traffic images to obtain high-definition traffic images;

[0047] S3. Extract the feature probabilities of each street image block in the high-definition traffic image to obtain the key texture locations of each street image block;

[0048] S4. Determine the number of vehicles present based on the key texture locations of each street image block;

[0049] S5. Generate a traffic flow heat map for the target area according to the number of vehicles present in each street image block.

[0050] In the embodiment of the present invention, S3 includes the following sub-steps:

[0051] S31. Divide the high-definition traffic image into several street image blocks;

[0052] S32. Use a convolution kernel to extract the multi-layer feature matrices of each street image block;

[0053] S33. Take the sum of the maximum singular values of all multi-layer feature matrices as the total sample value, and take the ratio between the maximum singular value of the multi-layer feature matrix of the street image block and the total sample value as the feature probability of the street image block;

[0054] S34. Calculate the background observation degree of the street image block according to the feature probability of the street image block;

[0055] S35. Obtain the key texture locations based on the background observation degrees of all street image blocks.

[0056] In the present invention, in a convolutional neural network, the input image is represented as a multi-layer feature matrix. Using a convolution kernel to extract the multi-layer feature matrices of each street image block can capture various features from low-level to high-level; by calculating the sum of the maximum singular values of the multi-layer feature matrices as the total sample value, and taking the ratio between the maximum singular value of the multi-layer feature matrix of the street image block and the total sample value as the feature probability, the feature saliency of each street image block can be quantified. Based on the background observation degrees of all street image blocks, the background texture locations can be obtained.

[0057] In S33, the formula for the feature probability of the th street image block is: ; where represents the The maximum singular value of the multi-layer feature matrix of the street image patches, represents the number of street image patches in the high-definition traffic image.

[0058] In the embodiment of the present invention, in S34, the background observability of the th street image patch is calculated as follows: ; where represents the feature probability of the th street image patch, represents the weight of the street image patch, represents the logarithmic function with base 2, represents taking a random value between 0 and 1.

[0059] The weight of the street image patch can be determined by the ratio of the area occupied by the street image patch in the high-definition traffic image.

[0060] In the embodiment of the present invention, S35 includes the following sub-steps:

[0061] S351. Generate local background constraint conditions;

[0062] S352. Input each street image patch and its adjacent street image patches into the local background constraint conditions, retain the pixel points in the street image patch that satisfy the local background constraints, and eliminate the pixel points in the street image patch that do not satisfy the local background constraints;

[0063] S353. Take the retained pixel points in the street image patch as the key texture areas.

[0064] In the present invention, inputting the street image patch and its adjacent image patches into the local background constraint conditions, this method can retain the pixel points that satisfy the constraint conditions, and at the same time eliminate the background pixel points to obtain the area that can be used for vehicle counting.

[0065] In the embodiment of the present invention, in S352, the expression of the local background constraint conditions is: ; where represents the background observability of the th street image patch, represents the background observability of the th adjacent street image patch to the th street image patch, represents the background observability of the th pixel point in the th street image patch, represents the sum of the background observabilities of all pixel points in the th street image patch, represents the The total number of adjacent street image patches of a street image patch.

[0066] In an embodiment of the present invention, S4 includes the following sub-steps:

[0067] S41. Extract the pixel values of all pixel points at the key texture of the street image patch as a texture vector;

[0068] S42. Obtain a texture regularization term according to the texture vector of the street image patch;

[0069] S43. Obtain the number of vehicles existing in the street image patch according to the texture regularization term.

[0070] In the present invention, by extracting the pixel values of all pixel points at the key texture of the street image patch as a texture vector, the key texture features in the street image can be accurately captured; using the texture regularization term as a parameter of the Snake model can, to a certain extent, reduce the risk of overfitting and improve the accuracy of vehicle number recognition.

[0071] In an embodiment of the present invention, in S42, the texture regularization term has the following expression:

[0072] ;

[0073] In the formula, represents the regularization coefficient, represents the number of pixel points at the key texture, represents a constant, represents the texture vector of the street image patch, represents the two-norm of the vector.

[0074] The regularization coefficient is a parameter for adding constraints (penalties) to the objective function in machine learning.

[0075] In an embodiment of the present invention, in S43, the processed texture regularization term is used as the elastic coefficient of the Snake model, and the Snake model is used to extract the contour number of the street image patch as the number of existing vehicles.

[0076] In the present invention, the SNAKE model is a deformable parametric curve and the corresponding energy function for extracting the target contour. By using the processed texture regularization term as the elastic coefficient of the Snake model, the texture features of the street image patch can be accurately incorporated into the contour extraction process.

[0077] In an embodiment of the present invention, in S43, the processed texture regularization term has the following expression:

[0078] ;

[0079] In the formula, represents the texture regularization term, represents the rounding function.

[0080] Those of ordinary skill in the art will realize that the embodiments described herein are for helping the reader understand the principles of the present invention, and it should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations without departing from the essence of the present invention according to these technical revelations disclosed by the present invention, and these deformations and combinations are still within the protection scope of the present invention.

Claims

1. A data visualization processing method for intelligent transportation, characterized in that It includes the following steps: S1. Collect the panoramic traffic images of the target area; S2. Denoise the panoramic traffic images to obtain high-definition traffic images; S3. Extract the feature probabilities of each street image block in the high-definition traffic images to obtain the key texture areas of each street image block; S4. Determine the number of existing vehicles based on the key texture areas of each street image block; S5. Generate a traffic flow heat map for the target area according to the number of existing vehicles in each street image block; The S3 includes the following sub-steps: S31. Divide the high-definition traffic images into several street image blocks; S32. Use the convolution kernel to extract the multi-layer feature matrices of each street image block; S33. Take the sum of the maximum singular values of all multi-layer feature matrices as the total sample value, and take the ratio between the maximum singular value of the multi-layer feature matrix of the street image block and the total sample value as the feature probability of the street image block; S34. Calculate the background observation degree of the street image block according to the feature probability of the street image block; S35. Obtain the key texture areas based on the background observation degrees of all street image blocks; In S34, the background observation degree of the th street image block is calculated as follows: ; in the formula, represents the feature probability of the th street image block, represents the weight of the street image block, represents the logarithmic function with base 2, represents taking a random value between 0 and 1.

2. The data visualization processing method for intelligent transportation according to claim 1, wherein The S35 includes the following sub-steps: S351. Generate local background constraint conditions; S352. Input each street image block and its adjacent street image blocks into the local background constraint conditions, retain the pixel points in the street image block that satisfy the local background constraints, and eliminate the pixel points in the street image block that do not satisfy the local background constraints; S353. Take the retained pixel points in the street image block as the key texture areas.

3. The data visualization processing method for intelligent transportation according to claim 2, wherein In the above S352, the expression of the local background constraint condition is as follows: ; where represents the background observation degree of the -th street image block, represents the background observation degree of the -th street image block adjacent to the -th street image block, represents the background observation degree of the -th pixel in the -th street image block, represents the sum of the background observation degrees of all pixels in the -th street image block, represents the total number of adjacent street image blocks to the -th street image block.

4. The data visualization processing method for intelligent transportation according to claim 1, characterized in that, The S4 includes the following sub-steps: S41. Extract the pixel values of all pixel points in the key texture areas of the street image block as the texture vector; S42. Obtain the texture regularization term according to the texture vector of the street image block; S43. Obtain the number of existing vehicles in the street image block according to the texture regularization term.

5. The data visualization processing method for intelligent transportation according to claim 4, characterized in that In S42, the texture regularization term has the following expression: ； In the formula, represents the regularization coefficient, represents the number of pixel points at the key texture, represents a constant, represents the texture vector of the street image block, represents the two-norm of the vector.

6. The data visualization processing method for intelligent transportation according to claim 4, wherein In the S43, take the processed texture regularization term as the elastic coefficient of the Snake model, and use the Snake model to extract the contour number of the street image block as the number of existing vehicles.

7. The data visualization processing method for intelligent transportation according to claim 6, wherein In S43, the processed texture regularization term has the following expression: ； In the formula, represents the texture regularization term, represents the rounding function.

Images