Road violation behavior monitoring method based on high-throughput computing technology

Abstract

The invention provides a road illegal behavior monitoring method based on a high-throughput computing technology, and the method comprises the steps: building an intelligent management and control platform, deploying video collection equipment and a sensor, carrying out algorithm training and optimization, and storing and managing data, and transmitting a road scene video to the intelligent management and control platform in real time. Illegal behaviors such as red light running of motor vehicles or non-motor vehicles, retrograde running of motor vehicles or non-motor vehicles, slow running of trucks and the like are recognized through an AI intelligent algorithm, and mass vehicle passing data are collected, stored, processed and analyzed in real time by using a big data analysis technology; a multi-modal data fusion algorithm, a dynamic resource allocation strategy and an adjustment extraction and prediction model based on deep learning are adopted, and a video monitoring and recognition technology is combined, so that the traffic monitoring efficiency and accuracy are improved, and powerful support is provided for traffic management and decision making.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a road violation monitoring method based on high-throughput computing technology. Background Art

[0002] With the continuous development of urban transportation, traffic flow is increasing, and the number of violations of road motor vehicles and non-motor vehicles is also increasing. The traditional traffic road monitoring method mainly relies on electronic monitoring equipment to capture and manual verification to confirm. Many places need to install a lot of equipment to monitor road traffic, such as cameras specifically for capturing violations, video cameras for monitoring road traffic conditions, equipment that needs to install geomagnetic coils and radar sensing linkage cameras, fill light equipment, etc., which have large data traffic, many control points, difficult equipment maintenance, high professional requirements for personnel, and strong influence of human subjective factors during judgment. It is difficult to comprehensively and timely discover and deal with various violations. To this end, the applicant proposes a road violation monitoring method based on high-throughput computing technology, which can monitor road violations more efficiently and accurately, effectively improve the level of traffic management and ensure traffic safety. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a road violation monitoring method based on high-throughput computing technology, by building an intelligent management and control platform, deploying video acquisition equipment and sensors, algorithm training and optimization, and data storage and management, by transmitting road scene video to the intelligent management and control platform in real time, and using AI intelligent algorithms to identify illegal and irregular behaviors such as motor or non-motor vehicles running red lights, motor or non-motor vehicles driving in the wrong direction, and slow-moving trucks, and use big data analysis technology to collect, store, process and analyze massive vehicle passing data in real time; adopt multimodal data fusion algorithms, dynamic resource allocation strategies and adjustment extraction and prediction models based on deep learning, and combine video monitoring and recognition technologies to improve traffic monitoring efficiency and accuracy, and provide strong support for traffic management and decision-making.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] A road violation monitoring method based on high-throughput computing technology, characterized in that: the road violation monitoring method based on high-throughput computing is mainly divided into building an intelligent management and control platform, deploying video acquisition equipment and sensors, algorithm training and optimization, and data storage and management, specifically:

[0006] Build an intelligent management and control platform: Build a server cluster with powerful computing and storage capabilities to receive and process road scene video data and vehicle passing data;

[0007] Deploy video acquisition equipment and sensors: Install high-definition cameras, speed sensors and other equipment at road intersections and sections to ensure full coverage of the monitoring area;

[0008] Algorithm training and optimization: Use a large amount of labeled road violation data to train the AI ​​intelligent algorithm, and continuously optimize the algorithm model to improve recognition accuracy.

[0009] Data storage and management: Establish an efficient database system to classify, store and manage the collected video data and vehicle passing data for quick retrieval and analysis.

[0010] Furthermore, the road violation monitoring method based on high-throughput computing specifically comprises the following steps:

[0011] Step 1: Real-time collection and transmission of road scene videos: Deploy video collection equipment at key locations on the road to transmit the scene videos to the intelligent management and control platform in real time;

[0012] Step 2: Identification of illegal behaviors based on AI intelligent algorithms: Use deep learning algorithms to analyze the transmitted videos in real time and identify various illegal behaviors;

[0013] Step 3: Application of big data analysis technology: real-time collection, storage, processing and analysis of massive amounts of vehicle data, accurate identification of vehicle information, and in-depth analysis of vehicle behavior using data mining and machine learning algorithms;

[0014] Step 4: Integration of video surveillance and recognition technology: Adopt high-definition video surveillance technology and image recognition technology, deploy high-definition cameras to achieve real-time monitoring of key areas, and use image recognition technology to automatically identify and track vehicles and pedestrians in surveillance videos.

[0015] Furthermore, the violation identification based on AI intelligent algorithm in step 2 of road violation based on high-throughput computing specifically includes multimodal data fusion algorithm, dynamic resource allocation strategy and feature extraction and prediction model based on deep learning.

[0016] Furthermore, the multimodal data fusion algorithm of the road violation monitoring method based on high-throughput computing is specifically:

[0017] The image features in the video data are deeply integrated with the vehicle information in the vehicle passing data. The vehicle feature vector in the video data is defined as V f =[v 1 , v 2 , ..., v n ], the vehicle attribute vector in the vehicle passing data is defined as D f =[d 1 , d 2, ..., d m , then through the fusion function F(V f , D f ) to achieve data fusion, the formula is:

[0018] F(V f , D f )=ω 1 V f +ω 2 D f

[0019] Where: 1 and ω 2 is the weight coefficient, and the optimal value is determined through a large amount of data training.

[0020] Furthermore, the dynamic resource allocation strategy of the road violation monitoring method based on high-throughput computing is specifically as follows:

[0021] According to the traffic flow changes in different time periods and road sections, when the estimated data processing volume of a road section in a certain time period is L t , the total amount of available computing resources is R t , through the resource allocation function A(L t , R t ) to dynamically allocate computing resources, the formula is:

[0022] A(L t , R t )=αL t / R t

[0023] Where: α is the adjustment coefficient.

[0024] Furthermore, the feature extraction and prediction model based on deep learning of the road violation monitoring method based on high-throughput computing is specifically:

[0025] Using deep learning technology, a feature extraction and prediction model specifically for road scenes is constructed. The convolutional neural network is: when the input video image is I, after multiple layers of convolution and pooling operations, the feature map F(I) is obtained; the formula is:

[0026] F(I)=Conv(I)×Pool(I)

[0027] The characteristics of the time series are analyzed through long short-term memory networks to predict the behavior trends of vehicles, thereby detecting potential violations in advance.

[0028] The benefits of this application are:

[0029] 1. The road violation monitoring method based on high-throughput computing can process and analyze road data more efficiently and intelligently, realize accurate and real-time monitoring of road violations, and reduce missed and misjudgment;

[0030] 2. The road violation monitoring method based on high-throughput computing can effectively improve processing efficiency while saving computing resources to the greatest extent, improve the overall performance and adaptability of the system, reduce the workload of manual monitoring, and improve the automation level of traffic management;

[0031] 3. The road violation monitoring method based on high-throughput computing can determine the optimal value through a large amount of data training. The fusion algorithm can more comprehensively and accurately describe vehicle behavior and improve the accuracy of identifying violations.

[0032] 4. The road violation monitoring method based on high-throughput computing provides scientific decision-making basis for traffic management departments based on big data analysis results, rationally plans traffic facilities and adjusts traffic strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic diagram of the working process of the present invention. DETAILED DESCRIPTION

[0034] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0035] like Figure 1 As shown, a road violation monitoring method based on high-throughput computing technology is shown. The road violation monitoring method based on high-throughput computing is mainly divided into building an intelligent management and control platform, deploying video acquisition equipment and sensors, algorithm training and optimization, and data storage and management, specifically:

[0036] Build an intelligent management and control platform: Build a server cluster with powerful computing and storage capabilities to receive and process road scene video data and vehicle passing data;

[0037] Deploy video acquisition equipment and sensors: Install high-definition cameras, speed sensors and other equipment at road intersections and sections to ensure full coverage of the monitoring area;

[0038] Algorithm training and optimization: Use a large amount of labeled road violation data to train the AI ​​intelligent algorithm, and continuously optimize the algorithm model to improve recognition accuracy.

[0039] Data storage and management: Establish an efficient database system to classify, store and manage the collected video data and vehicle passing data for quick retrieval and analysis.

[0040] The steps of the road violation monitoring method based on high-throughput computing are as follows:

[0041] Step 1: Real-time collection and transmission of road scene videos: Deploy video collection equipment at key locations on the road to transmit the scene videos to the intelligent management and control platform in real time;

[0042] Step 2: Identification of illegal behaviors based on AI intelligent algorithms: Use deep learning algorithms to analyze the transmitted videos in real time and identify various illegal behaviors;

[0043] Step 3: Application of big data analysis technology: real-time collection, storage, processing and analysis of massive amounts of vehicle data, accurate identification of vehicle information, and in-depth analysis of vehicle behavior using data mining and machine learning algorithms;

[0044] Step 4: Integration of video surveillance and recognition technology: Adopt high-definition video surveillance technology and image recognition technology, deploy high-definition cameras to achieve real-time monitoring of key areas, and use image recognition technology to automatically identify and track vehicles and pedestrians in surveillance videos.

[0045] The violation identification based on AI intelligent algorithm in step 2 of road violation based on high-throughput computing shown in the figure specifically includes multimodal data fusion algorithm, dynamic resource allocation strategy and feature extraction and prediction model based on deep learning.

[0046] The multimodal data fusion algorithm of the road violation monitoring method based on high-throughput computing is shown as follows:

[0047] The image features in the video data are deeply integrated with the vehicle information in the vehicle passing data. The vehicle feature vector in the video data is defined as V f =[v 1 , v 2 , ..., v n ], the vehicle attribute vector in the vehicle passing data is defined as D f =[d 1 , d 2 , ..., d m , then through the fusion function F(V f , D f ) to achieve data fusion, the formula is:

[0048] F(V f , D f )=ω 1 V f +ω 2 D f

[0049] Where: 1 and ω 2 is the weight coefficient, and the optimal value is determined through a large amount of data training.

[0050] The dynamic resource allocation strategy of the road violation monitoring method based on high-throughput computing is shown as follows:

[0051] According to the traffic flow changes in different time periods and road sections, when the estimated data processing volume of a road section in a certain time period is L t , the total amount of available computing resources is R t , through the resource allocation function A(L t , R t ) to dynamically allocate computing resources, the formula is:

[0052] A(L t , R t )=αL t / R t

[0053] Where: α is the adjustment coefficient.

[0054] The feature extraction and prediction model based on deep learning of the road violation monitoring method based on high-throughput computing is shown, specifically:

[0055] Using deep learning technology, a feature extraction and prediction model specifically for road scenes is constructed. The convolutional neural network is: when the input video image is I, after multiple layers of convolution and pooling operations, the feature map F(I) is obtained; the formula is:

[0056] F(I)=Conv(I)×Pool(I)

[0057] The characteristics of the time series are analyzed through long short-term memory networks to predict the behavior trends of vehicles, thereby detecting potential violations in advance.

[0058] When the road violation monitoring method based on high-throughput computing is used, the first step is to build an intelligent management and control platform and construct a server cluster with powerful computing and storage capabilities. Then, video acquisition equipment and sensors are deployed. Video acquisition equipment and sensors are installed at key locations on the road, and the original road monitoring equipment can be connected for use. After the installation and docking are completed, the on-site video of the road is collected in real time, and the on-site video is transmitted to the intelligent management and control platform in real time. The violation identification based on the AI ​​intelligent algorithm in the intelligent platform uses a deep learning algorithm to analyze the transmitted video in real time to identify various violations. The vehicle behavior is deeply analyzed by using big data analysis technology mining and machine learning algorithms to predict the behavior of the vehicle. If necessary, it can be reminded and corrected through the on-site guidance audio equipment to avoid traffic accidents. The monitoring method integrates video monitoring and recognition technology, conducts real-time monitoring in key areas, and automatically identifies and tracks vehicles and pedestrians in the monitoring video. It also makes predictions. In the process, a large amount of labeled road violation data is used to train the AI ​​intelligent algorithm, and the algorithm model is continuously optimized to improve the recognition accuracy. In addition, an efficient database system is established to classify, store and manage the collected video data and vehicle passing data for rapid retrieval and analysis, so as to facilitate timely traffic management.

[0059] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A road violation monitoring method based on high-throughput computing technology, characterized in that: The road violation monitoring method based on high-throughput computing is divided into building an intelligent management and control platform, deploying video acquisition equipment and sensors, algorithm training and optimization, and data storage and management, specifically: Build an intelligent management and control platform: Build a server cluster with powerful computing and storage capabilities to receive and process road scene video data and vehicle passing data; Deploy video acquisition equipment and sensors: Install high-definition cameras, speed sensors and other equipment at road intersections and sections to ensure full coverage of the monitoring area; Algorithm training and optimization: Use a large amount of labeled road violation data to train the AI ​​intelligent algorithm, and continuously optimize the algorithm model to improve recognition accuracy. Data storage and management: Establish an efficient database system to classify, store and manage the collected video data and vehicle passing data for quick retrieval and analysis.

2. A road violation monitoring method based on high-throughput computing technology, characterized in that: The road violation monitoring method based on high-throughput computing specifically comprises the following steps: Step 1: Real-time collection and transmission of road scene videos: Deploy video collection equipment at key locations on the road to transmit the scene videos to the intelligent management and control platform in real time; Step 2: Identification of illegal behaviors based on AI intelligent algorithms: Use deep learning algorithms to analyze the transmitted videos in real time and identify various illegal behaviors; Step 3: Application of big data analysis technology: real-time collection, storage, processing and analysis of massive amounts of vehicle data, accurate identification of vehicle information, and in-depth analysis of vehicle behavior using data mining and machine learning algorithms; Step 4: Integration of video surveillance and recognition technology: Adopt high-definition video surveillance technology and image recognition technology, deploy high-definition cameras to achieve real-time monitoring of key areas, and use image recognition technology to automatically identify and track vehicles and pedestrians in surveillance videos.

3. The method for monitoring road violations based on high-throughput computing technology according to claim 2, characterized in that: The violation identification based on AI intelligent algorithm in step 2 of the road violation based on high-throughput computing specifically includes a multimodal data fusion algorithm, a dynamic resource allocation strategy, and a feature extraction and prediction model based on deep learning.

4. The method for monitoring road violations based on high-throughput computing technology according to claim 3 is characterized in that: The multimodal data fusion algorithm of the road violation monitoring method based on high-throughput computing is specifically: The image features in the video data are deeply integrated with the vehicle information in the vehicle passing data. The vehicle feature vector in the video data is defined as V f =[υ1,υ2,...,υ n ], the vehicle attribute vector in the vehicle passing data is defined as D f =[d1, d2, ..., d m , then through the fusion function F(V f , D f ) to achieve data fusion, the formula is: F(V f D f )=ω1V f +ω2D f Among them: ω1 and ω2 are weight coefficients, and the optimal values ​​are determined through a large amount of data training.

5. The method for monitoring road violations based on high-throughput computing technology according to claim 3 is characterized in that: The dynamic resource allocation strategy of the road violation monitoring method based on high-throughput computing is specifically as follows: According to the traffic flow changes in different time periods and road sections, when the estimated data processing volume of a road section in a certain time period is L t , the total amount of available computing resources is R t , through the resource allocation function A(L t , R t ) to dynamically allocate computing resources, the formula is: A(L t ,R t )=αL t / R t Where: α is the adjustment coefficient.

6. The method for monitoring road violations based on high-throughput computing technology according to claim 3, characterized in that: The feature extraction and prediction model based on deep learning of the road violation monitoring method based on high-throughput computing is specifically: Using deep learning technology, a feature extraction and prediction model specifically for road scenes is constructed. The convolutional neural network is: when the input video image is I, after multiple layers of convolution and pooling operations, the feature map F(I) is obtained; the formula is: F(I)=Conυ(I)×Pool(I) The characteristics of the time series are analyzed through long short-term memory networks to predict the behavior trends of vehicles, thereby detecting potential violations in advance.

Images