Traffic flow monitoring method and system for traffic control points

Abstract

The application discloses a traffic flow monitoring method and system for a traffic control point, and relates to the technical field of intelligent traffic, and the method comprises the following steps: configuring a monitoring device at the traffic control point, wherein the monitoring device comprises an image acquisition device and a radar device; collecting video images of the traffic control point through the image acquisition device, and collecting radar data of the traffic control point through the radar device; performing feature extraction based on the video images and the radar data, performing multi-dimensional traffic flow change analysis according to a fusion traffic flow related feature set obtained, and dynamically adjusting signal light timing according to a predicted traffic flow change trend to generate a dynamic traffic scheduling scheme. The application solves the technical problems that the existing traffic flow monitoring data source is single, the dynamic response capability is insufficient, and the traffic passing efficiency and safety are low, and achieves the technical effects that the multi-dimensional traffic flow monitoring and signal light timing adjustment are realized through the fusion of image and radar data, and the traffic passing efficiency and safety are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent transportation, in particular to a traffic flow monitoring method and system for traffic control points. BACKGROUND

[0002] With the increase of urban traffic flow, traffic congestion problems are becoming increasingly serious. Existing traffic flow monitoring technologies mainly rely on a single data source, such as video monitoring or radar monitoring. These methods can provide certain traffic flow information, but have problems such as incomplete data and insufficient accuracy. For example, video-based traffic flow monitoring can be affected by lighting conditions and obstructions, while radar monitoring has difficulty in distinguishing vehicle types. In addition, existing technologies also have deficiencies in dynamic traffic scheduling, mostly based on fixed schedules or simple flow data. SUMMARY

[0003] The present application provides a traffic flow monitoring method and system for traffic control points, to solve the technical problems of single data source of existing traffic flow monitoring, insufficient dynamic response capability, and low traffic efficiency and safety.

[0004] In a first aspect, the present application provides a traffic flow monitoring method for traffic control points, the method comprising: configuring a monitoring device at a traffic control point, the monitoring device comprising an image acquisition device and a radar device; acquiring video images of the traffic control point through the image acquisition device, and acquiring radar data of the traffic control point through the radar device; performing feature extraction based on the video images and radar data to obtain a set of fused traffic flow related features; performing multi-dimensional flow change analysis according to the set of fused traffic flow related features to generate a predicted traffic flow change trend; and dynamically adjusting signal timing according to the predicted traffic flow change trend to generate a dynamic traffic scheduling scheme.

[0005] In a second aspect of the present application, a traffic flow monitoring system for a traffic control point is provided, which comprises: a monitoring device configuration module, configured to configure a monitoring device at a traffic control point, the monitoring device comprising an image acquisition device and a radar device; a traffic data acquisition module, configured to acquire video images of the traffic control point by the image acquisition device, and acquire radar data of the traffic control point by the radar device; a feature extraction module, configured to perform feature extraction based on the video images and radar data, and obtain a set of fused traffic flow related features; a multi-dimensional flow change analysis module, configured to perform multi-dimensional flow change analysis according to the set of fused traffic flow related features, and generate a predicted traffic flow change trend; and a signal light timing adjustment module, configured to dynamically adjust signal light timing according to the predicted traffic flow change trend, and generate a dynamic traffic scheduling scheme.

[0006] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0007] The traffic flow monitoring method and system for a traffic control point provided in the present application relate to the technical field of intelligent transportation, which configures an image acquisition device and a radar device at a traffic control point to respectively acquire video images and radar data, generates a set of traffic flow related features through feature extraction fusion, performs multi-dimensional flow change analysis to predict a traffic flow trend, dynamically adjusts signal light timing according to the traffic flow trend, and generates an intelligent traffic scheduling scheme, thereby improving traffic management efficiency and traffic capacity, solving the technical problems of single source of traffic flow monitoring data and insufficient dynamic response capability, and resulting in low traffic efficiency and safety, and achieving the technical effects of multi-dimensional flow monitoring and signal light timing adjustment through fusion of image and radar data, and improving traffic efficiency and safety. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort.

[0009] Figure 1 A traffic flow monitoring method flow chart for a traffic control point provided in the embodiments of the present application is shown in the figure.

[0010] Figure 2 A traffic flow monitoring system structure diagram for a traffic control point provided in the embodiments of the present application is shown in the figure.

[0011] The reference signs are explained as follows: a monitoring device configuration module 11, a traffic data collection module 12, a feature extraction module 13, a multi-dimensional traffic flow change analysis module 14, and a signal timing adjustment module 15. DETAILED DESCRIPTION

[0012] The present application provides a traffic flow monitoring method and system for a traffic control point, which solves the technical problems of single source of traffic flow monitoring data and insufficient dynamic response capability, resulting in low traffic efficiency and safety.

[0013] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0014] It should be noted that the terms "first", "second", and the like in the specification and the above drawings of the present application are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0015] Embodiment one, as shown in the present application provides a traffic flow monitoring method for a traffic control point, which comprises: Figure 1

[0016] P10: configuring a monitoring device at the traffic control point, the monitoring device comprising an image collection device and a radar device.

[0017] Specifically, the monitoring device for traffic flow monitoring is configured at the traffic control point. Specifically, a monitoring device including an image collection device and a radar device needs to be installed at the traffic control point. The image collection device is mainly used to capture real-time video images of the traffic control point, and through high-resolution cameras and optical technology, the appearance of the vehicle, the license plate information and the dynamic changes of the traffic flow are recorded. The radar device detects the speed, distance and direction of the vehicle by emitting and receiving electromagnetic waves, and has the advantages of not being affected by light conditions and being able to work all day.

[0018] ​To ensure the effective operation of the monitoring equipment, the image acquisition device needs to have high frame rate and high resolution characteristics to capture clear and continuous image data in complex traffic environments. At the same time, the radar device can use Doppler radar technology to accurately measure the speed and distance of vehicles and achieve coverage of multiple lanes through multi-beam design. The combination of these two devices provides a rich source of data for subsequent feature extraction and flow analysis.

[0019] During installation, the image acquisition device is usually installed high above the traffic control point, such as a street lamp pole or a dedicated support, to ensure that its field of view can cover the entire traffic control area. The radar device selects appropriate angles and positions according to its working principle to avoid signal interference and ensure accurate detection of vehicles. By reasonably arranging these two monitoring devices, comprehensive monitoring of the traffic control point can be achieved, providing reliable data support for subsequent traffic flow analysis and dynamic scheduling.

[0020] In addition, the monitoring equipment also needs to have data transmission and storage functions, which can transmit the collected image and radar data to the background processing system in real time and store them efficiently. The real-time and completeness of such data is the key to achieving accurate traffic flow prediction and dynamic scheduling.

[0021] P20: Collecting video images of the traffic control point through the image acquisition device and collecting radar data of the traffic control point through the radar device.

[0022] Optionally, in the traffic flow monitoring system, data acquisition is a key step to achieve accurate traffic management. Specifically, the image acquisition device and the radar device are responsible for collecting video images and radar data of the traffic control point, respectively, and these two types of data are the basis for subsequent feature extraction and flow analysis.

[0023] First, the image acquisition device captures real-time video images of the traffic control point through high-resolution cameras. These cameras are usually installed high above the traffic control point, such as a signal light pole or a dedicated monitoring support, to ensure that they can cover multiple lanes and traffic directions. The resolution and frame rate of the camera are important factors affecting image quality. High-resolution cameras can capture clearer images, including vehicle appearance, license plate numbers, and traffic sign details. At the same time, high-frame-rate cameras can record dynamic changes in vehicles, providing more abundant data for subsequent traffic flow analysis.

[0024] To ensure the stability and reliability of image acquisition, image acquisition devices are usually equipped with auto-focus, low-light enhancement, and anti-shake technologies. These technologies can effectively deal with complex environmental conditions, such as insufficient night lighting or vehicle vibration during high-speed driving. In addition, image acquisition devices also preprocess the collected video images, such as removing noise, enhancing contrast, and compressing images, to facilitate subsequent transmission and storage.

[0025] Next, the radar device detects the radar data of the traffic control point by emitting and receiving electromagnetic waves. The working principle of the radar device is based on the Doppler effect, that is, by measuring the frequency change of the reflected wave to determine the speed and distance of the vehicle. The radar device can monitor the driving speed, vehicle spacing, and movement direction of the vehicle in real time.

[0026] The radar device usually adopts a multi-beam design, which can cover multiple lanes at the same time, thereby achieving comprehensive monitoring of traffic flow. In addition, the radar device also has the ability to work all-weather, and can operate stably even in harsh weather conditions (such as rain, fog, snow, etc.). In order to improve the accuracy of radar data, the radar device will perform data calibration and filtering processing to remove possible interference signals.

[0027] The data collected by the image acquisition device and the radar device complement each other, forming a complete traffic flow monitoring data set. Image data provides appearance information of vehicles and visual details of traffic scenes, while radar data provides dynamic information such as speed and distance of vehicles. By combining these two types of data, a more comprehensive reflection of the traffic conditions at the traffic control point can be achieved, providing more abundant information for subsequent feature extraction and flow analysis. During data acquisition, the synchronization of the image acquisition device and the radar device is also very important. Through precise time synchronization, the consistency of image data and radar data in time can be ensured, thereby providing a reliable data foundation for subsequent data fusion and analysis.

[0028] P30: performing feature extraction based on the video image and radar data to obtain a set of fused traffic flow related features.

[0029] Further, the step P30 of the embodiments of the present application further includes:

[0030] P31: constructing a feature extraction unit, the feature extraction unit including an image feature extraction channel and a radar feature extraction channel; P32: based on the feature extraction unit, performing double-channel feature extraction on the video image and radar data, and fusing the extracted image features and radar data features to generate a set of fused traffic flow related features.

[0031] It should be understood that feature extraction is a key step in transforming the collected video images and radar data into traffic flow features that can be used for analysis. The present application builds a dual-channel feature extraction unit to process video images and radar data separately and fuse the extracted features to generate a set of fused traffic flow-related features. This process not only involves independent processing of image and radar data, but also includes collaborative analysis of both types of data to ensure that the final generated feature set can fully reflect the real-time traffic conditions at the traffic control point.

[0032] First, in order to effectively process video images and radar data, the present application designs a feature extraction unit containing two independent channels. The image feature extraction channel focuses on extracting visual information related to traffic flow from video images, such as vehicle appearance, license plate number, vehicle type, and traffic signs, etc. This process uses advanced image processing techniques such as convolutional neural networks (CNN) to extract hierarchical features of images through multiple convolution operations, providing rich visual details for subsequent analysis. The radar feature extraction channel focuses on extracting dynamic information such as vehicle speed, distance and direction of motion from radar data. Radar devices measure the frequency change of reflected waves through the Doppler effect to obtain real-time dynamic data of vehicles. The design of these two channels fully considers the different characteristics of video images and radar data, providing a solid foundation for subsequent feature fusion.

[0033] Next, after completing the independent extraction of image features and radar features, the two types of features are further fused to generate a set of fused traffic flow-related features. The fusion process first matches the feature types of the extracted image features and radar features to ensure their consistency in time and space. For example, by synchronizing time stamps and calibrating spatial positions, image features and radar features at the same time and at the same location are aligned. Subsequently, through feature comparison and fusion algorithms, feature values with high dispersion are filtered out to reduce the impact of noise and abnormal data. This process can use methods such as weighted summation, feature splicing or fusion networks in deep learning to finally generate a set of fused features that integrate visual information and dynamic information. This fusion strategy not only retains the rich details of image data, but also combines the dynamic advantages of radar data, providing more comprehensive and accurate data support for multi-dimensional analysis and prediction of traffic flow.

[0034] Through the above steps, the present application achieves the goal of extracting and fusing traffic flow-related features from video images and radar data, providing a high-quality data foundation for subsequent traffic flow analysis and dynamic scheduling.

[0035] Further, the step P32 of the embodiment of the present application further comprises:

[0036] P32-1: based on the image feature extraction channel of the feature extraction unit, performing semantic segmentation on the video image to generate a pure image set; P32-2: performing key frame extraction and image down-sampling on the pure image set respectively to generate a detail image set and a dynamic feature set; P32-3: performing traffic flow feature recognition according to the detail image set and the dynamic feature set to obtain a first traffic flow feature set; P32-4: based on the radar feature extraction channel of the feature extraction unit, performing feature extraction on the radar data to obtain a second traffic flow feature set.

[0037] In a possible embodiment of the present application, the process of dual-channel feature extraction can be further refined. The video image and radar data are processed through the image feature extraction channel and the radar feature extraction channel respectively, and finally a fusion traffic flow related feature set is generated.

[0038] In the image feature extraction channel, the collected video image is first subjected to semantic segmentation. Semantic segmentation is an advanced image processing technology that classifies each pixel in the image into different semantic categories (such as vehicles, roads, pedestrians, traffic signs, etc.), which can effectively remove background noise and other irrelevant information, thereby generating a pure image set. This process not only highlights the core objects of the traffic control point (such as vehicles and lanes), but also provides a clearer and more accurate image basis for subsequent feature extraction. Through semantic segmentation, the system can focus on the image content directly related to traffic flow, improving the efficiency and accuracy of feature extraction.

[0039] After generating the pure image set, further processing of the image data is carried out to extract more representative and efficient data. First, key frame extraction technology is used to select frames containing important information from continuous video frames. Key frame extraction can reduce redundant data while retaining the motion trajectories of vehicles and the dynamic changes of traffic flow. Subsequently, image down-sampling is performed on the pure image set to reduce the resolution of the images, reducing the amount of data while retaining the core features of the images. Through this process, a detail image set and a dynamic feature set are generated, which are used to capture the appearance details of vehicles and the dynamic changes of traffic flow, respectively.

[0040] Further, based on the generated detail image set and dynamic feature set, traffic flow feature recognition is performed. Deep learning algorithms (such as convolutional neural networks, CNN) are used to detect, classify and count vehicles in the images, while extracting features such as vehicle speed, type and color. These features are integrated into a first traffic flow feature set, providing rich visual information for subsequent flow analysis and prediction. This process not only identifies the features of individual vehicles, but also analyzes the overall state of traffic flow, such as congestion level and vehicle distribution.

[0041] Next, in the radar feature extraction channel, the collected radar data is processed to obtain dynamic information related to traffic flow. The radar device measures the speed, distance, and direction of motion of vehicles by transmitting and receiving electromagnetic waves using the Doppler effect. After filtering and calibration processing, a second traffic flow feature set is generated, which contains key information such as vehicle speed, acceleration, and spacing. The extraction process of radar data emphasizes the capture of dynamic changes, providing real-time motion information for traffic flow analysis and making up for the lack of dynamic perception in image data.

[0042] Through the above steps, the image feature extraction channel and the radar feature extraction channel generate the first and second traffic flow feature sets, respectively. These two feature sets will be further fused to generate a comprehensive traffic flow-related feature set, providing comprehensive data support for subsequent multi-dimensional flow change analysis and dynamic scheduling.

[0043] Further, the embodiment of the present application step P32 further comprises:

[0044] P32-5: Perform feature type matching on the first traffic flow feature set and the second traffic flow feature set to generate a plurality of feature comparison groups; P32-6: Based on the plurality of feature comparison groups, perform feature comparison and fusion, filter out feature values with a dispersion greater than a threshold value, and generate a fused traffic flow-related feature set.

[0045] Optionally, after completing the feature extraction of the image feature extraction channel and the radar feature extraction channel, the present application further performs fusion processing on the first traffic flow feature set and the second traffic flow feature set generated by the two channels.

[0046] In order to effectively fuse image features and radar features, it is necessary to first perform feature type matching on the two feature sets. Specifically, according to the physical meaning and timestamp information of the features, the image features (such as vehicle type, license plate number, vehicle color, etc.) are aligned with the radar features (such as vehicle speed, distance, direction of motion, etc.) to generate a plurality of feature comparison groups. For example, vehicles in the same lane at the same time will have their image features and radar features combined into a comparison group. This process ensures the correspondence of the two types of features in time and space, providing a basis for subsequent fusion analysis.

[0047] After generating the feature sets, further feature comparison and fusion are performed on these feature sets. During the fusion process, the correlation between image features and radar features within each feature set is calculated, and the fused features are evaluated. To ensure the reliability and accuracy of the fused features, the system filters out feature values with a dispersion greater than a pre-set threshold. Dispersion refers to the degree of deviation of a feature value in a statistical distribution, and a feature value with a high dispersion usually indicates noise or abnormal data. By setting a reasonable threshold based on empirical data, these unreliable feature values can be eliminated, and more representative and consistent features can be retained. Finally, after comparison, fusion, and dispersion filtering, a set of fused traffic flow-related features is generated, providing high-quality data support for subsequent traffic flow analysis and dynamic scheduling.

[0048] P40: Based on the set of fused traffic flow-related features, multi-dimensional flow change analysis is performed to generate a predicted traffic flow change trend.

[0049] Further, the step P40 of the embodiments of the present application further includes:

[0050] P41: The set of fused traffic flow-related features includes the number of vehicles passing through, vehicle types, vehicle speeds, and vehicle directions within the pre-audit period; P42: Based on the number of vehicles passing through, vehicle types, vehicle speeds, and vehicle directions within the pre-audit period, multi-dimensional flow change analysis is performed to generate flow change trends in multiple dimensions; P43: The flow change trends in multiple dimensions are fused to generate a predicted traffic flow change trend.

[0051] It should be understood that the key link is to perform multi-dimensional flow change analysis based on the set of fused traffic flow-related features and generate a predicted traffic flow change trend.

[0052] Firstly, the set of fused traffic flow-related features is the basis for traffic flow analysis, containing multi-dimensional information such as the number of vehicles passing through, vehicle types, vehicle speeds, and vehicle directions within the pre-audit period at the traffic control point. These information is obtained by feature extraction and fusion of image and radar data, and can fully reflect the traffic conditions at the traffic control point within the pre-audit period. The number of vehicles reflects the scale of traffic flow; vehicle types reveal the proportion of different types of vehicles; vehicle speeds and directions provide dynamic characteristics of traffic flow. These features provide rich data support for subsequent multi-dimensional flow change analysis.

[0053] Next, based on the number of vehicles, types, speed, and driving direction in the fused traffic flow-related feature set, multi-dimensional flow change analysis is performed. This process generates flow change trends in multiple dimensions by analyzing the change trends in each dimension separately. Specifically, the system first counts the number of vehicles in different time periods within the pre-audit period, analyzes the increase and decrease trends of traffic flow, and determines the traffic peak and valley periods. At the same time, the system analyzes the flow changes of different types of vehicles to understand their proportion changes in traffic flow, providing a basis for traffic planning. In addition, by monitoring the changes in vehicle speed, the system can determine the smoothness and congestion of traffic flow, and the changes in vehicle driving direction help to analyze the distribution and flow direction of traffic flow. These multi-dimensional flow change trends provide detailed and comprehensive information for subsequent comprehensive prediction.

[0054] After completing the multi-dimensional flow change analysis, the flow change trends in multiple dimensions are fused to generate a predicted vehicle flow change trend. This process analyzes the change trends of vehicle number, type, speed, and driving direction, combines the actual needs of traffic management, evaluates the importance of flow change trends in different dimensions, and assigns weights. For example, the change in vehicle speed may be more critical to traffic congestion judgment, so it will be given a higher weight. By weighted summation or other fusion algorithms, the multi-dimensional trends are integrated into a comprehensive trend, combined with historical data and machine learning models, to generate a predicted vehicle flow change trend. This trend not only reflects the dynamic changes of current traffic flow, but also predicts the traffic flow changes in the future, providing forward-looking guidance for traffic scheduling, effectively alleviating traffic congestion, and improving road traffic efficiency.

[0055] Further, the step P43 of the embodiments of the present application further includes:

[0056] P43-1: Based on the flow-related features in multiple dimensions, respectively evaluate the importance and generate flow impact weights for each feature according to the evaluation results; P43-2: According to the flow impact weights, fuse the flow change trends in multiple dimensions to generate a predicted vehicle flow change trend.

[0057] Specifically, after completing the multi-dimensional flow change analysis, how to fuse the flow change trends in multiple dimensions to generate an accurate predicted vehicle flow change trend can be further refined.

[0058] Firstly, to more scientifically fuse the traffic change trends of different dimensions, the importance of each dimension's traffic-related features is evaluated. These features include the number of vehicles, vehicle types, vehicle speeds, and vehicle driving directions, etc. The purpose of importance evaluation is to determine the influence of each feature on traffic flow changes, so as to assign a traffic impact weight to each feature. For example, the change of vehicle speed may be more critical to the judgment of traffic congestion, so it will be given a higher weight; while the change of vehicle type may have an important impact on traffic planning, but less impact on immediate traffic changes, so the weight is relatively low. Through this evaluation, the system can dynamically adjust the weight of each feature according to the actual traffic management needs, ensuring that the fusion process is more scientific and reasonable.

[0059] After determining the traffic impact weight of each feature, the traffic change trends of multiple dimensions are fused according to these weights. Specifically, the change trend of each dimension is multiplied by the corresponding weight, and then these trends are integrated into a comprehensive predicted traffic change trend through weighted summation. This process not only considers the independent changes of each dimension, but also adjusts the contribution of different dimensions to the overall trend through weights. The final generated predicted traffic change trend can more accurately reflect the real-time traffic conditions at the traffic control point, and provide forward-looking guidance for future traffic changes. This weight-based trend fusion method enables the system to dynamically adapt to different traffic scenarios, improving the accuracy and reliability of the prediction. This process provides a scientific basis for dynamic traffic scheduling and signal light optimization, effectively improving the intelligent level of traffic management.

[0060] P50: According to the predicted traffic change trend, dynamically adjust the signal light timing to generate a dynamic traffic scheduling scheme.

[0061] Further, the step P50 of the embodiments of the present application further includes:

[0062] P51: Receive a plurality of predicted traffic change trends of a plurality of traffic control points; P52: Obtain signal light timing information of a target urban road network, a pre-trained traffic scheduling model; P53: According to the traffic scheduling model, combine the plurality of predicted traffic change trends to optimize the signal light timing, and generate the dynamic traffic scheduling scheme.

[0063] Optionally, by dynamically adjusting the signal light timing, a dynamic traffic scheduling scheme suitable for multiple traffic control points is generated, which converts the predicted traffic change trend into actual traffic management measures, thereby realizing intelligent management of urban traffic network.

[0064] To achieve effective management of the entire urban traffic network, first, the predicted traffic flow trends from multiple traffic control points are received. These trends are based on the analysis of the fusion traffic flow-related feature set of each control point, which can reflect the traffic flow changes of each control point in the future period. By integrating the data of multiple control points, the system can comprehensively understand the dynamic situation of the urban traffic network, providing a global perspective for subsequent signal timing optimization and traffic scheduling. This process ensures that traffic management decisions are not based on local conditions at a single intersection, but take into account the dynamic changes of the entire traffic network.

[0065] Further, before generating the dynamic traffic scheduling scheme, the system needs to obtain the current signal timing information of the target urban road network and the pre-trained traffic scheduling model. The signal timing information includes the signal cycle, phase setting, and current green light duration of each intersection, which are the basis for dynamic adjustment and reflect the current traffic management operation status. At the same time, the pre-trained traffic scheduling model is an intelligent model based on historical data and machine learning algorithms, which can quickly generate an optimized signal timing scheme according to the input traffic flow trends. This model learns the patterns and rules in historical traffic data, effectively dealing with complex traffic scenarios, ensuring the scientificity and practicality of the scheduling scheme.

[0066] Further, based on the pre-trained traffic scheduling model, combined with the predicted traffic flow trends of multiple traffic control points, signal timing optimization is performed. Specifically, the predicted traffic flow trends are input into the traffic scheduling model, which analyzes the distribution and changes of future traffic flow based on these trends, and combines the current signal timing information to calculate the optimal signal timing scheme. The optimization goal is to reduce traffic congestion, improve road traffic efficiency, and ensure the smoothness of traffic flow, and generate a dynamic traffic scheduling scheme based on the optimized signal timing scheme. This scheme not only includes signal timing adjustment, but also may involve traffic flow guidance, priority passing strategies, etc. By dynamically adjusting the signal timing, it can respond to traffic flow changes in real time, effectively alleviate traffic congestion, and improve the overall operation efficiency of urban traffic. This process not only improves the flexibility and adaptability of traffic management, but also provides strong support for efficient operation of urban traffic.

[0067] In summary, the embodiments of the present application have at least the following technical effects:

[0068] The present application configures image acquisition devices and radar devices at traffic control points to respectively acquire video images and radar data, generates traffic flow-related feature sets through feature extraction and fusion, analyzes multi-dimensional flow changes to predict traffic trends, and dynamically adjusts signal timing to generate intelligent traffic scheduling schemes, improving traffic management efficiency and traffic capacity.

[0069] The technical effects of achieving multi-dimensional traffic monitoring and signal timing adjustment by fusing image and radar data, improving traffic efficiency and safety are achieved.

[0070] In the second embodiment, based on the same inventive concept as the traffic flow monitoring method for traffic control points in the foregoing embodiments, the application provides a traffic flow monitoring system for traffic control points, and the system and method embodiments in the application are based on the same inventive concept. The system comprises: Figure 2

[0071] The monitoring device configuration module 11 is configured to configure a monitoring device at a traffic control point, and the monitoring device comprises an image acquisition device and a radar device.

[0072] The traffic data acquisition module 12 is configured to acquire video images of the traffic control point through the image acquisition device and radar data of the traffic control point through the radar device.

[0073] The feature extraction module 13 is configured to perform feature extraction based on the video images and radar data to obtain a set of fused traffic flow related features.

[0074] The multi-dimensional traffic change analysis module 14 is configured to perform multi-dimensional traffic change analysis according to the set of fused traffic flow related features to generate a predicted traffic flow change trend.

[0075] The signal timing adjustment module 15 is configured to dynamically adjust the signal timing according to the predicted traffic flow change trend to generate a dynamic traffic scheduling scheme.

[0076] Further, the feature extraction module 13 is further configured to perform the following steps:

[0077] The feature extraction module 13 is further configured to perform the following steps:

[0078] Further, the feature extraction module 13 is further configured to perform the following steps:

[0079] ​Based on the image feature extraction channel of the feature extraction unit, the video image is semantically segmented to generate a pure image set; the pure image set is respectively subjected to key frame extraction and image down-sampling to generate a detail image set and a dynamic feature set; traffic flow feature recognition is performed according to the detail image set and the dynamic feature set to obtain a first traffic flow feature set; based on the radar feature extraction channel of the feature extraction unit, feature extraction is performed on the radar data to obtain a second traffic flow feature set.

[0080] Further, the feature extraction module 13 is further configured to perform the following steps:

[0081] The first traffic flow feature set and the second traffic flow feature set are subjected to feature type matching to generate a plurality of feature comparison groups; based on the plurality of feature comparison groups, feature comparison fusion is performed to screen out feature values with a dispersion greater than a threshold value to generate a fusion traffic flow related feature set.

[0082] Further, the multi-dimensional flow change analysis module 14 is further configured to perform the following steps:

[0083] The fusion traffic flow related feature set includes the number of vehicles passing through, vehicle types, vehicle speeds and vehicle driving directions within a pre-trial period; based on the number of vehicles passing through, vehicle types, vehicle speeds and vehicle driving directions within the pre-trial period, multi-dimensional flow change analysis is performed to generate a plurality of dimensional flow change trends; the plurality of dimensional flow change trends are fused to generate a predicted vehicle flow change trend.

[0084] Further, the multi-dimensional flow change analysis module 14 is further configured to perform the following steps:

[0085] Based on the multi-dimensional flow related features, importance evaluation is respectively performed, and flow impact weights of each feature are generated according to the evaluation results; according to the flow impact weights, the plurality of dimensional flow change trends are fused to generate a predicted vehicle flow change trend.

[0086] Further, the signal light timing adjustment module 15 is further configured to perform the following steps:

[0087] A plurality of predicted vehicle flow change trends of a plurality of traffic control points are received; signal light timing information of a target urban road network and a pre-trained traffic scheduling model are obtained; according to the traffic scheduling model, signal light timing optimization is performed in combination with the plurality of predicted vehicle flow change trends to generate the dynamic traffic scheduling scheme.

[0088] It should be noted that the above-mentioned embodiment sequences of the present application are merely for description only, but not for representing the advantages and disadvantages of the embodiments. And the above-mentioned embodiments of the present specification have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0089] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0090] The specification and drawings are merely exemplary of the present application, and any and all modifications, variations, combinations or equivalents that are within the scope of the present application should be included. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A traffic flow monitoring method for a traffic control point, characterized by, The method comprises: configuring a monitoring device at a traffic control point, the monitoring device comprising an image acquisition device and a radar device; acquiring video images of the traffic control point by the image acquisition device and radar data of the traffic control point by the radar device; based on the video images and radar data, feature extraction is performed to obtain a set of fused traffic flow-related features, which comprises: constructing a feature extraction unit, the feature extraction unit comprising an image feature extraction channel and a radar feature extraction channel; based on the feature extraction unit, dual-channel feature extraction is performed on the video images and radar data, and the extracted image features and radar data features are fused to generate a set of fused traffic flow-related features; based on the feature extraction unit, dual-channel feature extraction is performed on the video images and radar data, which comprises: based on the image feature extraction channel of the feature extraction unit, semantic segmentation is performed on the video images to generate a set of pure images; the set of pure images is respectively subjected to key frame extraction and image down-sampling to generate a set of detail images and a set of dynamic features; traffic flow feature recognition is performed according to the set of detail images and the set of dynamic features to obtain a first set of traffic flow features; based on the radar feature extraction channel of the feature extraction unit, feature extraction is performed on the radar data to obtain a second set of traffic flow features; fusing the extracted image features and radar data features comprises: performing feature type matching on the first set of traffic flow features and the second set of traffic flow features to generate a plurality of feature comparison groups; based on the plurality of feature comparison groups, feature comparison and fusion are performed to filter out feature values with a dispersion greater than a threshold to generate a set of fused traffic flow-related features; based on the set of fused traffic flow-related features, multi-dimensional flow change analysis is performed to generate a predicted vehicle flow change trend, which comprises: the set of fused traffic flow-related features includes the number of vehicles passing through, vehicle types, vehicle speeds, and vehicle directions within a pre-trial period; based on the number of vehicles passing through, vehicle types, vehicle speeds, and vehicle directions within the pre-trial period, multi-dimensional flow change analysis is performed to generate flow change trends in multiple dimensions; fusing the flow change trends in multiple dimensions to generate a predicted vehicle flow change trend; fusing the flow change trends in multiple dimensions to generate a predicted vehicle flow change trend comprises: based on the flow-related features in multiple dimensions, importance evaluation is respectively performed, and flow impact weights of each feature are generated according to the evaluation results; based on the flow impact weights, the flow change trends in multiple dimensions are fused to generate a predicted vehicle flow change trend; based on the set of fused traffic flow-related features, multi-dimensional flow change analysis is performed to generate a predicted vehicle flow change trend; based on the predicted vehicle flow change trend, signal timing is dynamically adjusted to generate a dynamic traffic scheduling scheme.

2. The traffic flow monitoring method for a traffic control point as claimed in claim 1, wherein, based on the predicted vehicle flow change trend, signal timing is dynamically adjusted to generate a dynamic traffic scheduling scheme, which comprises: receiving a plurality of predicted vehicle flow change trends of a plurality of traffic control points; Obtaining signal timing information of a target urban road network, pre-training a traffic scheduling model; According to the traffic scheduling model, combining the multiple predicted traffic flow change trends to optimize the signal timing, and generating the dynamic traffic scheduling scheme.

3. Traffic flow monitoring system for a traffic control point for implementing the method according to claim 1, characterized in that The system comprises: A monitoring device configuration module is configured to configure a monitoring device at a traffic control point, the monitoring device comprising an image acquisition device and a radar device; A traffic data acquisition module is configured to acquire video images of the traffic control point through the image acquisition device and radar data of the traffic control point through the radar device; A feature extraction module is configured to extract features based on the video images and radar data, and obtain a set of fused traffic flow related features; A multi-dimensional flow change analysis module is configured to analyze multi-dimensional flow changes based on the set of fused traffic flow related features, and generate predicted traffic flow change trends; A signal timing adjustment module is configured to dynamically adjust signal timing based on the predicted traffic flow change trends, and generate a dynamic traffic scheduling scheme.

Images