Intelligent Transportation Control Method and System Based on Street Lamp Poles

Abstract

The present application discloses an intelligent traffic control method and system based on street lamp poles. It adopts artificial intelligence monitoring technology based on deep learning to extract the implicit feature distribution information of road information in the monitoring images of each road, and uses the spatial connectivity topological features between the roads to strengthen the expression of the image features of each road, so as to identify and judge the degree of road congestion. In this way, it can accurately detect and judge the traffic environment information comprehensively and intelligently to realize the construction of a smart city, thereby improving the efficiency of resource utilization and optimizing urban management and services.

Description

Technical Field

[0001] The present application relates to the field of intelligent control technologies, and more specifically, to an intelligent transportation control method and system based on street lamp poles. Background Art

[0002] Today's street lamp poles can utilize technologies such as image processing, single-lamp control, and system integration, integrating multiple functions such as intelligent lighting, mobile network coverage, charging, and video surveillance. While reducing the total space occupied by equipment, they provide great convenience for people's lives.

[0003] The country vigorously promotes the construction of smart cities, encouraging the use of various information technologies or innovative concepts to integrate and connect the systems and services of cities, so as to improve the efficiency of resource utilization, optimize urban management and services, and improve the quality of life of citizens. The intelligent transportation system is the core part of the construction of smart cities. However, the current transportation system is not comprehensive and accurate enough in collecting traffic environment information.

[0004] Therefore, an optimized intelligent transportation control solution based on street lamp poles is expected. Summary of the Invention

[0005] To solve the above technical problems, the present application is proposed. Embodiments of the present application provide an intelligent transportation control method and system based on street lamp poles. It adopts artificial intelligence monitoring technology based on deep learning to extract the implicit feature distribution information of road information in the monitoring images of each road, and uses the spatial connectivity topological features between the roads to strengthen the expression of the image features of each road, so as to identify and judge the degree of road congestion. In this way, the traffic environment information can be accurately, comprehensively, and intelligently detected and judged to realize the construction of smart cities, thereby improving the efficiency of resource utilization and optimizing urban management and services.

[0006] According to one aspect of the present application, an intelligent transportation control method based on street lamp poles is provided, which includes:

[0007] Obtain the monitoring images of each road collected by the cameras of the intelligent street lamp poles deployed on each road;

[0008] Respectively pass the monitoring images of each road through a first convolutional neural network model as a filter to obtain a plurality of road monitoring feature vectors;

[0009] Construct the spatial connectivity topological matrix between each road, where the values of each position in the non-diagonal positions of the spatial connectivity topological matrix are used to represent whether the corresponding two roads are connected;

[0010] Pass the spatial connectivity topology matrix through a second convolutional neural network model acting as a feature extractor to obtain a spatial connectivity topology feature matrix;

[0011] Arrange the multiple road monitoring feature vectors two-dimensionally to obtain a road monitoring global feature matrix;

[0012] Pass the road monitoring global feature matrix and the spatial connectivity topology feature matrix through a graph neural network to obtain a topological road monitoring global feature matrix;

[0013] Perform feature distribution correction on the topological road monitoring global feature matrix to obtain a classification feature matrix;

[0014] Pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to represent the grade label of the road congestion degree.

[0015] In the above intelligent transportation control method based on street lamp poles, the step of passing the monitoring images of each road through a first convolutional neural network model acting as a filter to obtain multiple road monitoring feature vectors includes: using each layer of the first convolutional neural network model acting as a filter to respectively perform the following operations on the input data during the forward pass of the layer:

[0016] Perform convolution processing on the input data to obtain a convolution feature map;

[0017] Perform mean pooling processing based on a feature matrix on the convolution feature map to obtain a pooled feature map;

[0018] Perform non-linear activation on the pooled feature map to obtain an activation feature map;

[0019] Among them, the output of the last layer of the first convolutional neural network model acting as a filter is the multiple road monitoring feature vectors, and the input of the first layer of the first convolutional neural network model acting as a filter is the monitoring images of each road.

[0020] In the above intelligent transportation control method based on street lamp poles, the step of passing the spatial connectivity topology matrix through a second convolutional neural network model acting as a feature extractor to obtain a spatial connectivity topology feature matrix includes: using each layer of the second convolutional neural network model to respectively perform convolution processing, pooling processing along the channel dimension, and non-linear activation processing on the input data during the forward pass of the layer, so that the last layer of the second convolutional neural network model outputs the spatial connectivity topology feature matrix, where the input of the first layer of the second convolutional neural network model is the spatial connectivity topology matrix.

[0021] In the above-mentioned intelligent transportation control method based on street lamp poles, the step of obtaining the topological road monitoring global feature matrix by passing the road monitoring global feature matrix and the spatial connectivity topological feature matrix through a graph neural network includes:

[0022] The graph neural network processes the road monitoring global feature matrix and the spatial connectivity topological feature matrix through learnable neural network parameters to obtain the topological road monitoring global feature matrix containing irregular spatial connectivity topological features and high-dimensional road monitoring image features.

[0023] In the above-mentioned intelligent transportation control method based on street lamp poles, the step of performing feature distribution correction on the topological road monitoring global feature matrix to obtain a classification feature matrix includes:

[0024] Performing feature distribution correction on the topological road monitoring global feature matrix according to the following formula to obtain the classification feature matrix;

[0025] where the formula is:

[0026]

[0027] where M and M′ are the topological road monitoring global feature matrix and the classification feature matrix respectively, ReLU(·) represents the ReLU activation function, represents matrix multiplication, and the division between the numerator matrix and the denominator matrix is the element-wise division of matrix eigenvalues, exp(·) represents the exponential operation of the matrix, and the exponential operation of the matrix represents calculating the natural exponential function values with the eigenvalues at each position in the matrix as the exponents.

[0028] In the above-mentioned intelligent transportation control method based on street lamp poles, the step of passing the classification feature matrix through a classifier to obtain a classification result, where the classification result is used to represent the grade label of road congestion degree, includes:

[0029] Expanding the classification feature matrix into a classification feature vector according to row vectors or column vectors;

[0030] Using multiple fully connected layers of the classifier to perform fully connected encoding on the classification feature vector to obtain an encoded classification feature vector;

[0031] Passing the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0032] According to another aspect of the present application, there is provided an intelligent transportation control system based on street lamp poles, which includes:

[0033] A monitoring image acquisition module, configured to acquire monitoring images of each road collected by cameras of intelligent street lamp poles deployed on each road;

[0034] A first convolution module, configured to respectively pass the monitoring images of each road through a first convolutional neural network model as a filter to obtain a plurality of road monitoring feature vectors;

[0035] A topological matrix construction module, configured to construct a spatial connectivity topological matrix between each road, wherein the values of each position in the non-diagonal positions of the spatial connectivity topological matrix are used to indicate whether the corresponding two roads are connected;

[0036] A second convolution module, configured to pass the spatial connectivity topological matrix through a second convolutional neural network model as a feature extractor to obtain a spatial connectivity topological feature matrix;

[0037] A feature matrix arrangement module, configured to perform two-dimensional arrangement on the plurality of road monitoring feature vectors to obtain a road monitoring global feature matrix;

[0038] A graph neural encoding module, configured to pass the road monitoring global feature matrix and the spatial connectivity topological feature matrix through a graph neural network to obtain a topological road monitoring global feature matrix;

[0039] A feature distribution correction module, configured to perform feature distribution correction on the topological road monitoring global feature matrix to obtain a classification feature matrix;

[0040] A classification result generation module, configured to pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to represent a grade label of the road congestion degree.

[0041] In the above intelligent transportation control system based on street lamp poles, the first convolution module is further configured to: use each layer of the first convolutional neural network model as a filter to respectively perform the following operations on the input data during the forward pass of the layer:

[0042] Perform convolution processing on the input data to obtain a convolution feature map;

[0043] Perform mean pooling processing based on a feature matrix on the convolution feature map to obtain a pooling feature map;

[0044] Perform non-linear activation on the pooling feature map to obtain an activation feature map;

[0045] Wherein, the output of the last layer of the first convolutional neural network model as a filter is the plurality of road monitoring feature vectors, and the input of the first layer of the first convolutional neural network model as a filter is the monitoring images of each road.

[0046] In the above-mentioned intelligent transportation control system based on street lamp poles, the second convolutional module is further configured to: use each layer of the second convolutional neural network model to perform convolutional processing, pooling processing along the channel dimension, and non-linear activation processing on the input data respectively during the forward pass of the layer, so as to output the spatial connectivity topological feature matrix by the last layer of the second convolutional neural network model, wherein the input of the first layer of the second convolutional neural network model is the spatial connectivity topological matrix.

[0047] In the above-mentioned intelligent transportation control system based on street lamp poles, the graph neural coding module is further configured to: process the road monitoring global feature matrix and the spatial connectivity topological feature matrix through learnable neural network parameters to obtain the topological road monitoring global feature matrix including irregular spatial connectivity topological features and high-dimensional road monitoring image features.

[0048] Compared with the prior art, the intelligent transportation control method and system based on street lamp poles provided by this application adopt artificial intelligence monitoring technology based on deep learning to extract the implicit feature distribution information about road information in the monitoring images of each road, and use the spatial connectivity topological features between each road to strengthen the expression of the image features of each road, so as to identify and judge the degree of road congestion. In this way, it is possible to accurately and comprehensively and intelligently detect and judge the traffic environment information to realize the construction of a smart city, thereby improving the efficiency of resource utilization and optimizing urban management and services. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0050] Figure 1 FIG. is a schematic diagram of the scenario of the intelligent transportation control method based on street lamp poles according to the embodiment of the present application.

[0051] Figure 2 FIG. is a flowchart of the intelligent transportation control method based on street lamp poles according to the embodiment of the present application.

[0052] Figure 3 FIG. is a schematic diagram of the architecture of the intelligent transportation control method based on street lamp poles according to the embodiment of the present application.

[0053] Figure 4It is a flowchart of sub-steps of step S180 in the intelligent traffic control method based on street lamp poles according to an embodiment of the present application.

[0054] Figure 5 It is a block diagram of the intelligent traffic control method system based on street lamp poles according to an embodiment of the present application. Detailed implementation manners

[0055] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0056] Scene overview

[0057] As mentioned above, since today's traffic systems do not collect traffic environment information comprehensively and accurately enough. Therefore, an optimized intelligent traffic control solution based on street lamp poles is expected.

[0058] Currently, deep learning and neural networks have been widely applied in fields such as computer vision, natural language processing, and speech signal processing. In addition, deep learning and neural networks have also shown levels close to or even exceeding that of humans in fields such as image classification, object detection, semantic segmentation, and text translation.

[0059] In recent years, the development of deep learning and neural networks has provided new solutions and ideas for the intelligent recognition of traffic environment information in intelligent transportation.

[0060] Correspondingly, considering that when collecting and analyzing traffic environment information, it can be carried out by collecting surveillance images of each road through the cameras of intelligent street lamp poles deployed on each road. However, since there are spatial correlations between the characteristics of the surveillance images of each road, that is, there are spatial correlation characteristics of whether each two roads among the multiple roads are connected. If such spatial correlation characteristics can be used to strengthen the feature expression of the surveillance images of each road, it is obvious that the accuracy of road information recognition can be improved.

[0061] Based on this, in the technical solution of the present application, an artificial intelligence monitoring technology based on deep learning is adopted to extract the implicit feature distribution information of road information in the surveillance images of each road, and the spatial connectivity topological characteristics between each road are used to strengthen the expression of the image features of each road, so as to identify and judge the degree of road congestion. In this way, it is possible to comprehensively and intelligently detect and judge traffic environment information accurately, so as to realize the construction of a smart city, thereby improving the efficiency of resource utilization and optimizing urban management and services.

[0062] Specifically, in the technical solution of the present application, first, the monitoring images of each road are collected by the cameras of the intelligent street lamp poles deployed on each road. Then, a first convolutional neural network model used as a filter, which has excellent performance in extracting implicit features of images, is used to mine the features of the monitoring images of each road, so as to separately extract the implicit feature distribution information about road information in the monitoring images of each road, thereby obtaining multiple road monitoring feature vectors.

[0063] Then, considering that when monitoring road information, due to the characteristic distribution of spatial correlation among the roads, that is, there is a correlation relationship of whether the roads are connected among the roads. Therefore, in the technical solution of the present application, in order to improve the global monitoring of the road information and thus improve the judgment accuracy of the road congestion degree, the spatial connectivity topological features among the roads are further used to optimize the expression of the monitoring image features of each road. Specifically, first, a spatial connectivity topological matrix among the roads is constructed. Here, the values of each position in the non-diagonal positions of the spatial connectivity topological matrix are used to represent whether the corresponding two roads are connected. Then, the spatial connectivity topological matrix is subjected to feature mining through a second convolutional neural network model used as a feature extractor to extract the spatial connectivity topological feature distribution information among the roads, thereby obtaining a spatial connectivity topological feature matrix.

[0064] Furthermore, taking each road monitoring feature vector in the multiple road monitoring feature vectors as the feature representation of the node, and taking the spatial connectivity topological feature matrix as the feature representation of the edge between nodes, the road monitoring global feature matrix obtained by two-dimensional arrangement of the multiple road monitoring feature vectors and the spatial connectivity topological feature matrix are passed through a graph neural network to obtain a topological road monitoring global feature matrix. Specifically, the graph neural network encodes the road monitoring global feature matrix and the spatial connectivity topological feature matrix through learnable neural network parameters to obtain the topological road monitoring global feature matrix containing the spatial connectivity topological features of the irregular roads and the high-dimensional implicit features of the road images.

[0065] Then, taking the topological road monitoring global feature matrix as a classification feature matrix, classification processing is performed through a classifier to obtain a classification result of a grade label representing the road congestion degree. In this way, the traffic environment information can be comprehensively and intelligently detected and judged to realize the construction of a smart city.

[0066] In particular, in the technical solution of the present application, each topological road monitoring global feature vector of the topological road monitoring global feature matrix is a row vector of the topological road monitoring global feature matrix, and this row vector is a topological association expression of the image feature semantics obtained by a single camera with respect to the camera space topology. Therefore, it has a certain degree of independence. In this way, within the overall feature distribution of the topological road monitoring global feature matrix formed by arranging the topological road monitoring global feature vectors, there will be a negative correlation relationship between local distributions, thereby affecting the accuracy of the classification result obtained by the topological road monitoring global feature matrix through the classifier.

[0067] Therefore, the applicant of the present application corrects the topological road monitoring global feature matrix by means of fully positive projection non-linear reweighting, which is specifically expressed as:

[0068]

[0069] Here, M and M′ are the topological road monitoring global feature matrices before and after correction respectively. Among them, the topological road monitoring global feature matrix is pre-converted into diagonal features through linear transformation, and the division between the numerator matrix and the denominator matrix is the element-wise division of matrix eigenvalues.

[0070] Specifically, the fully positive projection non-linear reweighting uses the ReLU function to ensure the full positivity of the projection to avoid aggregating negatively correlated information, and at the same time introduces a non-linear reweighting mechanism to aggregate the local eigenvalue distributions of the topological road monitoring global feature matrix with each other, so that the internal structure of the topological road monitoring global feature matrix can penalize long-distance connections and strengthen local coupling. In this way, the synergistic effect of the spatial feature transformation corresponding to the fully positive projection reweighting of the topological road monitoring global feature matrix in the high-dimensional feature space is realized, thereby improving the accuracy of the classification result obtained by the topological road monitoring global feature matrix through the classifier. In this way, it is possible to accurately detect and judge the traffic environment information comprehensively and intelligently to realize the construction of a smart city, and further improve the efficiency of resource utilization and optimize urban management and services.

[0071] A smart traffic control method based on street light poles is provided, which includes: obtaining monitoring images of each road collected by cameras of smart street light poles deployed on each road; passing the monitoring images of each road through a first convolutional neural network model as a filter to obtain multiple road monitoring feature vectors; constructing a spatial connectivity topology matrix between each road, wherein the values ​​of each position on the non-diagonal position in the spatial connectivity topology matrix are used to indicate whether the corresponding two roads are connected; passing the spatial connectivity topology matrix through a second convolutional neural network model as a feature extractor to obtain a spatial connectivity topology feature matrix; arranging the multiple road monitoring feature vectors in two dimensions to obtain a road monitoring global feature matrix; passing the road monitoring global feature matrix and the spatial connectivity topology feature matrix through a graph neural network to obtain a topological road monitoring global feature matrix; performing feature distribution correction on the topological road monitoring global feature matrix to obtain a classification feature matrix; and passing the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to represent a grade label of the road congestion degree.

[0072] Figure 1 FIG. 1 is an application scenario diagram of the intelligent traffic control method based on street light poles according to an embodiment of the present application. Figure 1 As shown, in this application scenario, the intelligent street light poles (such as Figure 1 The camera (L) shown in FIG. Figure 1 C) collecting monitoring images of each road (such as Figure 1 Then, the monitoring images of each road are input to a server (such as a server) that is deployed with a smart traffic control algorithm based on street lamp poles. Figure 1 In S) shown in , the server is capable of generating classification results of grade labels for indicating the degree of road congestion based on the smart traffic control algorithm based on street light poles.

[0073] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0074] Exemplary Methods

[0075] Figure 2 Flow chart of the intelligent traffic control method based on street light poles according to an embodiment of the present application. Figure 2As shown in the figure, the intelligent transportation control method based on street lamp poles according to an embodiment of the present application includes the steps of: S110, acquiring monitoring images of each road collected by cameras of intelligent street lamp poles deployed on each road; S120, respectively passing the monitoring images of each road through a first convolutional neural network model as a filter to obtain a plurality of road monitoring feature vectors; S130, constructing a spatial connectivity topological matrix between each road, wherein the values of each position at non-diagonal positions in the spatial connectivity topological matrix are used to represent whether two corresponding roads are connected; S140, passing the spatial connectivity topological matrix through a second convolutional neural network model as a feature extractor to obtain a spatial connectivity topological feature matrix; S150, arranging the plurality of road monitoring feature vectors two-dimensionally to obtain a road monitoring global feature matrix; S160, passing the road monitoring global feature matrix and the spatial connectivity topological feature matrix through a graph neural network to obtain a topological road monitoring global feature matrix; S170, performing feature distribution correction on the topological road monitoring global feature matrix to obtain a classification feature matrix; and S180, passing the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to represent a grade label of the road congestion degree.

[0076] Figure 3 It is a schematic architecture diagram of the intelligent transportation control method based on street lamp poles according to an embodiment of the present application. As Figure 3 shown in the figure, in this network architecture, first, acquire monitoring images of each road collected by cameras of intelligent street lamp poles deployed on each road; then, respectively pass the monitoring images of each road through a first convolutional neural network model as a filter to obtain a plurality of road monitoring feature vectors; then, construct a spatial connectivity topological matrix between each road, wherein the values of each position at non-diagonal positions in the spatial connectivity topological matrix are used to represent whether two corresponding roads are connected; then, pass the spatial connectivity topological matrix through a second convolutional neural network model as a feature extractor to obtain a spatial connectivity topological feature matrix; then, arrange the plurality of road monitoring feature vectors two-dimensionally to obtain a road monitoring global feature matrix; then, pass the road monitoring global feature matrix and the spatial connectivity topological feature matrix through a graph neural network to obtain a topological road monitoring global feature matrix; then, perform feature distribution correction on the topological road monitoring global feature matrix to obtain a classification feature matrix; finally, pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to represent a grade label of the road congestion degree.

[0077] More specifically, in step S110, the monitoring images of each road are acquired by the cameras of the intelligent street lamp poles deployed on each road. When collecting and analyzing traffic environment information, the monitoring images of each road can be collected by the cameras of the intelligent street lamp poles deployed on each road. However, since there is a spatial correlation between the monitoring image features of each road, that is, there is a spatial correlation feature of whether each two roads among the multiple roads are connected, therefore, this spatial correlation feature is used to strengthen the feature expression of the monitoring images of each road to improve the accuracy of road information recognition. Based on this, an artificial intelligence monitoring technology based on deep learning is adopted to extract the implicit feature distribution information about road information in the monitoring images of each road, and the spatial connectivity topological features between each road are used to strengthen the expression of the image features of each road, so as to identify and judge the degree of road congestion.

[0078] More specifically, in step S120, the monitoring images of each road are respectively passed through the first convolutional neural network model as a filter to obtain multiple road monitoring feature vectors. The first convolutional neural network model as a filter, which has excellent performance in extracting implicit features of images, is used to mine the features of the monitoring images of each road, so as to respectively extract the implicit feature distribution information about road information in the monitoring images of each road, thereby obtaining multiple road monitoring feature vectors.

[0079] Correspondingly, in a specific example, the step of passing the monitoring images of each road through the first convolutional neural network model as a filter to obtain multiple road monitoring feature vectors includes: using each layer of the first convolutional neural network model as a filter to respectively perform the following operations on the input data during the forward pass of the layer: performing convolutional processing on the input data to obtain a convolutional feature map; performing mean pooling processing based on a feature matrix on the convolutional feature map to obtain a pooled feature map; and performing non-linear activation on the pooled feature map to obtain an activation feature map; wherein, the output of the last layer of the first convolutional neural network model as a filter is the multiple road monitoring feature vectors, and the input of the first layer of the first convolutional neural network model as a filter is the monitoring images of each road.

[0080] More specifically, in step S130, a spatial connectivity topology matrix between the respective roads is constructed, wherein the values at the non-diagonal positions in the spatial connectivity topology matrix are used to represent whether the corresponding two roads are connected. Considering that when monitoring road information, due to the characteristic distribution of the relevance in space between the respective roads, that is, there is a relevance relationship of whether the respective roads are connected. Therefore, in the technical solution of the present application, in order to improve the global monitoring of the respective road information, so as to improve the accuracy of judging the road congestion degree, the spatial connectivity topology features between the respective roads are further used to optimize the expression of the monitoring image features of the respective roads.

[0081] More specifically, in step S140, the spatial connectivity topology matrix is passed through a second convolutional neural network model serving as a feature extractor to obtain a spatial connectivity topology feature matrix. The spatial connectivity topology matrix is subjected to feature mining in the second convolutional neural network model serving as a feature extractor to extract the spatial connectivity topology feature distribution information between the respective roads, so as to obtain a spatial connectivity topology feature matrix.

[0082] Correspondingly, in a specific example, the passing the spatial connectivity topology matrix through a second convolutional neural network model serving as a feature extractor to obtain a spatial connectivity topology feature matrix includes: using each layer of the second convolutional neural network model to perform convolution processing, pooling processing along the channel dimension, and non-linear activation processing on the input data respectively in the forward pass of the layer, so that the spatial connectivity topology feature matrix is output by the last layer of the second convolutional neural network model, wherein the input of the first layer of the second convolutional neural network model is the spatial connectivity topology matrix.

[0083] More specifically, in step S150, the multiple road monitoring feature vectors are arranged two-dimensionally to obtain a road monitoring global feature matrix.

[0084] More specifically, in step S160, the road monitoring global feature matrix and the spatial connectivity topology feature matrix are passed through a graph neural network to obtain a topological road monitoring global feature matrix.

[0085] Using each of the multiple road monitoring feature vectors as the feature representation of nodes, and using the spatial connectivity topological feature moment as the feature representation of the edges between nodes, the road monitoring global feature matrix obtained by two-dimensionally arranging the multiple road monitoring feature vectors and the spatial connectivity topological feature matrix are passed through a graph neural network to obtain a topological road monitoring global feature matrix. Specifically, the graph neural network encodes the road monitoring global feature matrix and the spatial connectivity topological feature matrix through learnable neural network parameters to obtain the topological road monitoring global feature matrix containing the spatial connectivity topological features of the respective irregular roads and the high-dimensional implicit features of the respective road images.

[0086] Correspondingly, in a specific example, the step of passing the road monitoring global feature matrix and the spatial connectivity topological feature matrix through a graph neural network to obtain a topological road monitoring global feature matrix includes: the graph neural network processes the road monitoring global feature matrix and the spatial connectivity topological feature matrix through learnable neural network parameters to obtain the topological road monitoring global feature matrix containing the irregular spatial connectivity topological features and high-dimensional road monitoring image features.

[0087] Particularly, in the technical solution of this application, each topological road monitoring global feature vector of the topological road monitoring global feature matrix is a row vector of the topological road monitoring global feature matrix, and this row vector is a topological association expression of the image feature semantics obtained by a single camera relative to the camera space topology. Therefore, it has a certain degree of independence. In this way, within the overall feature distribution of the topological road monitoring global feature matrix arranged by the topological road monitoring global feature vectors, there will be a negative correlation relationship between local distributions, thereby affecting the accuracy of the classification result obtained by the topological road monitoring global feature matrix through the classifier. Therefore, the applicant of this application uses the method of fully positive projection non-linear reweighting to correct the topological road monitoring global feature matrix.

[0088] More specifically, in step S170, the feature distribution of the topological road monitoring global feature matrix is corrected to obtain a classification feature matrix.

[0089] Correspondingly, in a specific example, the step of correcting the feature distribution of the topological road monitoring global feature matrix to obtain a classification feature matrix includes: correcting the feature distribution of the topological road monitoring global feature matrix with the following formula to obtain the classification feature matrix; where the formula is:

[0090]

[0091] where M and M' are the global feature matrix of the topological road monitoring and the classification feature matrix respectively, ReLU(·) represents the ReLU activation function, represents matrix multiplication, and the division between the numerator matrix and the denominator matrix is the element-wise division of matrix eigenvalues. exp(·) represents the exponential operation of a matrix, and the exponential operation of the matrix means calculating the natural exponential function values with the eigenvalues at each position in the matrix as the exponents.

[0092] Specifically, the fully positive projection non-linear reweighting uses the ReLU function to ensure the full positivity of the projection to avoid aggregating negatively correlated information, and at the same time introduces a non-linear reweighting mechanism to aggregate the local eigenvalue distributions of the global feature matrix of the topological road monitoring with each other, so that the internal structure of the global feature matrix of the topological road monitoring can penalize long-distance connections and strengthen local coupling. In this way, the synergistic effect of the spatial feature transformation corresponding to the fully positive projection reweighting of the global feature matrix of the topological road monitoring in the high-dimensional feature space is achieved, thereby improving the accuracy of the classification result obtained by the global feature matrix of the topological road monitoring through the classifier.

[0093] More specifically, in step S180, the classification feature matrix is passed through a classifier to obtain a classification result, and the classification result is used to represent the grade label of the road congestion level.

[0094] Correspondingly, in a specific example, as Figure 4 shown, passing the classification feature matrix through a classifier to obtain a classification result, where the classification result is used to represent the grade label of the road congestion level, includes: S181, expanding the classification feature matrix into a classification feature vector according to row vectors or column vectors; S182, using multiple fully connected layers of the classifier to perform fully connected encoding on the classification feature vector to obtain an encoded classification feature vector; and S183, passing the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0095] In summary, the intelligent transportation control method based on street lamp poles according to the embodiments of the present application uses artificial intelligence monitoring technology based on deep learning to extract the implicit feature distribution information about road information in the monitoring images of each road, and uses the spatial connectivity topological features between each road to strengthen the expression of the image features of each road, so as to identify and judge the road congestion level. In this way, the traffic environment information can be accurately, comprehensively and intelligently detected and judged to realize the construction of a smart city, thereby improving the efficiency of resource utilization and optimizing urban management and services.

[0096] Exemplary system

[0097] Figure 5 It is a block diagram of an intelligent transportation control system 100 based on a street lamp pole according to an embodiment of the present application. As Figure 5 shown, the intelligent transportation control system 100 based on a street lamp pole according to an embodiment of the present application includes: a monitoring image acquisition module 110, configured to acquire monitoring images of each road collected by cameras of intelligent street lamp poles deployed on each road; a first convolution module 120, configured to respectively pass the monitoring images of each road through a first convolutional neural network model as a filter to obtain a plurality of road monitoring feature vectors; a topological matrix construction module 130, configured to construct a spatial connectivity topological matrix between each road, where the values at each position in the non-diagonal positions of the spatial connectivity topological matrix are used to represent whether there is connectivity between the corresponding two roads; a second convolution module 140, configured to pass the spatial connectivity topological matrix through a second convolutional neural network model as a feature extractor to obtain a spatial connectivity topological feature matrix; a feature matrix arrangement module 150, configured to perform two-dimensional arrangement on the plurality of road monitoring feature vectors to obtain a road monitoring global feature matrix; a graph neural encoding module 160, configured to pass the road monitoring global feature matrix and the spatial connectivity topological feature matrix through a graph neural network to obtain a topological road monitoring global feature matrix; a feature distribution correction module 170, configured to perform feature distribution correction on the topological road monitoring global feature matrix to obtain a classification feature matrix; and a classification result generation module 180, configured to pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to represent a level label of the road congestion degree.

[0098] In one example, in the above-mentioned intelligent transportation control system 100 based on a street lamp pole, the first convolution module 120 is further configured to: use each layer of the first convolutional neural network model as a filter to respectively perform the following operations on the input data during the forward pass of the layer: perform convolution processing on the input data to obtain a convolution feature map; perform mean pooling processing based on the feature matrix on the convolution feature map to obtain a pooling feature map; and perform non-linear activation on the pooling feature map to obtain an activation feature map; where the output of the last layer of the first convolutional neural network model as a filter is the plurality of road monitoring feature vectors, and the input of the first layer of the first convolutional neural network model as a filter is the monitoring images of each road.

[0099] In one example, in the above-mentioned streetlight-pole-based intelligent transportation control system 100, the second convolution module 140 is further configured to: use each layer of the second convolutional neural network model to perform convolution processing, pooling processing along the channel dimension, and non-linear activation processing on the input data respectively during the forward pass of the layer, so as to output the spatial connectivity topology feature matrix by the last layer of the second convolutional neural network model, where the input of the first layer of the second convolutional neural network model is the spatial connectivity topology matrix.

[0100] In one example, in the above-mentioned streetlight-pole-based intelligent transportation control system 100, the graph neural encoding module 160 is further configured to: process the road monitoring global feature matrix and the spatial connectivity topology feature matrix through learnable neural network parameters to obtain the topological road monitoring global feature matrix including irregular spatial connectivity topology features and high-dimensional road monitoring image features.

[0101] In one example, in the above-mentioned streetlight-pole-based intelligent transportation control system 100, the feature distribution correction module 170 is further configured to: perform feature distribution correction on the topological road monitoring global feature matrix according to the following formula to obtain the classification feature matrix; where the formula is:

[0102]

[0103] where M and M′ are the topological road monitoring global feature matrix and the classification feature matrix respectively, ReLU(·) represents the ReLU activation function, represents matrix multiplication, and the division between the numerator matrix and the denominator matrix is the element-wise division of matrix eigenvalues, exp(·) represents the exponential operation of the matrix, and the exponential operation of the matrix represents calculating the natural exponential function values with the eigenvalues at each position in the matrix as the exponents.

[0104] In one example, in the above-mentioned streetlight-pole-based intelligent transportation control system 100, the classification result generation module is further configured to: expand the classification feature matrix into a classification feature vector according to a row vector or a column vector; perform fully connected encoding on the classification feature vector using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; and pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

[0105] Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the above-mentioned streetlight-pole-based intelligent transportation control system 100 have been introduced in detail in the description of the Figures 1 to 4 streetlight-pole-based intelligent transportation control method, and therefore, the repeated description thereof will be omitted.

[0106] As described above, the intelligent transportation control system 100 based on street lamp poles according to the embodiments of the present application can be implemented in various wireless terminals, such as a server with an intelligent transportation control algorithm based on street lamp poles. In one example, the intelligent transportation control system 100 based on street lamp poles according to the embodiments of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the intelligent transportation control system 100 based on street lamp poles can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the intelligent transportation control system 100 based on street lamp poles can also be one of the many hardware modules of the wireless terminal.

[0107] Alternatively, in another example, the intelligent transportation control system 100 based on street lamp poles and the wireless terminal can also be separate devices, and the intelligent transportation control system 100 based on street lamp poles can be connected to the wireless terminal through a wired and / or wireless network, and transmit interaction information according to a predefined data format.

[0108] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present application. In addition, the above-disclosed specific details are only for the purpose of illustration and easy understanding, rather than limitations, and the above details do not limit the present application to necessarily adopt the above specific details for implementation.

[0109] The block diagrams of the devices, apparatuses, equipment, and systems involved in the present application are only illustrative examples and do not intend to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open-ended words, meaning "including but not limited to", and can be used interchangeably with each other. The words "or" and "and" used herein refer to the word "and / or", and can be used interchangeably with each other, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to", and can be used interchangeably with each other.

[0110] It should also be noted that in the devices, equipment, and methods of the present application, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent solutions of the present application.

[0111] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Thus, the present application is not intended to be limited to the aspects shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0112] The above description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the present application to the form disclosed herein. Although numerous example aspects and embodiments have been discussed above, those skilled in the art will recognize some of their variations, modifications, alterations, additions, and subcombinations.

Claims

1. A smart traffic control method based on street lamp poles, characterized in that, Including: Obtaining monitoring images of each road collected by cameras of intelligent street lamp poles deployed on each road; Respectively passing the monitoring images of each road through a first convolutional neural network model as a filter to obtain multiple road monitoring feature vectors; Constructing a spatial connectivity topology matrix between each road, where the values of each position in the non-diagonal positions of the spatial connectivity topology matrix are used to represent whether the corresponding two roads are connected; Passing the spatial connectivity topology matrix through a second convolutional neural network model as a feature extractor to obtain a spatial connectivity topology feature matrix; Performing two-dimensional arrangement on the multiple road monitoring feature vectors to obtain a road monitoring global feature matrix; Passing the road monitoring global feature matrix and the spatial connectivity topology feature matrix through a graph neural network to obtain a topological road monitoring global feature matrix; Performing feature distribution correction on the topological road monitoring global feature matrix to obtain a classification feature matrix; Passing the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to represent a grade label of road congestion degree.

2. The intelligent transportation control method based on a street lamp pole according to claim 1, wherein The step of respectively passing the monitoring images of each road through a first convolutional neural network model as a filter to obtain multiple road monitoring feature vectors includes: using each layer of the first convolutional neural network model as a filter to respectively perform the following operations on the input data during the forward pass of the layer: Performing convolutional processing on the input data to obtain a convolutional feature map; Performing mean pooling processing based on a feature matrix on the convolutional feature map to obtain a pooled feature map; Performing non-linear activation on the pooled feature map to obtain an activated feature map; Wherein, the output of the last layer of the first convolutional neural network model as a filter is the multiple road monitoring feature vectors, and the input of the first layer of the first convolutional neural network model as a filter is the monitoring images of each road.

3. The intelligent transportation control method based on a street lamp pole according to claim 1, wherein The step of passing the spatial connectivity topology matrix through a second convolutional neural network model as a feature extractor to obtain a spatial connectivity topology feature matrix includes: using each layer of the second convolutional neural network model to respectively perform convolutional processing, pooling processing along the channel dimension, and non-linear activation processing on the input data during the forward pass of the layer, and outputting the spatial connectivity topology feature matrix by the last layer of the second convolutional neural network model, where the input of the first layer of the second convolutional neural network model is the spatial connectivity topology matrix.

4. The intelligent transportation control method based on a street lamp pole according to claim 1, wherein The step of passing the road monitoring global feature matrix and the spatial connectivity topology feature matrix through a graph neural network to obtain a topological road monitoring global feature matrix includes: The graph neural network processes the road monitoring global feature matrix and the spatial connectivity topology feature matrix through learnable neural network parameters to obtain the topological road monitoring global feature matrix including irregular spatial connectivity topology features and high-dimensional road monitoring image features.

5. The intelligent transportation control method based on a street lamp pole according to claim 1, characterized in that, The step of performing feature distribution correction on the topological road monitoring global feature matrix to obtain a classification feature matrix includes: The global feature matrix of the topological road monitoring is corrected for feature distribution by the following formula to obtain the classification feature matrix; wherein, the formula is: Wherein, M and M' are respectively the global feature matrix of the topological road monitoring and the classification feature matrix, ReLU(·) represents the ReLU activation function, represents matrix multiplication, and the division between the numerator matrix and the denominator matrix is the element-wise division of matrix eigenvalues. exp(·) represents the exponential operation of the matrix. The exponential operation of the matrix means calculating the natural exponential function values with the eigenvalues at each position in the matrix as the exponents. M T is the transpose matrix of M.

6. The intelligent transportation control method based on a street lamp pole according to claim 1, characterized in that The classification feature matrix is passed through a classifier to obtain a classification result, and the classification result is used to represent the grade label of the road congestion degree, including: The classification feature matrix is expanded into a classification feature vector according to a row vector or a column vector; The classification feature vector is subjected to fully connected encoding by multiple fully connected layers of the classifier to obtain an encoded classification feature vector; The encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result.

7. An intelligent transportation control system based on a street lamp pole, characterized in that, Including: A monitoring image acquisition module, configured to acquire monitoring images of each road collected by cameras of intelligent street lamp poles deployed on each road; A first convolution module, configured to respectively pass the monitoring images of each road through a first convolutional neural network model as a filter to obtain a plurality of road monitoring feature vectors; A topological matrix construction module, configured to construct a spatial connectivity topological matrix between each road, wherein the values of each position in the non-diagonal positions of the spatial connectivity topological matrix are used to represent whether two corresponding roads are connected; A second convolution module, configured to pass the spatial connectivity topological matrix through a second convolutional neural network model as a feature extractor to obtain a spatial connectivity topological feature matrix; A feature matrix arrangement module, configured to perform two-dimensional arrangement on the plurality of road monitoring feature vectors to obtain a global feature matrix of road monitoring; A graph neural encoding module, configured to pass the global feature matrix of road monitoring and the spatial connectivity topological feature matrix through a graph neural network to obtain a global feature matrix of topological road monitoring; A feature distribution correction module, configured to correct the feature distribution of the global feature matrix of topological road monitoring to obtain a classification feature matrix; A classification result generation module, configured to pass the classification feature matrix through a classifier to obtain a classification result, and the classification result is used to represent the grade label of the road congestion degree.

8. The intelligent transportation control system based on a street lamp pole according to claim 7, characterized in that The first convolution module is further configured to: use each layer of the first convolutional neural network model as a filter to respectively perform the following operations on the input data during the forward propagation of the layer: Perform convolution processing on the input data to obtain a convolution feature map; Perform mean pooling processing based on the feature matrix on the convolution feature map to obtain a pooled feature map; Perform non-linear activation on the pooled feature map to obtain an activated feature map; wherein, the output of the last layer of the first convolutional neural network model as a filter is the plurality of road monitoring feature vectors, and the input of the first layer of the first convolutional neural network model as a filter is the monitoring images of each road.

9. The intelligent transportation control system based on a street lamp pole according to claim 7, characterized in that, The second convolution module is further configured to: use each layer of the second convolutional neural network model to respectively perform convolution processing, pooling processing along the channel dimension, and non-linear activation processing on the input data during the forward propagation of the layer, so as to output the spatial connectivity topological feature matrix by the last layer of the second convolutional neural network model, wherein the input of the first layer of the second convolutional neural network model is the spatial connectivity topological matrix.

10. The intelligent transportation control system based on a street lamp pole according to claim 7, wherein, The graph neural coding module is further configured to: process the road monitoring global feature matrix and the spatial connectivity topological feature matrix through learnable neural network parameters to obtain the topological road monitoring global feature matrix including irregular spatial connectivity topological features and high-dimensional road monitoring image features.

Images