Traffic parameter processing method and apparatus, and device and medium

Abstract

A traffic parameter processing method and apparatus, and a device and a medium. The traffic parameter processing method comprises: acquiring real-time traffic data collected by at least one sensor for a target region at a current monitoring moment (S100); on the basis of the real-time traffic data and lane information of the target region, obtaining a pending vehicle queue of at least one lane in the target region (S200); comparing the pending vehicle queue with a reference vehicle queue, and determining at least one occluded vehicle target in the pending vehicle queue (S300); on the basis of reference position information and position information and speed information of the occluded vehicle target in the reference vehicle queue, predicting current position information of the occluded vehicle target (S400); and on the basis of the current position information of the at least one occluded vehicle target, updating the pending vehicle queue to obtain a complete vehicle queue in the lane at the current moment (S500).

Description

Traffic parameter processing method, device, equipment and medium Technical Field

[0001] The present application relates to the field of traffic monitoring technology, and in particular to a traffic parameter processing method, device, equipment and medium. Background Art

[0002] Traffic parameter processing methods mainly include radar method, ultrasonic method, GPS data-based method, video image processing-based method, etc., in order to collect traffic parameters such as the real physical position, speed, direction, etc. of vehicle targets.

[0003] Summary of the Invention

[0004] In a first aspect, the present application provides a traffic parameter processing method, comprising: obtaining real-time traffic data collected by at least one sensor for a target area at a current monitoring moment; obtaining a pending queue for at least one lane in the target area based on the real-time traffic data and lane information of the target area, wherein the pending queue includes a plurality of vehicle targets arranged in sequence, and the vehicle targets have position information and speed information; for each lane in the at least one lane, comparing the pending queue with a reference queue to determine at least one obscured vehicle target in the pending queue; the reference queue is a complete queue for the same lane at the previous monitoring moment; for each obscured vehicle target in the at least one obscured vehicle target, predicting the current position information of the obscured vehicle target based on reference position information and the position information and speed information of the obscured vehicle target in the reference queue, wherein the reference position information is the position information of the previous vehicle target of the obscured vehicle target in the pending queue in the reference queue; updating the pending queue based on the current position information of the at least one obscured vehicle target to obtain the complete queue for the lane at the current moment.

[0005] In a possible embodiment of the present application, the current position information of the obscured vehicle target is predicted based on the reference position information and the position information and speed information of the obscured vehicle target in the reference queue, including: obtaining the first predicted position information of the obscured vehicle target based on the reference position information and the preset vehicle distance information; obtaining the second predicted position information of the obscured vehicle target based on the position information and speed information of the obscured vehicle target in the reference queue; and taking the maximum value of the first predicted position information and the second predicted position information as the current position information of the obscured vehicle target.

[0006] In a possible embodiment of the present application, the pending queue is compared with the reference queue to determine at least one obscured vehicle target in the pending queue, including: comparing the pending queue with the reference queue to determine at least one suspected obscured target; for each suspected obscured target in the at least one suspected obscured target, determining the distance and intersection-and-union ratio between the suspected obscured target and a detectable vehicle target from the reference queue, wherein the detectable vehicle target is the previous vehicle target of the suspected obscured target in the reference queue; if the distance is less than a first preset threshold and the intersection-and-union ratio is greater than a second preset threshold, the suspected obscured target is determined as an obscured vehicle target.

[0007] In a possible embodiment of the present application, the at least one sensor includes: a traffic radar for collecting real-time traffic point cloud data for the target area; and a visual sensor for collecting real-time traffic image data for the target area. The obtaining of a pending queue for at least one lane in the target area based on the real-time traffic data and lane information of the target area comprises: performing target detection on the real-time traffic point cloud data to obtain one or more first vehicle targets in a radar coordinate system, wherein the radar coordinate system generates one or more corresponding virtual lanes based on the lane information of the target area; performing target detection on the real-time traffic image data to obtain one or more second vehicle targets in an image coordinate system; determining one or more target matching groups by converting the one or more second vehicle targets into the radar coordinate system, wherein each target matching group includes a matching first vehicle target and a matching second vehicle target; for each target matching group, determining a valid vehicle target from the first vehicle target and the second vehicle target based on a confidence comparison result between the first vehicle target and the second vehicle target in the target matching group; and obtaining a pending queue for at least one lane in the target area by generating the valid vehicle target in the corresponding virtual lane based on at least one valid vehicle target.

[0008] In a possible embodiment of the present application, determining a valid vehicle target from the first vehicle target and the second vehicle target based on a confidence comparison result between the first vehicle target and the second vehicle target in the target matching group includes: determining a confidence area of ​​the target matching group in the radar coordinate system, wherein, in the radar coordinate system, along the direction away from the traffic radar, it is sequentially divided into a visual high confidence area, an equal confidence area, and a radar high confidence area; if the confidence area is a visual high confidence area, the second vehicle target is taken as a valid vehicle target; if the confidence area is a radar high confidence area, the first vehicle target is taken as a valid vehicle target; if the confidence area is an equal confidence area, the first vehicle target and the second vehicle target are fused to obtain a valid vehicle target.

[0009] In a possible embodiment of the present application, the method further includes: identifying a background area brightness value of real-time traffic image data; obtaining a first distance between a first distal boundary of the visual high confidence area and the traffic radar based on the background area brightness value and a preset dynamic coefficient; obtaining a second distance between a second distal boundary of the equal confidence area and the traffic radar based on the farthest distance that the visual sensor can collect and a preset width value of the equal confidence area; based on the first distance and the second distance, in the radar coordinate system, sequentially separating the visual high confidence area, the equal confidence area, and the radar high confidence area along a direction away from the traffic radar.

[0010] In a possible embodiment of the present application, after determining the valid vehicle target from the first vehicle target and the second vehicle target in the target matching group, the method includes: generating a valid vehicle target in a radar coordinate system; based on the position information of the valid vehicle target at the current monitoring moment and multiple previous consecutive monitoring moments, judging whether the valid vehicle target enters from one end of the virtual section identification area of ​​the virtual lane and exits from the other end of the virtual section identification area; if the vehicle target enters from one end of the virtual section identification area of ​​the virtual lane and exits from the other end, determining that the valid vehicle target belongs to the virtual lane, and updating the traffic flow of the lane corresponding to the virtual lane.

[0011] In a second aspect, the present application further provides a traffic parameter processing device, comprising: a data acquisition module, configured to acquire real-time traffic data collected by at least one sensor for a target area at a current monitoring moment; a queue generation module, configured to obtain a pending queue for at least one lane in the target area based on the real-time traffic data and lane information of the target area, wherein the pending queue includes a plurality of vehicle targets arranged in sequence, and the vehicle targets have position information and speed information; a target determination module, configured to compare the pending queue with a reference queue for each lane in the at least one lane, and determine at least one obscured vehicle target in the pending queue; the reference queue is the complete queue of the lane at the previous monitoring moment; A position prediction module is used to predict the current position information of each of the at least one obscured vehicle targets based on reference position information and the position information and speed information of the obscured vehicle target in the reference queue, wherein the reference position information is the position information of the previous vehicle target of the obscured vehicle target in the pending queue in the reference queue; a queue improvement module is used to update the pending queue based on the current position information of the at least one obscured vehicle target to obtain the complete queue of the lane at the current moment.

[0012] In a third aspect, the present application further provides a traffic parameter processing device, comprising: a processor, a memory, and a computer program stored in the memory, wherein the computer program is executed by the processor to implement the traffic parameter processing method of the first aspect.

[0013] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the traffic parameter processing method of the first aspect.

[0014] A traffic parameter processing method proposed in an embodiment of the present application includes: obtaining real-time traffic data collected by at least one sensor for a target area at a current monitoring moment; obtaining a pending queue of at least one lane in the target area based on the real-time traffic data and lane information of the target area, wherein the pending queue includes position information and speed information of each vehicle target; for each lane in the at least one lane, comparing the pending queue with a reference queue at a previous monitoring moment to determine at least one obscured vehicle target in the pending queue; for each obscured vehicle target in the at least one obscured vehicle target, predicting the current position information of the obscured vehicle target based on reference position information and the position information and speed information of the obscured vehicle target in the reference queue; updating the pending queue based on the current position information of the at least one obscured vehicle target to obtain a complete queue of the lane at the current moment.

[0015] It can be seen that after the present application obtains the pending queues of each lane based on the real-time traffic data collected by at least one sensor, for the suspected obscured targets, since the rear vehicle cannot overtake the front vehicle in the same lane, the current position information of the previous vehicle target and the position information and speed information of the obscured vehicle target in the reference queue, that is, the position information and speed information before the obscured rear vehicle disappears, can be used to predict the current position information of the suspected obscured target, so that it no longer relies entirely on the perceived real-time traffic data to judge the current data of the vehicle, reducing the impact of the difficulty of occlusion on the traffic parameter processing algorithm, thereby improving the traffic parameter measurement accuracy and traffic induction effect in dense scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIG1 is a schematic diagram of the structure of the traffic parameter processing device of the present application.

[0017] FIG2 is a flow chart of the first embodiment of the traffic parameter processing method of the present application.

[0018] FIG3 is a schematic diagram showing the arrangement of the radar-visual vehicle detector of the present application.

[0019] FIG4 is a flow chart of the traffic parameter information collection process of this application.

[0020] FIG5 is a schematic diagram of the partitioning of the confidence region of this application.

[0021] FIG6 is a schematic diagram of parameters for traffic information collection in this application.

[0022] FIG7 is a schematic diagram of movement between the vehicle target and the virtual section identification area of ​​the present application.

[0023] FIG8 is a schematic diagram of the queue of the present application.

[0024] FIG9 is a schematic diagram showing the principle of queue position prediction of the present application.

[0025] FIG10 is a module diagram of the first embodiment of the traffic parameter processing device of the present application.

[0026] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0027] It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0028] With the rapid growth of my country's automobile population, the pressure on urban road traffic is becoming increasingly severe. Therefore, it is necessary to appropriately guide and dispatch road traffic to alleviate traffic pressure and improve travel efficiency. Traffic parameters (such as traffic volume, speed, and queue length) can reflect current traffic information in real time and are crucial data sources for traffic guidance. Traffic guidance, however, requires real-time collection of traffic parameters. Currently, traffic parameter processing methods primarily include radar (microwave), ultrasonic, GPS-based, and video image processing-based methods. However, regardless of the traffic collection method employed, in actual use scenarios, when multiple vehicles are queued within a lane in the monitored area, the vehicle ahead (i.e., closest to the sensor) often obscures the vehicle behind it, partially or completely concealing the vehicle behind it. This makes it difficult for all types of sensors to accurately detect the obscured vehicle, thus affecting the data measurement of the traffic parameter monitoring system.

[0029] To this end, the present application provides a solution that no longer relies entirely on perception information to obtain the vehicle queue. Instead, it uses the current position information of the previous vehicle target and the position information and speed information of the obscured vehicle target in the reference queue, that is, the position information and speed information of the obscured rear vehicle before it disappears, to predict the current position information of the suspected obscured vehicle target, thereby completing the lane queue, and further improving the traffic parameter measurement accuracy and traffic induction effect in dense scenarios.

[0030] The inventive concept of the present application is further described below with reference to some specific embodiments.

[0031] Refer to Figure 1, which is a structural diagram of a traffic parameter processing device in a hardware operating environment involved in an embodiment of the present application.

[0032] As shown in Figure 1, the traffic parameter processing device may include: a processor 1001, such as a CPU, a user interface 1003, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to achieve connection and communication between these components. Optionally, the user interface 1003 may also be a display screen (Display) and / or an input unit such as a keyboard (Keyboard). The memory 1005 may be a high-speed RAM (Random Access Memory) memory or a non-volatile memory (non-volatile memory), such as a disk storage. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0033] It is understandable that the traffic parameter processing device may also include a network interface 1004, which may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface). Optionally, the image processing device may also include an RF (Radio Frequency) circuit, a sensor, an audio circuit, a WiFi module, and the like. The traffic parameter processing device may also include a display screen, which is used to display a UI (User Interface). The UI may include graphics, text, icons, videos, and any combination thereof. When the display screen is a touch screen display, the display screen also has the ability to collect touch signals on the surface or above the surface of the display screen. The touch signal can be input into the processor 301 as a control signal for processing.

[0034] Those skilled in the art will understand that the traffic parameter processing device structure shown in FIG1 does not constitute a limitation on the traffic parameter processing device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0035] Based on the above structure but not limited to the above structure, a first embodiment of the traffic parameter processing method of the present application is proposed. Please refer to Figure 2, which is a flow chart of the first embodiment of the traffic parameter processing method of the present application.

[0036] It should be noted that although a logical order is shown in the flowcharts, in some cases, the steps shown or described may be performed in an order different from that shown or described here.

[0037] In this embodiment, the traffic parameter processing method includes steps S100 to S500.

[0038] Step S100: Acquire real-time traffic data collected by at least one sensor for a target area at a current monitoring moment.

[0039] In this embodiment, the executor of the traffic parameter processing method is a traffic parameter processing device, which can be constructed as a computer or server connected to the sensor for communication, or the traffic parameter processing device can also be constructed as a virtual cluster arranged in the cloud.

[0040] The target area can be an intersection or other area requiring traffic guidance or monitoring. The target area includes a main road, which can have two lanes, four lanes, or more. Sensors are fixedly mounted above the main road to measure and collect traffic parameters on the main road in the target area. Therefore, the lanes in this embodiment are lanes on the main road.

[0041] As will be appreciated, sensors can include millimeter-wave radar, microwave radar, ultrasonic radar, and visual sensors. Microwave radar and ultrasonic radar, however, cannot accurately, real-timely, or conveniently detect road traffic conditions and cannot support functions such as vehicle type classification. While visual sensors can obtain more information, they struggle to determine a target's true position and velocity. Furthermore, due to limitations in resolution and lighting conditions, they struggle to control costs and achieve all-weather recognition. Therefore, in one specific embodiment, the sensor is a radar-based vehicle detector. This refers to a device that utilizes video image processing and millimeter-wave radar sensing technologies, fusing the results of these two technologies to address the shortcomings of individual sensors and detect certain traffic flow parameters or traffic events. The device possesses multiple functions and the ability to combine these functions, most notably the ability to acquire and store multi-dimensional information about the presence of targets in the current scene. As can be seen, the radar-based vehicle detector includes a traffic radar and a visual sensor, with the traffic radar optionally being a millimeter-wave radar. Referring to Figure 3, the traffic radar and visual sensor are positioned in the same direction of the main lane, thereby monitoring real-time traffic data in the same target area. Of course, in some embodiments, the radar vehicle detector can simultaneously support the collection of forward and reverse traffic real-time data, thereby obtaining more real-time traffic data for the target area.

[0042] Traffic radar is used to collect real-time traffic point cloud data for the target area at the current monitoring moment. The visual sensor is used to collect real-time traffic image data for the target area at the current monitoring moment. This embodiment fuses radar and visual tracking to obtain information such as the distance and speed of vehicle targets compared to a purely video-based acquisition system. Furthermore, compared to a purely radar-based acquisition system, it can obtain more accurate vehicle classification information and the location of stationary vehicles. This allows for adaptability to more complex scenarios and acquisition of more accurate and comprehensive information around the clock.

[0043] The sensors can collect real-time traffic data in real time, or can also collect data according to a preset collection period, which is not limited in this embodiment. It is understood that after each sensor collects real-time traffic data, it can be transmitted to the traffic parameter processing device for processing.

[0044] Step S200: Based on the real-time traffic data and the lane information of the target area, obtain the pending queue of at least one lane in the target area.

[0045] Refer to Figure 4. In the area of ​​traffic parameter collection, real-time traffic data collected by various sensors can be used to perform target detection to identify individual vehicle targets. The data of the identified vehicle targets is then continuously recorded and fused into a common coordinate system to generate fused trajectories. Data analysis is then performed based on the fused trajectories and lane information in the target area to extract traffic parameters such as lane flow, vehicle speed, and lane queue length. Lane queue length is a key parameter in traffic guidance.

[0046] In a specific implementation, step S200 specifically includes steps S210 to S250.

[0047] Step S210: Perform target detection on the real-time traffic point cloud data included in the real-time traffic data to obtain one or more first vehicle targets in the radar coordinate system.

[0048] Among them, the radar coordinate system generates a corresponding virtual lane based on the lane information.

[0049] Step S220: Perform target detection on the real-time traffic image data included in the real-time traffic data to obtain one or more second vehicle targets in the image coordinate system.

[0050] In this embodiment, a target detection algorithm can be used to perform target detection on real-time traffic point cloud data collected by traffic radar, thereby identifying one or more first vehicle targets. The data of the identified vehicle targets is then continuously recorded. The millimeter-wave radar point cloud data can be converted into feature vectors and then input into a neural network model pre-trained using calibration samples for target detection. This process can generate one or more first vehicle targets.

[0051] It is worth mentioning that after detecting the first vehicle target, the first vehicle target can also be recorded, thereby forming a radar recording queue. Specifically, after detecting the first vehicle target, the target frame, ID number, license plate information, position information, speed, distance and other associated information of the first vehicle target can also be obtained. Then, the first vehicle target at the current monitoring moment is matched one by one with each vehicle target in the radar recording queue, and the target frame overlap and license plate information matching degree of the first vehicle target at the current monitoring moment and the vehicle target to be matched in the radar recording queue are calculated. If the calculated target frame overlap and license plate information matching degree meet the preset conditions, such as being greater than the corresponding threshold value, it is considered that the target association matching is successful, and it is considered that the currently detected first vehicle target and the recorded vehicle target in the radar recording queue are the same target. Then, the associated information related to the first vehicle target in the radar recording queue can be updated based on the associated information of the first vehicle target obtained at the current monitoring moment. If the target association matching fails, it is considered that the first vehicle target is a newly detected vehicle, thereby a new first vehicle target can be created in the radar recording queue.

[0052] Real-time traffic image data collected by visual sensors can also be detected using target detection algorithms to identify the secondary vehicle targets. For example, real-time traffic image data can be converted into RGB images and then input into a pre-trained YOLOv3 neural network model for target detection, thereby identifying the secondary vehicle targets.

[0053] Similarly, after detecting the second vehicle target, the second vehicle target can also be recorded, thereby forming a visual record queue. Specifically, after detecting the second vehicle target, the associated information such as the target frame, ID number, license plate information and position information of the second vehicle target can also be obtained. Then the second vehicle target at the current monitoring moment is matched one by one with each vehicle target in the visual record queue, and the target frame overlap and license plate information matching degree of the second vehicle target at the current monitoring moment and the vehicle target to be matched in the visual record queue are calculated. If the calculated target frame overlap and license plate information matching degree meet the preset conditions, such as being greater than the corresponding threshold value, then it is considered that the target association matching is successful, and it is considered that the currently detected second vehicle target is the same target as the recorded vehicle target in the visual record queue. Then the associated information related to the second vehicle target in the visual record queue can be updated according to the associated information of the second vehicle target obtained at the current monitoring moment. If the target association matching fails, then it is considered that the second vehicle target is a newly detected vehicle, thereby a new second vehicle target can be created in the visual record queue.

[0054] Step S230: convert the one or more second vehicle targets into the radar coordinate system, and determine one or more target matching groups.

[0055] Each target matching group includes a first vehicle target and a second vehicle target that match each other.

[0056] Step S240 : For each target matching group, based on a confidence comparison result between the first vehicle target and the second vehicle target in the target matching group, determine a valid vehicle target from the first vehicle target and the second vehicle target.

[0057] It is understandable that in order to facilitate the subsequent extraction of traffic parameters, it is also necessary to unify the first vehicle target and the second vehicle target detected above into the same coordinate system and represent them through the generated vehicle target. When generating the vehicle target in the same coordinate system, the data relied on is the valid vehicle target. In this embodiment, the second vehicle target captured by the visual sensor is projected into the radar coordinate system, so that the perception and tracking results of the video and millimeter-wave radar are fused from the perspective of BEV (Bird's-Eye View), optimizing the problems of missed detection and false detection existing in a single sensor, and outputting richer information and more accurate vehicle target trajectories.

[0058] As you can understand, the traffic parameter processing equipment maintains two record queues: one for radar and one for vision. It can use intra-frame correlation to match the first and second vehicle targets, generating a target matching group representing the same vehicle. However, due to the differences in visual and radar perception characteristics, unstable performance on either the visual or radar side can easily lead to trajectory jitter, false detections, or missed detections, resulting in a target matching group that requires further processing.

[0059] It is understandable that due to the characteristics of visual perception, when the distance of the vehicle target in the real world is too far, resulting in the vehicle target in the visual image being too small, or when the imaging conditions are poor (such as rainy and foggy days, at night), the visual perception results will be poor and the vehicle target will be difficult to record continuously. At this time, it is necessary to lower the visual confidence. Due to the characteristics of millimeter-wave radar, when the vehicle target is along the radar tangent or when the target speed is low, the radar detection will be unstable, which will affect the accuracy of radar tracking. Therefore, in this case, it is necessary to lower the radar confidence. In this way, for any target matching group, based on the confidence comparison result between the first vehicle target and the second vehicle target, the data with higher confidence can be selected to generate the final valid vehicle target in the radar coordinate system.

[0060] It can be understood that, as an option of this specific embodiment, when determining the confidence comparison result, the confidence of the first vehicle target can be calculated first, and the confidence of the second vehicle target can be calculated, so as to compare the sizes of the two and obtain the confidence comparison result.

[0061] Of course, as another option of this specific embodiment, when executing step 240, the traffic parameter processing method can determine the valid vehicle target from the first vehicle target and the second vehicle target in the target matching group based on the confidence comparison result between the first vehicle target and the second vehicle target, thereby determining the confidence area of ​​the target matching group in the radar coordinate system; if the confidence area is a visual high confidence area, the second vehicle target is taken as the valid vehicle target; if the confidence area is a radar high confidence area, the first vehicle target is taken as the valid vehicle target; if the confidence area is an equal confidence area, the first vehicle target and the second vehicle target are fused to obtain the valid vehicle target.

[0062] Specifically, since both the traffic radar and the visual sensor are fixedly mounted in the target area, they can be divided into high-confidence visual areas, equal-confidence areas, and high-confidence radar areas based on the distance from the captured scene to the sensor. Refer to Figure 5. In the radar coordinate system, moving away from the traffic radar, the areas are divided into high-confidence visual areas A1, equal-confidence areas A2, and high-confidence radar areas A3. Different vehicle target generation strategies are required for different confidence areas:

[0063] (1) Visual high confidence region: Within this region, a valid vehicle target is generated based on the visual sensor, that is, based on the second vehicle target. Of course, in some examples, the first vehicle target may also be considered. In this case, when generating a valid vehicle target, the weight of the first vehicle target is less than the weight of the second vehicle target.

[0064] (2) Equal confidence region: In this region, the confidence levels for radar and vision are equal. Therefore, within this region, the first vehicle target and the second vehicle target are matched as much as possible. That is, when generating a valid vehicle target, the first vehicle target and the second vehicle target have the same weight.

[0065] (3) Radar High Confidence Zone: Within this zone, a valid vehicle target is generated based on the radar sensor, that is, based on the first vehicle target. Of course, in some examples, a second vehicle target that matches the radar sensor may also be considered. In this case, when generating a valid vehicle target, the weight of the second vehicle target is less than that of the first vehicle target.

[0066] The visual high confidence area, the equal confidence area, and the radar high confidence area can be determined as follows:

[0067] The traffic parameter processing equipment identifies the brightness value of the background area of ​​the real-time traffic image data; based on the brightness value of the background area and a preset dynamic coefficient, obtains a first distance between a first far-side boundary of the visual high-confidence area and the traffic radar; based on the farthest distance that the visual sensor can collect and a preset width value of the equal-confidence area, obtains a second distance between a second far-side boundary of the equal-confidence area and the traffic radar; based on the first distance and the second distance, in the radar coordinate system, along the direction away from the traffic radar, the high-confidence visual area, the equal-confidence area and the high-confidence radar area are separated in sequence.

[0068] Specifically, the traffic parameter processing device identifies the background brightness V of the real-time traffic image data, that is, the image frame collected at the current monitoring moment. The value range of V is 0-255. Therefore, the boundary line of the visual high confidence area is dynamically adjusted according to the background brightness V. Specifically, in the radar coordinate system, a 0-meter line is constructed based on the location of the traffic radar. It is not difficult to see that the 0-meter line is perpendicular to the lane. Taking the 0-meter line as the starting point and the first distance as the length, the visual high confidence area can be divided in the radar coordinate system. The first distance can be obtained according to the following formula 1: THR1 = V*α.

[0069] Here, THR1 is the first distance, which is the optimal perception distance of the visual sensor and is proportional to the background brightness V. α is the preset dynamic coefficient, which can be set based on experience. It is worth noting that the preset dynamic coefficient α can be adjusted based on the maximum perception distance of the traffic radar. In this case, the preset dynamic coefficient α is proportional to the maximum perception distance. In the example shown in Figure 5, THR1 = 120m.

[0070] Then, starting from the far boundary line of the high-confidence visual area and using the second distance as the length, the radar coordinate system can be divided into equal confidence areas. The second distance can be obtained according to the following formula: THR2 = THR1 + DIS1.

[0071] Where THR2 is the second distance, and DIS1 is the preset width of the confidence zone. Of course, in one example, with good vision, THR2 may exceed the maximum detection distance of the traffic radar. To avoid this, the second distance can be calculated using the following formula: THR2 = min(THR1 + DIS1, range).

[0072] Where range is the maximum detection range of the traffic radar. This means that the resulting confidence equalization area will not exceed the maximum detection range of the traffic radar. In the example shown in Figure 5, THR2 = 180m.

[0073] Then, starting from the far boundary of the equal confidence zone and ending at the line of the traffic radar's maximum sensing distance (range), THR3 = range, the radar high confidence zone is demarcated in the radar coordinate system. In the example shown in Figure 5, THR3 = 300m.

[0074] As can be seen, this implementation configures high-confidence visual areas, equal-confidence areas, and high-confidence radar areas based on the perception performance and capabilities of the two sensors at different distances, enabling rapid determination of confidence comparison results. Furthermore, this implementation uses the brightness information of the collected real-time traffic image data to determine the visual detection performance of the current scene, dynamically adjusting the positions of the three confidence areas to achieve more accurate target information acquisition.

[0075] Step S250: Generate the valid vehicle target in the corresponding virtual lane based on at least one valid vehicle target, and obtain a pending queue of at least one lane in the target area.

[0076] After determining the valid vehicle target, the valid vehicle target can be generated in the corresponding virtual lane in the radar coordinate system, that is, the trajectory of the same vehicle target is updated in the radar coordinate system. After the trajectory is updated, the first vehicle target or the second vehicle target of a single sensor can use Kalman filtering to maintain the original record queue. However, the trajectory of the generated valid vehicle target is updated over time. Due to the different perception modes of the visual and radar sensors, such as the distance and speed measurement of the visual sensor are inaccurate, and the low-speed target and lateral positioning of the radar sensor are inaccurate, it is also necessary to dynamically adjust the sensor measurement noise during the update. Specifically, the collected measurement information includes x (lateral positioning), y (longitudinal positioning), vx (lateral speed), and vy (longitudinal speed). For visual noise, the measurement information can be set as follows: [noise_v x ,noise_v y ,10*noise_v vx ,10*noise_v vy ].

[0077] Where noise_v is the empirical value of visual measurement noise.

[0078] For radar noise, the measurement information can be set as follows: [noise_r x ,2*noise_r y ,noise_r vx ,2*noise_r vy ].

[0079] Where noise_r is the empirical value of radar measurement noise. Therefore, using the empirical values ​​of radar and visual measurement noise, the fused trajectory can be updated twice (once with radar information and once with visual information) using the Kalman filter algorithm. This way, when updating the existing fused trajectory, different measurement noise levels are empirically set based on the radar's high radial velocity and radial distance measurement accuracy and poor lateral measurement accuracy, as well as the fact that visual measurement is incapable of velocity measurement. This utilizes more accurate measurement information, thereby reducing the impact of information with large measurement errors on the trajectory and improving overall trajectory accuracy.

[0080] It is worth mentioning that after the fusion trajectory is updated, information collection and status judgment can be performed based on the real-time trajectory of each vehicle target in the radar coordinate system.

[0081] Specifically, please refer to Figure 6. The sensor can collect the following information:

[0082] Vehicle target status, including but not limited to traffic flow, speed information, headway, and headway spacing. Traffic flow is the number of vehicles passing through the virtual section marking area (arrival and departure signals) within a period of time. Speed ​​information: The speed of each vehicle target in the lane when passing through the virtual section marking area. Headway: The time interval between two vehicles when counting (arrival). Headway spacing: The headway spacing between two vehicles when counting (arrival).

[0083] Lane Status: This includes but is not limited to congestion level, queue length, time occupancy, and space occupancy. Congestion level includes three states: smooth, slow, and congested. Queue Length: The length and number of vehicles in the queue. Vehicle Count: The number of vehicles in each lane area. Time Occupancy: This reflects the percentage of vehicles occupied per unit time in the virtual section area. Space Occupancy: This reflects the percentage of vehicles in the lane per unit time.

[0084] The time occupancy ratio is the overlap ratio between the target frame and the virtual section mark area, and can be calculated as follows: time_ratio=occupy_frame / total_frame.

[0085] Where time_ratio is the time occupancy ratio, occupy_frame is the number of overlapping moments, and total_frame is the total statistical period. Spatial occupancy ratio is the ratio of the vehicle target in the current lane to the virtual lane.

[0086] It is worth mentioning that in the radar coordinate system of this embodiment, multiple virtual cross-section identification areas can be generated in the virtual lane, so that the above parameters can be recorded and obtained based on them.

[0087] Taking the traffic flow statistics function as an example, the traffic parameter processing equipment can generate a valid vehicle target in the radar coordinate system based on the valid vehicle target; based on the position information of the valid vehicle target at the current monitoring moment and the previous multiple consecutive monitoring moments, it is determined whether the valid vehicle target enters from one end of the virtual section identification area of ​​the virtual lane and exits from the other end; if so, it is determined that the valid vehicle target belongs to the virtual lane, and the traffic flow of the lane corresponding to the virtual lane is updated.

[0088] For details, please refer to FIG7 . Specifically, in this embodiment, a corresponding virtual lane is configured in the radar coordinate system. It will be understood that the virtual lane corresponds to the actual lane in the target area. Thus, based on the trajectory and the virtual lane, the lane number to which the lane target belongs can be calculated, which can be idx. If the current lane is the forward lane (e.g., from far to near, with a Bird's-Eye View (BEV) perspective from top to bottom), and a valid vehicle target enters the virtual section marking area from below the virtual section marking area, a signal indicating that the vehicle target has arrived at the virtual section marking area is generated. If the current lane is the forward lane (e.g., from far to near, with a BEV perspective from top to bottom), and a valid vehicle target exits the virtual section marking area and reaches below the virtual section marking area, a signal indicating that the target has left the virtual section marking area is generated. When the same vehicle target gives a departure arrival signal, it is considered that the vehicle target has completed the traffic counting state in the virtual section identification area, and the traffic flow of the virtual lane where the virtual section identification area is located is increased by one, that is, the traffic flow of the lane corresponding to the virtual lane is updated.

[0089] In other words, after determining the valid vehicle target from the first vehicle target and the second vehicle target, the method also includes: generating the valid vehicle target in the radar coordinate system; judging whether the valid vehicle target enters from one end of the virtual section identification area of ​​the virtual lane and exits from the other end of the virtual section identification area based on the position information of the valid vehicle target at the current monitoring moment and the previous multiple consecutive monitoring moments; if the vehicle target enters from one end of the virtual section identification area of ​​the virtual lane and exits from the other end, determining that the valid vehicle target belongs to the virtual lane, and updating the traffic flow of the lane corresponding to the virtual lane.

[0090] It is worth mentioning that after the same vehicle target is counted in the virtual section identification area, a tag signal will be recorded, thereby filtering out the situation where the same vehicle target is counted multiple times in the virtual section identification area.

[0091] Compared with the purely visual virtual section identification area, the BEV virtual section identification area in this embodiment is closer to the state of a real vehicle entering and exiting the buried solid coil.

[0092] In addition, in this embodiment, traffic statistics can also be calculated by category, for example, including classification statistics functions for motor vehicles, two-wheeled vehicles, three-wheeled vehicles, etc. Alternatively, it can also be classified according to vehicle type: supporting classification functions for large vehicles, medium vehicles, and small vehicles.

[0093] Each pending queue includes a plurality of vehicle targets arranged in sequence, and each of the plurality of vehicle targets has position information and speed information.

[0094] Refer to Figure 8 . When the current lane is congested and vehicles are stationary, the queue length l can be accumulated step by step. Specifically, the direction of the lane in which the current vehicle is located is determined. If it is in the forward direction, the distance between the vehicle and the previous vehicle in the same lane and the speed of the vehicle are determined. When the distance information and speed information are each less than a set threshold, the current vehicle is determined to have entered a queue. For lanes in a queue state, a pending queue is generated. This pending queue includes multiple vehicles, the number of vehicles in the pending queue, the distance from the last vehicle in the pending queue to the stop line, and the position and speed information of each vehicle. This allows for real-time update of the queue length and number of vehicles in the queue for traffic guidance.

[0095] Step S300: Compare the pending queue with the reference queue to determine at least one obscured vehicle target in the pending queue.

[0096] The reference queue is the complete queue of the same lane at the previous monitoring moment.

[0097] It's understandable that the pending queue at the current monitoring moment only includes vehicle targets that can be detected by the sensor. If a vehicle disappears from the sensor's detection range, it will not be included in the pending queue. It's worth noting that a vehicle disappearing from the sensor's detection range could mean it's exited the target area or was obscured.

[0098] Therefore, the complete queue and the pending queue of the same lane at the previous monitoring moment can be compared, and the disappeared vehicle target can be determined as the obscured vehicle target of the pending queue.

[0099] It is understandable that this determination method easily identifies a vehicle target that has left the target area as an obscured vehicle target. Therefore, in a specific embodiment, step S300 specifically includes steps S310 to S330.

[0100] Step S310: Compare the pending queue with the reference queue to determine at least one suspected blocked target.

[0101] Step S320: for each suspected obscured target in the at least one suspected obscured target, determine the distance and intersection-over-union ratio between the suspected obscured target and the detectable vehicle target from the reference queue.

[0102] Step S330: If the distance is less than the first preset threshold and the intersection-over-union ratio is greater than the second preset threshold, the suspected obscured target is determined to be an obscured vehicle target.

[0103] Specifically, the data from the previous monitoring moment, known as the reference queue, is extracted and compared with the pending queue. Disappearing vehicles are identified as suspected occluded targets. The distance and intersection-over-union ratio between the suspected occluded vehicle targets and detectable vehicle targets are then extracted from the reference queue.

[0104] The distance between the suspected occluded vehicle target and the detectable vehicle target can be calculated based on the position information of each vehicle target in the reference queue. The intersection over union (IoU) can be the overlap ratio between the target box of the detectable vehicle target in the visual coordinate system from the BEV perspective at the last monitoring moment and the target box of the suspected occluded vehicle target in the visual coordinate system from the BEV perspective at the last monitoring moment.

[0105] It can be understood that for a queue in a congested state, the obscured vehicle target does not disappear suddenly. From the perspective of the visual sensor, it is gradually obscured by the vehicle target in front over time, that is, the obscured vehicle target changes from being partially obscured to being completely obscured. Therefore, if the distance between the obscured vehicle target and the vehicle target in front (detectable vehicle target) is less than the first preset threshold, and the intersection-and-union ratio is greater than the second preset threshold, the suspected obscured target is determined to be an obscured vehicle target. Among them, the first preset threshold and the second preset threshold can be obtained based on experience. If this condition is not met, the suspected obscured vehicle target is considered to be an invalid target, that is, it is not an obscured vehicle target.

[0106] It is worth mentioning that, in order to ensure the accuracy of the judgment, the detectable vehicle target may be the vehicle target preceding the suspected obscured target.

[0107] In real-world congestion scenarios, multiple vehicles are often obscured in a lane's queue. Therefore, if the previous vehicle is also a suspected obscured vehicle, the system continues traversing the reference queue until it finds a vehicle in the pending queue (not the obscured one). This detected vehicle is then used as the detectable vehicle.

[0108] Step S400: For each of the at least one obscured vehicle target, predict the current position information of the obscured vehicle target based on the reference position information and the position information and speed information of the obscured vehicle target in the reference queue.

[0109] The reference position information is the position information of the previous vehicle target of the blocked vehicle target in the pending queue in the reference queue.

[0110] Specifically, after the obscured vehicle target is determined, the distance between the obscured vehicle target and the preceding vehicle target will not be very large due to the lane being congested. After all, if the distance is too large, the obscured vehicle target will not be obscured. Furthermore, in a congested state, vehicles sometimes move at a snail's pace, and this snail's pace is maintained. Therefore, in a queue generated in a congested state, the current position information of the obscured vehicle target can be predicted based on the position information of the preceding vehicle target in the reference queue, the position information and speed information of the obscured vehicle target in the reference queue, and the position information and speed information of the obscured vehicle target in the reference queue, i.e., the position information and speed information of the obscured vehicle target before it disappears.

[0111] In a specific implementation, step S400 specifically includes steps S410 to S430.

[0112] Step S410: Obtain first predicted position information of the obscured vehicle target based on the reference position information and the preset vehicle distance information.

[0113] Step S420: Obtain second predicted position information of the obscured vehicle target based on the position information and speed information of the obscured vehicle target in the reference queuing queue.

[0114] Step S430: Taking the maximum value of the first predicted position information and the second predicted position information as the current position information of the obscured vehicle target.

[0115] Specifically, in this embodiment, the current position information of the blocked vehicle target in the pending queue is determined according to the following formula 4: Y1=max(Y0+length, Y′1-vt).

[0116] Among them, Y1 is the current position information of the obscured vehicle target in the pending queue, Y0 is the reference position information of the vehicle target before the obscured vehicle target in the reference queue, length is the preset vehicle distance information, Y′1 is the position information of the obscured vehicle target in the reference queue, v is the speed information of the obscured vehicle target in the reference queue, and t is the time difference between the current monitoring time and the previous monitoring time.

[0117] Therefore, in this embodiment, in a congested state, please refer to Figure 9. Five vehicles are detected in the queue in the left figure. When the sixth vehicle is identified as entering the queue, the vehicle targets will be queued from near to far. When the first vehicle leaves the lane, vehicle target 1 will be kicked out of the queue.

[0118] As can be seen, adhering to the principle of front-to-rear vehicle order within the same lane, i.e., a rear vehicle cannot overtake a front vehicle within the same lane, this embodiment can configure preset inter-vehicle distance information based on the empirical value of the detected inter-vehicle distance in a congested state, thereby predicting the first predicted position information of the obscured vehicle target. It is worth noting that the specific value of the preset inter-vehicle distance information length can be adjusted based on the vehicle types of the front and rear vehicles. For example, the length between small vehicles can be 2 meters, while the length between large vehicles can be 3 meters.

[0119] However, it is understood that in congested conditions, vehicles may be traveling at high speeds, making the preset inter-vehicle distance information difficult to accurately reflect the inter-vehicle distance. Therefore, the speed at which the obscured vehicle disappears can be used to predict a second predicted position of the obscured vehicle. The maximum of the first and second predicted positions is then used as the current position of the obscured vehicle in the pending queue.

[0120] Furthermore, in one example, since the obscured vehicle target is in an obscured state, to ensure that it is obscured, it is also necessary to restrict the velocity information of the obscured vehicle target within the reference queue. This prevents the calculated second predicted position information of the obscured vehicle target from actually being in an obscured state. In this case, the adaptive function of Formula 4 can be changed to: Y1 = max(Y0 + length, Y′1 - max(v, 3) * t).

[0121] That is, in this example, a speed threshold of 3 m / s is introduced to limit the speed information of the obscured vehicle target in the reference queue, so as to avoid obvious distortion of the prediction result calculated accordingly.

[0122] Step S500: Based on the current position information of the at least one obscured vehicle target, the pending queue is updated to obtain a complete queue for the lane at the current moment.

[0123] Specifically, once the current position of the obscured vehicle in the pending queue is predicted, the current position information can be added to the pending queue, thereby filling in the missing data of the obscured vehicle in the pending queue and obtaining a complete queue. At this point, the traffic parameter processing equipment can collect traffic parameters such as the vehicle target status and lane status based on this information.

[0124] In a specific scenario, at the previous monitoring moment, all vehicle targets in the target lane are sorted from nearest to farthest to obtain a complete queue. The position information of the first vehicle in the pending queue at the current monitoring moment is then searched in the fused trajectory. If the pending queue obtained at the current monitoring moment still includes the first vehicle, the position and speed information of the first vehicle are updated. A query is then performed to obtain the second vehicle from the previous monitoring moment. If the pending queue obtained at the current monitoring moment still includes the second vehicle, the position and speed information of the second vehicle are updated. However, if the second vehicle is not detected, the second vehicle is determined to be a suspected obstructed vehicle target. A validity check is then performed on the second vehicle to determine whether it is an obstructed vehicle target. Once the second vehicle is determined to be an obstructed vehicle target, the current position information of the second vehicle in the pending queue at the current monitoring moment is predicted and added to the pending queue. This cycle continues until the current position information of the last vehicle in the current lane is added to the pending queue.

[0125] This demonstrates that this embodiment achieves more accurate queuing information acquisition and supports real-time output of the number of vehicles within a lane based on configured lane information. Furthermore, this more accurate queuing information acquisition improves the detection performance of vehicle trajectories within lanes, thereby increasing the accuracy of measuring the number of vehicles in the target area.

[0126] It is not difficult to see that after obtaining the pending queues of each lane based on the real-time traffic data collected by at least one sensor, this embodiment can use the current position information of the previous vehicle target and the position information and speed information of the obscured vehicle target in the reference queue, that is, the position information and speed information of the obscured rear vehicle before it disappears, to predict the current position information of the suspected obscured target in the pending queue. This no longer completely relies on the perceived real-time traffic data to judge the current data of the vehicle in the lane, reducing the impact of the difficulty of occlusion on the traffic parameter processing algorithm, thereby improving the traffic parameter measurement accuracy and traffic induction effect in dense scenes.

[0127] Based on the same inventive concept, referring to FIG10 , the present application further provides a traffic parameter processing device, comprising: a data acquisition module 101, configured to acquire real-time traffic data collected by at least one sensor for a target area at a current monitoring moment; a queue generation module 102, configured to obtain a pending queue for at least one lane in the target area based on the real-time traffic data and lane information of the target area, wherein the pending queue includes a plurality of vehicle targets arranged in sequence, and the vehicle targets have position information and speed information; a target determination module 103, configured to compare the pending queue with a reference queue for each lane in the at least one lane, and determine at least one obscured vehicle target in the pending queue. The reference queue is the complete queue of the lane at the last monitoring moment; the position prediction module 104 is used to predict the current position information of each obscured vehicle target among the at least one obscured vehicle target based on the reference position information and the position information and speed information of the obscured vehicle target in the reference queue, wherein the reference position information is the position information of the previous vehicle target of the obscured vehicle target in the pending queue in the reference queue; the queue improvement module 105 is used to update the pending queue based on the current position information of the at least one obscured vehicle target to obtain the complete queue of the lane at the current moment.

[0128] In a possible embodiment of the present application, the position prediction module 104 is specifically used to: obtain first predicted position information of the obscured vehicle target based on reference position information and preset vehicle distance information; obtain second predicted position information of the obscured vehicle target based on the position information and speed information of the obscured vehicle target in the reference queue; and use the maximum value of the first predicted position information and the second predicted position information as the current position information of the obscured vehicle target.

[0129] In a possible embodiment of the present application, the target determination module 103 is specifically used to: compare the pending queue queue with the reference queue queue to determine at least one suspected obscured target; for each suspected obscured target in the at least one suspected obscured target, determine the distance and intersection-and-union ratio between the suspected obscured target and the detectable vehicle target from the reference queue queue, wherein the detectable vehicle target is the previous vehicle target of the suspected obscured target in the reference queue queue; if the distance is less than a first preset threshold and the intersection-and-union ratio is greater than a second preset threshold, the suspected obscured target is determined as the obscured vehicle target.

[0130] In a possible embodiment of the present application, the sensor includes: a traffic radar for collecting real-time traffic point cloud data for the target area; and a visual sensor for collecting real-time traffic image data for the target area. Accordingly, the data acquisition module 101 is specifically configured to: perform target detection on the real-time traffic point cloud data to obtain one or more first vehicle targets in a radar coordinate system, wherein the radar coordinate system generates one or more corresponding virtual lanes based on lane information of the target area; perform target detection on the real-time traffic image data to obtain one or more second vehicle targets in an image coordinate system; determine one or more target matching groups by converting the one or more second vehicle targets into the radar coordinate system, wherein each target matching group includes a matching first vehicle target and a matching second vehicle target; for each target matching group, determine a valid vehicle target from the first vehicle target and the second vehicle target based on a confidence comparison result between the first vehicle target and the second vehicle target in the target matching group; and obtain a pending queue for at least one lane in the target area by generating the valid vehicle target in the corresponding virtual lane based on at least one valid vehicle target.

[0131] In a possible embodiment of the present application, the data acquisition module 101 is specifically used to: determine the confidence area of ​​the target matching group in the radar coordinate system, wherein, in the radar coordinate system, along the direction away from the traffic radar, it is sequentially divided into a visual high confidence area, an equal confidence area, and a radar high confidence area; if the confidence area is a visual high confidence area, the second vehicle target is taken as a valid vehicle target; if the confidence area is a radar high confidence area, the first vehicle target is taken as a valid vehicle target; if the confidence area is an equal confidence area, the first vehicle target and the second vehicle target are fused to obtain a valid vehicle target.

[0132] In a possible embodiment of the present application, the data acquisition module 101 is specifically used to identify the brightness value of the background area of ​​the real-time traffic image data; based on the brightness value of the background area and a preset dynamic coefficient, obtain a first distance between the first distal boundary of the visual high confidence area and the traffic radar; based on the farthest distance that the visual sensor can collect and the preset width value of the equal confidence area, obtain a second distance between the second distal boundary of the equal confidence area and the traffic radar; based on the first distance and the second distance, in the radar coordinate system, in the direction away from the traffic radar, sequentially separate the visual high confidence area, the equal confidence area and the radar high confidence area.

[0133] In a possible embodiment of the present application, the data acquisition module 101 is specifically used to: generate a valid vehicle target in a radar coordinate system; based on the position information of the valid vehicle target at the current monitoring moment and multiple previous consecutive monitoring moments, determine whether the valid vehicle target enters from one end of the virtual section identification area of ​​the virtual lane and exits from the other end of the virtual section identification area; if the vehicle target enters from one end of the virtual section identification area of ​​the virtual lane and exits from the other end, determine that the valid vehicle target belongs to the virtual lane, and update the traffic flow of the lane corresponding to the virtual lane.

[0134] It should be noted that the various implementations of the behavior recognition device in this embodiment and the technical effects achieved can refer to the various implementations of the behavior recognition method in the aforementioned embodiments, and will not be repeated here.

[0135] In addition, an embodiment of the present application further proposes a computer storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the traffic parameter processing method as described above are implemented. Therefore, no further description will be given here. In addition, the description of the beneficial effects of adopting the same method will not be repeated. For technical details not disclosed in the computer-readable storage medium embodiment involved in this application, please refer to the description of the method embodiment of this application. As an example, the program instructions can be deployed to be executed on one computing device, or on multiple computing devices located at one location, or on multiple computing devices distributed at multiple locations and interconnected by a communication network.

[0136] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware through a computer program. The above-described program can be stored in a computer-readable storage medium. When executed, the program can include the processes in the above-described method embodiments. The above-described storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0137] It should also be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive work.

[0138] Through the description of the above embodiments, it is clear to those skilled in the art that the present application can be implemented by means of software plus necessary general-purpose hardware, and of course it can also be implemented by means of dedicated hardware including application-specific integrated circuits, dedicated CPUs, dedicated memories, dedicated components, etc. In general, all functions performed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structures used to implement the same function can also be diverse, such as analog circuits, digital circuits, or dedicated circuits, etc. However, for the present application, software program implementation is a better implementation method in most cases. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer's floppy disk, USB flash drive, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., and includes a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0139] The above are some embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the description and drawings of this application, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A traffic parameter processing method, comprising: Acquire real-time traffic data collected by at least one sensor for a target area at a current monitoring time; Based on the real-time traffic data and the lane information of the target area, obtaining a pending queue of at least one lane in the target area, wherein the pending queue includes a plurality of vehicle targets arranged in sequence, and the vehicle targets have position information and speed information; For each lane of the at least one lane, Comparing the pending queue with a reference queue to determine at least one obscured vehicle target in the pending queue, wherein the reference queue is a complete queue of the lane at the last monitoring moment; For each of the at least one obscured vehicle target, predicting the current position information of the obscured vehicle target based on the reference position information and the position information and speed information of the obscured vehicle target in the reference queue, wherein the reference position information is the position information of the previous vehicle target of the obscured vehicle target in the pending queue in the reference queue; Based on the current position information of the at least one obscured vehicle target, the pending queue is updated to obtain a complete queue of the lane at the current moment.

2. The traffic parameter processing method according to claim 1, wherein: The predicting the current position information of the obscured vehicle target based on the reference position information and the position information and speed information of the obscured vehicle target in the reference queuing queue includes: Based on the reference position information and the preset vehicle distance information, obtaining first predicted position information of the obscured vehicle target; Based on the position information and speed information of the obscured vehicle target in the reference queuing queue, obtaining second predicted position information of the obscured vehicle target; The maximum value of the first predicted position information and the second predicted position information is used as the current position information of the obscured vehicle target.

3. The traffic parameter processing method according to claim 1, wherein: The comparing the pending queue and the reference queue to determine at least one obscured vehicle target in the pending queue includes: Comparing the pending queue with the reference queue to determine at least one suspected obscured target; For each suspected occluded target in the at least one suspected occluded target, Determine the distance and intersection-over-joint ratio between the suspected obscured target and the detectable vehicle target from the reference queue, wherein the detectable vehicle target is the vehicle target preceding the suspected obscured target in the reference queue; If the distance is smaller than a first preset threshold and the intersection-over-union ratio is larger than a second preset threshold, the suspected obscured target is determined to be the obscured vehicle target.

4. The traffic parameter processing method according to any one of claims 1 to 3, wherein: The at least one sensor comprises: Traffic radar, used to collect real-time traffic point cloud data for the target area; and A visual sensor is used to collect real-time traffic image data for the target area. The obtaining, based on the real-time traffic data and the lane information of the target area, a pending queue of at least one lane in the target area comprises: Performing target detection on the real-time traffic point cloud data to obtain one or more first vehicle targets in a radar coordinate system, wherein the radar coordinate system generates one or more corresponding virtual lanes based on lane information of the target area; Performing target detection on the real-time traffic image data to obtain one or more second vehicle targets in an image coordinate system; Determine one or more target matching groups by converting the one or more second vehicle targets into the radar coordinate system, wherein each of the target matching groups includes one first vehicle target and one second vehicle target that match each other; For each of the target matching groups, based on a confidence comparison result between the first vehicle target and the second vehicle target in the target matching group, determining a valid vehicle target from the first vehicle target and the second vehicle target; By generating the valid vehicle target in the corresponding virtual lane based on at least one valid vehicle target, a pending queue queue of at least one lane in the target area is obtained.

5. The traffic parameter processing method according to claim 4, wherein: The determining a valid vehicle target from the first vehicle target and the second vehicle target based on the confidence comparison result between the first vehicle target and the second vehicle target in the target matching group includes: Determine a confidence region of the target matching group in the radar coordinate system, wherein, in the radar coordinate system, along a direction away from the traffic radar, the confidence region is sequentially divided into a visual high confidence region, an equal confidence region, and a radar high confidence region; If the confidence region is a visual high confidence region, taking the second vehicle target as the valid vehicle target; If the confidence area is a radar high confidence area, taking the first vehicle target as the valid vehicle target; If the confidence region is an equal confidence region, the first vehicle target and the second vehicle target are fused to obtain the valid vehicle target.

6. The traffic parameter processing method according to claim 5, further comprising: Identify the background area brightness value of the real-time traffic image data; Based on the brightness value of the background area and a preset dynamic coefficient, obtaining a first distance between a first far-side boundary of the visual high confidence area and the traffic radar; Based on the farthest distance that can be collected by the visual sensor and the preset width value of the equal confidence area, obtaining a second distance between a second far side boundary of the equal confidence area and the traffic radar; Based on the first distance and the second distance, in the radar coordinate system, along the direction away from the traffic radar direction, and sequentially separates the visual high confidence area, the equal confidence area, and the radar high confidence area.

7. The traffic parameter processing method according to claim 4, wherein: After determining the valid vehicle target from the first vehicle target and the second vehicle target based on the confidence comparison result between the first vehicle target and the second vehicle target in the target matching group, the method further includes: generating the valid vehicle target in the radar coordinate system; Based on the position information of the valid vehicle target at the current monitoring time and the previous multiple consecutive monitoring times, determining whether the valid vehicle target enters from one end of the virtual section identification area of ​​the virtual lane and exits from the other end of the virtual section identification area; If the vehicle target enters from one end of the virtual section identification area of ​​the virtual lane and exits from the other end, it is determined that the valid vehicle target belongs to the virtual lane, and the traffic flow of the lane corresponding to the virtual lane is updated.

8. A traffic parameter processing device, wherein: include: A data acquisition module, used to acquire real-time traffic data collected by at least one sensor for a target area at a current monitoring time; a queue generation module, configured to obtain a pending queue of at least one lane in the target area based on the real-time traffic data and the lane information of the target area, wherein the pending queue includes a plurality of vehicle targets arranged in sequence, and the vehicle targets have position information and speed information; a target determination module, for each of the at least one lane, comparing the pending queue with a reference queue to determine at least one obscured vehicle target in the pending queue; the reference queue is a complete queue of the lane at the last monitoring moment; A position prediction module, configured to predict, for each of the at least one obscured vehicle target, current position information of the obscured vehicle target based on reference position information and position information and speed information of the obscured vehicle target in the reference queue, wherein the reference position information is position information of a previous vehicle target of the obscured vehicle target in the pending queue in the reference queue; The queue improvement module is used to update the pending queue based on the current position information of the at least one obscured vehicle target, and obtain the complete queue of the lane at the current moment.

9. A traffic parameter processing device, comprising: A processor, a memory and a computer program stored in the memory, wherein the computer program, when executed by the processor, implements the traffic parameter processing method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the traffic parameter processing method according to any one of claims 1 to 7.

Images