Method, device and equipment for identifying road congestion state in real time and storage medium

Abstract

The application discloses a method and device for identifying road congestion state in real time, equipment and storage medium. Including: obtaining trajectory data corresponding to an idling vehicle equipped with a front image device, filtering a target vehicle driving on a road based on the trajectory data; extracting the front view image of the target vehicle, inputting the front view image into a pre-trained traffic object recognition model, and outputting the traffic object recognition result in the image; based on the traffic object recognition result, filtering an initial congestion image and automatically labeling the traffic light category corresponding to the initial congestion image; based on the traffic light category, inputting the initial congestion image into the corresponding congestion recognition model, and outputting the congestion level determination result of the current vehicle. Through image data recognition and congestion level recognition, the application accurately determines the congestion level, improves the accuracy and real-time performance of congestion recognition, provides strong support for traffic management and optimization, and improves road traffic efficiency.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation technology, and more specifically, to a method, apparatus, device, and storage medium for real-time identification of road congestion status. Background Technology

[0002] With the acceleration of urbanization and the continuous increase in the number of motor vehicles, traffic congestion has become increasingly serious, becoming one of the key factors restricting urban development. Traditional traffic congestion monitoring methods mainly rely on fixed traffic cameras and sensors. However, the fixed installation locations of these cameras and sensors make it difficult to cover all roads, especially in areas without traffic lights, where coverage is even lower. Furthermore, large-scale deployment of fixed cameras and sensors requires high installation and maintenance costs, limiting their widespread application. The limited data update frequency of fixed equipment also makes it difficult to reflect changes in traffic conditions in real time. Summary of the Invention

[0003] This application provides a method, apparatus, device, and storage medium for real-time identification of road congestion status, so as to at least solve the technical problem in the related art that it is difficult to accurately identify road congestion status in real time.

[0004] According to one aspect of the embodiments of this application, a method for real-time identification of road congestion status is provided, comprising:

[0005] Acquire trajectory data of idling vehicles equipped with front-view imaging devices, and filter target vehicles traveling on the road based on the trajectory data.

[0006] Extract the onboard front view image of the target vehicle, input the onboard front view image into a pre-trained traffic object recognition model, and output the traffic object recognition result in the image.

[0007] Based on the traffic object recognition results, initial congestion images are filtered, and the traffic light categories corresponding to the initial congestion images are automatically labeled;

[0008] Based on the traffic light category, the initial congestion image is input into the corresponding congestion recognition model, and the congestion level determination result of the current vehicle is output.

[0009] In one implementation, filtering target vehicles traveling on the road based on the trajectory data includes:

[0010] The vehicle's location is determined by matching the trajectory data with the road data.

[0011] Calculate the distance between the vehicle and the center line of the road;

[0012] If the distance is less than or equal to a preset threshold, the vehicle is identified as a target vehicle traveling on the road.

[0013] In one implementation, based on the traffic object recognition results, initial congestion images are filtered, and the traffic light categories corresponding to the initial congestion images are automatically labeled, including:

[0014] Based on the traffic object recognition results, the number and location of vehicles in the image and the type of traffic lights are determined;

[0015] The image showing a preset number or more vehicles in the area in front of the target vehicle is used as the initial congestion image.

[0016] Based on the automatic labeling of the traffic light categories, the traffic light categories corresponding to the initial congestion image include those with traffic lights and those without traffic lights.

[0017] In one implementation, based on the traffic light category, the initial congestion image is input into the corresponding congestion recognition model, and the current vehicle's congestion level determination result is output, including:

[0018] If the initial congestion image shows traffic lights, the initial congestion image is input into the pre-trained first congestion recognition model, and the congestion level determination result of the current vehicle is output.

[0019] If the initial congestion image is without traffic lights, the initial congestion image is input into a pre-trained second congestion recognition model, which outputs the congestion level determination result for the current vehicle.

[0020] In one implementation, before extracting the onboard front view image of the target vehicle and inputting the onboard front view image into a pre-trained traffic object recognition model, the method further includes:

[0021] The vehicle front view image is extracted based on historical trajectory data and taken by a vehicle driving on the road and idling.

[0022] The YOLO open-source model was used to identify the location and number of vehicles in the vehicle front view image, and positive and negative sample data were filtered based on the image recognition results.

[0023] The outlines and types of traffic objects in the positive sample data are labeled, and the YOLO model is fine-tuned based on the labeled data and negative sample data to obtain the traffic object recognition model.

[0024] In one implementation, before inputting the initial congestion image into the corresponding congestion recognition model based on the traffic light category, the method further includes:

[0025] Collect vehicle front view images with traffic lights, label the traffic light status and congestion level in the image data, and obtain the first training dataset;

[0026] Based on the first training dataset and the ResNet network architecture, train the first congestion recognition model;

[0027] Collect vehicle front view images without traffic lights, label the congestion level in the images, and obtain the second training dataset;

[0028] A second congestion identification model is trained based on the second training dataset and the ResNet network architecture.

[0029] In one implementation, after outputting the current vehicle's congestion level determination result, the method further includes:

[0030] Based on the image labels corresponding to the congestion level determination results, the corresponding vehicle trajectories are associated;

[0031] Based on the vehicle trajectory matching road area, congested road sections are identified.

[0032] According to another aspect of the embodiments of this application, an apparatus for real-time identification of road congestion status is provided, comprising:

[0033] The filtering module is used to acquire trajectory data of idling vehicles equipped with front-view imaging devices, and to filter target vehicles traveling on the road based on the trajectory data.

[0034] The traffic object recognition module is used to extract the onboard front view image of the target vehicle, input the onboard front view image into the pre-trained traffic object recognition model, and output the traffic object recognition result in the image.

[0035] The labeling module is used to filter initial congestion images based on the traffic object recognition results and automatically label the traffic light categories corresponding to the initial congestion images;

[0036] The congestion recognition module is used to input the initial congestion image into the corresponding congestion recognition model based on the traffic light category, and output the congestion level determination result of the current vehicle.

[0037] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-described method for real-time identification of road congestion status through the computer program.

[0038] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer program, and the computer program is configured to execute the above-described method for real-time identification of road congestion status when it is run.

[0039] The technical solutions provided in this application embodiment may include the following beneficial effects:

[0040] This application achieves efficient congestion level determination by combining vehicle trajectory data and onboard forward-view images. First, target vehicles on the road are screened using trajectory data to ensure the accuracy of the analysis. Next, a pre-trained traffic object recognition model analyzes the onboard forward-view images to quickly select initial congestion images and automatically label traffic light categories. Finally, the images are input into the corresponding congestion recognition model according to different traffic light categories to achieve accurate classification and determination of the congestion level. Classification based on the presence or absence of traffic lights improves the accuracy of the results. The entire process is automated and intelligent, improving the accuracy and real-time performance of congestion recognition, providing strong support for traffic management and optimization, helping to alleviate traffic pressure and improve road traffic efficiency. Attached Figure Description

[0041] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0042] Figure 1 This is a flowchart of a method for real-time identification of road congestion status according to an embodiment of this application;

[0043] Figure 2 This is a schematic diagram of a model construction method according to an embodiment of this application;

[0044] Figure 3 This is a schematic diagram of a road congestion status identification method according to an embodiment of this application;

[0045] Figure 4 This is a schematic diagram of a road vehicle according to an embodiment of this application;

[0046] Figure 5 This is a schematic diagram of a road vehicle according to an embodiment of this application;

[0047] Figure 6 This is a schematic diagram of a device for real-time identification of road congestion status according to an embodiment of this application;

[0048] Figure 7 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0049] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0050] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0051] The method for real-time identification of road congestion status according to embodiments of this application will be described in detail below with reference to the accompanying drawings. Figure 1 As shown, the method mainly includes the following steps:

[0052] S101 acquires the trajectory data of idling vehicles equipped with front-view imaging devices, and filters target vehicles traveling on the road based on the trajectory data.

[0053] In one implementation, for vehicles equipped with front-facing cameras, the vehicle's current trajectory data, including its real-time location, speed, and direction of travel, is first obtained through the vehicle's positioning system (such as GPS). Using this trajectory data, combined with a preset speed threshold (e.g., speed less than 5 km / h), vehicles in an idling state are filtered out.

[0054] Furthermore, the vehicle's location is determined by matching trajectory data with road data; the distance between the vehicle and the road centerline is calculated; and if the distance is less than or equal to a preset threshold, the vehicle is identified as the target vehicle traveling on the road.

[0055] The vehicle trajectory data is spatially matched with pre-stored road network data. By calculating the vertical distance between the vehicle trajectory point and the nearest road centerline, it is determined whether the vehicle is on the road. If the distance is less than or equal to a preset threshold (e.g., 10 meters), the vehicle is identified as the target vehicle traveling on the road.

[0056] This method can accurately filter vehicles traveling on the road, effectively eliminating false positives caused by parking in parking lots or other reasons. By setting reasonable thresholds, the accuracy of the filtering results can be ensured, providing a reliable data foundation for subsequent applications such as traffic flow analysis and congestion monitoring.

[0057] S102 extracts the onboard front view image of the target vehicle, inputs the onboard front view image into the pre-trained traffic object recognition model, and outputs the traffic object recognition results in the image.

[0058] In this embodiment of the application, a trained traffic object recognition model is used to identify targets such as vehicles, pedestrians, road lines, and traffic lights in the image.

[0059] Specifically, the currently captured front-view image is extracted from the target vehicle's front-facing camera. For example... Figure 4 and Figure 5 The image shown is a schematic diagram of the extracted vehicle front view image. These images are then input into a pre-trained traffic object recognition model. This model is typically based on a deep learning framework, such as the YOLO network, and has been trained on large-scale traffic scene datasets, enabling it to recognize various traffic objects, such as vehicles, pedestrians, and traffic lights. The final output is the location and category of each traffic object in the image.

[0060] Based on the traffic object recognition results, S103 filters the initial congestion images and automatically labels the traffic light categories corresponding to the initial congestion images.

[0061] In one implementation, based on the traffic object recognition results, the number and location of vehicles in the image and the type of traffic lights are determined; an image in which there are more than a preset number of vehicles in the area in front of the target vehicle is used as the initial congestion image.

[0062] First, based on the output of the traffic object recognition model, the number and location information of vehicles in the image are statistically analyzed. A preset vehicle number threshold (e.g., 5 vehicles) is set. If the number of vehicles in the area in front of the target vehicle reaches or exceeds this threshold, the image is marked as an initial congestion image. This process is implemented through automated algorithms, which can quickly filter out potentially congested scenarios and provide basic data for further congestion level determination.

[0063] Furthermore, based on the automatic labeling of traffic light categories, the traffic light categories corresponding to the initial congestion image include those with traffic lights and those without traffic lights.

[0064] During implementation, the initial congestion images are further analyzed, and the images are automatically labeled based on the traffic light category information output by the traffic object recognition model. Specifically, if a traffic light is detected in the image, it is labeled as "traffic light present"; conversely, if no traffic light is detected, it is labeled as "no traffic light". This automatic labeling step can efficiently distinguish different scenarios, providing an accurate basis for subsequently selecting the appropriate congestion recognition model based on the traffic light status, thereby improving the intelligence level and recognition accuracy of the entire congestion recognition system.

[0065] Based on traffic light categories, S104 inputs the initial congestion image into the corresponding congestion recognition model and outputs the congestion level determination result for the current vehicle.

[0066] In this embodiment of the application, when the initial congestion image shows traffic lights, the initial congestion image is input into the pre-trained first congestion recognition model, and the congestion level determination result of the current vehicle is output.

[0067] If the initial congestion image is without traffic lights, the initial congestion image is input into the pre-trained second congestion recognition model, and the congestion level judgment result of the current vehicle is output.

[0068] The congestion levels are categorized as no congestion, light congestion, moderate congestion, and heavy congestion.

[0069] In the embodiments of this application, different congestion recognition models are used to determine the congestion level based on whether the initial congestion image contains traffic lights. Specifically, when the initial congestion image is automatically labeled as "with traffic lights," the image is input into a pre-trained first congestion recognition model. This model is specifically optimized for scenarios with traffic lights and can comprehensively consider traffic light status factors to output the congestion level determination result for the current vehicle. Conversely, when the initial congestion image is labeled as "without traffic lights," the image is input into a pre-trained second congestion recognition model. This model focuses on analyzing features such as the absence of traffic lights to determine the congestion level of the current vehicle. In this way, this application can select the most suitable model for congestion level determination according to different traffic scenarios, thereby improving the accuracy and reliability of the determination.

[0070] Furthermore, based on the image identifiers corresponding to the congestion level determination results, the corresponding vehicle trajectories are associated, and road areas are matched based on the vehicle trajectories to determine congested road sections.

[0071] First, based on the image identifier corresponding to the congestion level determination result, vehicle trajectory data with the same timestamp as the image is extracted. Using the location information (such as latitude and longitude) in the vehicle trajectory data, combined with high-precision road network data, spatial analysis algorithms are used to match the vehicle trajectories with the road network.

[0072] Furthermore, the number and distribution of vehicles traveling on the road and experiencing congestion are counted to determine the start and end points of the congested section.

[0073] In one implementation, the number of vehicles in congestion on each road is calculated. The distribution range of congested vehicles is determined by analyzing their location information. Clustering algorithms (such as DBSCAN) can be used to group congested vehicles, with each group representing a congestion area.

[0074] For each congested area, calculate its boundary points. These boundary points can be the foremost and rearmost positions of the congested vehicles. Determine the start and end points: Map the boundary points of each congested area onto the road to determine the start and end points of the congested road segment.

[0075] Furthermore, information such as the start and end points of congested road sections, congestion levels, and affected areas are integrated into a complete description. This information is then fed back to traffic management departments or navigation systems in real time to facilitate timely traffic management measures. Different traffic management strategies can be quickly developed based on different congestion levels.

[0076] This process is achieved through automated algorithms, which can update congestion information in real time and provide traffic management departments with accurate data on congested road sections so that timely traffic management measures can be taken.

[0077] In one implementation, before extracting the onboard front view image of the target vehicle and inputting the onboard front view image into the pre-trained traffic object recognition model, the process further includes training the traffic object recognition model and the congestion recognition model.

[0078] Specifically, the vehicle-mounted front view image is extracted from a vehicle that is driving on the road and idling, based on historical trajectory data.

[0079] In one implementation, by combining vehicle trajectory data and national road network data, trajectory points with speeds below 5 km / h and distances from the route centerline are selected by matching trajectory points with the route centerline. The vehicle IDs and trajectory times corresponding to these trajectory points are then extracted. Next, images matching the vehicle IDs and times selected from the vehicle-mounted image data are chosen for subsequent analysis.

[0080] Furthermore, the YOLO open-source model is used to identify the position and number of vehicles in the vehicle's front view image, and positive and negative sample data are filtered based on the image recognition results. Based on the YOLO open-source model, the vehicle positions in each image are identified. Images with multiple vehicles arranged vertically in the middle area are selected as candidate positive samples. Images with a similar number of samples as those selected in the previous step, but which do not meet the above conditions, are selected as candidate negative samples.

[0081] Furthermore, the outlines and types of traffic objects in the positive sample data are labeled, such as various vehicles, pedestrians, signs, traffic lights, roads, obstacles, etc. The YOLO model is then fine-tuned using the labeled data and negative sample data to obtain the traffic object recognition model.

[0082] It also includes collecting vehicle-mounted front view images with traffic lights, labeling the traffic light status and congestion level in the image data to obtain the first training dataset; and training the first congestion recognition model based on the first training dataset and the ResNet network architecture.

[0083] First, we collected onboard front-view images of vehicles equipped with front-facing imaging devices while driving in areas with traffic lights. These images were then annotated in detail, including information such as traffic light status (red, green, yellow) and congestion level (no congestion, light, moderate, heavy), to construct the first training dataset.

[0084] Next, the ResNet network architecture was chosen as the base model due to its powerful feature extraction capabilities, making it suitable for processing complex image data. The ResNet model was trained using a pre-labeled training dataset, with appropriate training parameters set, such as learning rate and batch size. Cross-validation and other methods were employed to monitor the training process and prevent overfitting. In this way, the trained first congestion recognition model can accurately determine the congestion level based on the input vehicle-mounted forward view image, providing strong support for traffic congestion monitoring.

[0085] In one implementation, for complex scenarios with traffic lights, a hybrid architecture is formed by combining ResNet and Transformer modules. The Transformer module is introduced to capture long-range dependencies and global features. A first congestion recognition model is trained based on this hybrid architecture.

[0086] Specifically, the system includes a feature extraction layer that uses the first few layers of ResNet to extract local features; a Transformer module inserted after ResNet to capture global features; a fusion layer that combines the output features of ResNet and the Transformer module to form the final feature representation; and a classification layer that uses fully connected layers to perform the final congestion level classification.

[0087] Furthermore, vehicle-mounted forward view images without traffic lights were collected, and the congestion level in the images was labeled to obtain a second training dataset; based on the second training dataset and the ResNet network architecture, a second congestion recognition model was trained.

[0088] Based on the trained traffic object recognition model and congestion recognition model, road congestion status analysis is performed.

[0089] To facilitate understanding of the methods in the embodiments of this application, the following description is provided in conjunction with the appendix. Figure 2 Further description, such as Figure 2 The diagram illustrates the model construction method. First, based on vehicle trajectory data and national road network data, the trajectory and route data are matched. From the matching results, trajectory points with vehicle speeds less than 5 km / h and distances from the route centerline less than 10 m are selected, and the vehicle ID and trajectory time of these trajectory points are obtained.

[0090] Further, select images from the vehicle's front view images that meet the above conditions in terms of vehicle ID and time.

[0091] Furthermore, based on the YOLO open-source model, the positions of vehicles in each image are identified. Images with multiple vehicles arranged vertically in the middle region are selected as candidate positive samples. Images with a similar number of samples as those selected in the previous step, but which do not meet the above conditions, are selected as candidate negative samples.

[0092] The selected images are manually annotated, including the outlines and types of each traffic-related object in the images, as well as the congestion level. This includes annotating the outlines and types of each traffic-related object (vehicles, pedestrians, signs, traffic lights, roads, obstacles, etc.) in the images, and the congestion level.

[0093] Using labeled contour and type data, the open-source YOLO model was fine-tuned to train a traffic object recognition model.

[0094] Further, images with traffic light labels are selected. A congestion classification model is trained using the ResNet network architecture when traffic lights are present. Images without traffic light labels are selected. A congestion classification model is trained using the ResNet network architecture when traffic lights are absent.

[0095] like Figure 3 As shown, a road congestion model recognition method is described. First, the trajectories of vehicles equipped with onboard forward-view imaging devices and traveling at speeds less than 5 km / h are selected in real time. The trajectories are matched with route data, and if the distance to the route is less than 10m, the current image is acquired as a candidate image.

[0096] Using a traffic object recognition model, identify traffic objects in the image and select images where multiple vehicles are present in the foreground area.

[0097] Further, determine whether a traffic light object exists in the image. If a traffic light object exists, select the "Traffic Light Present" classification model. If no traffic light object exists, select the "Traffic Light Absent" classification model.

[0098] Based on the classification model, the congestion level of the current image is determined. For congested images, the congested road sections are reconstructed based on route matching data.

[0099] The flowchart describes an automated traffic congestion identification system that identifies and reconstructs congested road sections by analyzing onboard forward-view images and vehicle trajectory data.

[0100] This solution integrates vehicle trajectory data, onboard forward-view images, and advanced image recognition technology to accurately identify and analyze traffic congestion. First, idling vehicles are identified using vehicle trajectory data, and onboard forward-view images capture actual road conditions, providing a rich data source for congestion identification. Second, deep learning models such as YOLO and ResNet are used to process and recognize the images, improving the accuracy and efficiency of congestion identification. Furthermore, by distinguishing between scenarios with and without traffic lights, different classification models are used to determine congestion levels, making congestion identification more refined and scenario-based. Finally, by reconstructing congested road sections, intuitive and accurate information is provided for traffic management and optimization. Overall, this solution significantly improves the intelligence level of traffic congestion monitoring, providing strong technical support for alleviating urban traffic pressure and improving road efficiency.

[0101] According to another aspect of the embodiments of this application, an apparatus for real-time identification of road congestion status is also provided for implementing the above-described method for real-time identification of road congestion status. For example... Figure 6 As shown, the device includes:

[0102] The filtering module 601 is used to acquire trajectory data of idling vehicles equipped with front-view imaging devices, and filter target vehicles traveling on the road based on the trajectory data.

[0103] The traffic object recognition module 602 is used to extract the onboard front view image of the target vehicle, input the onboard front view image into the pre-trained traffic object recognition model, and output the traffic object recognition result in the image.

[0104] The annotation module 603 is used to filter the initial congestion images based on the traffic object recognition results and automatically annotate the traffic light categories corresponding to the initial congestion images;

[0105] The congestion recognition module 604 is used to input the initial congestion image into the corresponding congestion recognition model based on the traffic light category, and output the congestion level judgment result of the current vehicle.

[0106] It should be noted that the above embodiments of the device for real-time identification of road congestion status, when executing the method for real-time identification of road congestion status, are only illustrative examples of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the device for real-time identification of road congestion status and the method embodiments for real-time identification of road congestion status provided in the above embodiments belong to the same concept, and their implementation process is detailed in the method embodiments, which will not be repeated here.

[0107] According to another aspect of the present application, an electronic device corresponding to the method for real-time identification of road congestion status provided in the foregoing embodiments is also provided, for executing the above-described method for real-time identification of road congestion status.

[0108] Please refer to Figure 7 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 7 As shown, the electronic device includes: a processor 700, a memory 701, a bus 702, and a communication interface 703. The processor 700, the communication interface 703, and the memory 701 are connected via the bus 702. The memory 701 stores a computer program that can run on the processor 700. When the processor 700 runs the computer program, it executes the method for real-time identification of road congestion status provided in any of the foregoing embodiments of this application.

[0109] The memory 701 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 703 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0110] Bus 702 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. Memory 701 is used to store programs. After receiving execution instructions, processor 700 executes the program. The method for real-time identification of road congestion status disclosed in any of the aforementioned embodiments of this application can be applied to processor 700, or implemented by processor 700.

[0111] The processor 700 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 700 or by instructions in software form. The processor 700 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 701. Processor 700 reads the information in memory 701 and, in conjunction with its hardware, completes the steps of the above method.

[0112] The electronic device provided in this application embodiment and the method for real-time identification of road congestion status provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0113] According to another aspect of the present application, a computer-readable storage medium corresponding to the method for real-time identification of road congestion status provided in the foregoing embodiments is also provided, wherein a computer program (i.e., a program product) is stored thereon, and when the computer program is run by a processor, it executes the method for real-time identification of road congestion status provided in any of the foregoing embodiments.

[0114] It should be noted that examples of computer-readable storage media may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0115] The computer-readable storage medium provided in the above embodiments of this application and the method for real-time identification of road congestion status provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0116] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for real-time identification of road congestion status, characterized in that, include: Acquire trajectory data of idling vehicles equipped with front-view imaging devices, and filter target vehicles traveling on the road based on the trajectory data. Extract the onboard front view image of the target vehicle, input the onboard front view image into a pre-trained traffic object recognition model, and output the traffic object recognition result in the image. Based on the traffic object recognition results, initial congestion images are filtered, and the traffic light categories corresponding to the initial congestion images are automatically labeled; Based on the traffic light category, the initial congestion image is input into the corresponding congestion recognition model, and the congestion level determination result of the current vehicle is output; including: when the initial congestion image has traffic lights, the initial congestion image is input into a pre-trained first congestion recognition model, and the congestion level determination result of the current vehicle is output; when the initial congestion image has no traffic lights, the initial congestion image is input into a pre-trained second congestion recognition model, and the congestion level determination result of the current vehicle is output. Before inputting the initial congestion image into the corresponding congestion recognition model based on the traffic light category, the method further includes: collecting vehicle-mounted front view images with traffic lights, labeling the traffic light status and congestion level in the image data to obtain a first training dataset; training a first congestion recognition model based on the first training dataset and a hybrid network architecture of ResNet and Transformer; collecting vehicle-mounted front view images without traffic lights, labeling the congestion level in the images to obtain a second training dataset; and training a second congestion recognition model based on the second training dataset and the ResNet network architecture.

2. The method according to claim 1, characterized in that, Based on the trajectory data, target vehicles traveling on the road are filtered, including: The vehicle's location is determined by matching the trajectory data with the road data. Calculate the distance between the vehicle and the center line of the road; If the distance is less than or equal to a preset threshold, the vehicle is identified as a target vehicle traveling on the road.

3. The method according to claim 1, characterized in that, Based on the traffic object recognition results, initial congestion images are filtered, and the traffic light categories corresponding to the initial congestion images are automatically labeled, including: Based on the traffic object recognition results, the number and location of vehicles in the image and the type of traffic lights are determined; The image showing a preset number or more vehicles in the area in front of the target vehicle is used as the initial congestion image. Based on the automatic labeling of the traffic light categories, the traffic light categories corresponding to the initial congestion image include those with traffic lights and those without traffic lights.

4. The method according to claim 1, characterized in that, Before extracting the onboard front view image of the target vehicle and inputting the onboard front view image into the pre-trained traffic object recognition model, the process further includes: The vehicle front view image is extracted based on historical trajectory data and taken by a vehicle driving on the road and idling. The YOLO open-source model was used to identify the location and number of vehicles in the vehicle front view image, and positive and negative sample data were filtered based on the image recognition results. The outlines and types of traffic objects in the positive sample data are labeled, and the YOLO model is fine-tuned based on the labeled data and negative sample data to obtain the traffic object recognition model.

5. The method according to claim 1, characterized in that, After outputting the current vehicle's congestion level assessment result, it also includes: Based on the image labels corresponding to the congestion level determination results, the corresponding vehicle trajectories are associated; Based on the vehicle trajectory matching road area, congested road sections are identified.

6. A device for real-time identification of road congestion status, characterized in that, include: The filtering module is used to acquire trajectory data of idling vehicles equipped with front-view imaging devices, and to filter target vehicles traveling on the road based on the trajectory data. The traffic object recognition module is used to extract the onboard front view image of the target vehicle, input the onboard front view image into the pre-trained traffic object recognition model, and output the traffic object recognition result in the image. The labeling module is used to filter initial congestion images based on the traffic object recognition results and automatically label the traffic light categories corresponding to the initial congestion images; The congestion recognition module is used to input the initial congestion image into the corresponding congestion recognition model based on the traffic light category, and output the congestion level determination result of the current vehicle; including: when the initial congestion image has traffic lights, inputting the initial congestion image into a pre-trained first congestion recognition model and outputting the congestion level determination result of the current vehicle; when the initial congestion image has no traffic lights, inputting the initial congestion image into a pre-trained second congestion recognition model and outputting the congestion level determination result of the current vehicle; Before inputting the initial congestion image into the corresponding congestion recognition model based on the traffic light category, the method further includes: collecting vehicle-mounted front view images with traffic lights, labeling the traffic light status and congestion level in the image data to obtain a first training dataset; training a first congestion recognition model based on the first training dataset and a hybrid network architecture of ResNet and Transformer; collecting vehicle-mounted front view images without traffic lights, labeling the congestion level in the images to obtain a second training dataset; and training a second congestion recognition model based on the second training dataset and the ResNet network architecture.

7. An electronic device, characterized in that, The device includes a processor and a memory storing program instructions, the processor being configured to perform, when executing the program instructions, the method for real-time identification of road congestion status as described in any one of claims 1 to 5.

8. A computer-readable medium, characterized in that, It stores computer-readable instructions that are executed by a processor to implement a method for real-time identification of road congestion status as described in any one of claims 1 to 5.

Images