CN111540201A - Vehicle queuing length real-time estimation method and system based on roadside laser radar - Google Patents

Abstract

The invention discloses a real-time estimation method and system of vehicle queue length based on roadside laser radar, which acquires all position points scanned by roadside laser radar for vehicles on the road, and obtains three-dimensional point cloud data to be processed; The data is filtered in sequence; the 3D point cloud data after background filtering is clustered; the target recognition is performed on the 3D point cloud data after clustering, and the vehicles on the road are identified; Lane recognition; estimate the speed of each vehicle on the road based on the 3D point cloud data of adjacent frames of the lidar; determine the last vehicle on each lane based on the speed of each vehicle on the road, and then estimate the queue length. The vehicle queue length is estimated in real time, new traffic control measures are implemented in real time according to the vehicle queue length, the overall vehicle traffic efficiency is optimized, and the problems of vehicle congestion and queue spread during peak periods are solved.

Description

Real-time estimation method and system of vehicle queue length based on roadside lidar

technical field

The present disclosure relates to the technical field of traffic engineering, and in particular, to a method and system for real-time estimation of vehicle queuing length based on roadside lidar.

Background technique

The statements in this section merely mention background related to the present disclosure and do not necessarily constitute prior art.

The information of vehicle queue length has applications in many aspects of the transportation field, such as using it to evaluate the performance of signalized intersections, adaptive signal control, trip routing, etc. Some of these applications, such as optimal signal control and trip routing, require real-time access to vehicle queue lengths. Existing research shows that the vehicle queue length can be obtained by either estimation or direct detection. However, traditional estimation methods have problems such as being unable to reflect the vehicle queue length in real time. Therefore, in recent years, with the development of "smart transportation" related technologies, under the background of intelligent vehicles and roads, many scholars have begun to use the technology of Internet of Vehicles to conduct related research on vehicle queuing length. It is based on the assumption that all vehicles are in the "Internet of Vehicles", and the fact today is that the application rate of vehicle network technology in current vehicles is very low, and will be in a low application rate for a certain period of time in the future. in the case of.

In the process of realizing the present disclosure, the inventor found that the following technical problems exist in the prior art:

Although the Internet of Vehicles technology can indeed provide accurate and valuable information for estimating the length of vehicle queues, related methods are limited due to the low application rate of the vehicle network. In order to solve the above problems, it is necessary to find a method that does not depend on the "Internet of Vehicles", and can also estimate the queue length of vehicles in real time and with high accuracy.

SUMMARY OF THE INVENTION

In order to solve the deficiencies of the prior art, the present disclosure provides a real-time estimation method and system for vehicle queuing length based on roadside LiDAR; by estimating the vehicle queuing length, new traffic control measures can be taken in real time to optimize the overall vehicle traffic efficiency of the transportation network , to solve the problems of vehicle congestion and queuing spread during peak hours.

In a first aspect, the present disclosure provides a real-time estimation method of vehicle queue length based on roadside lidar;

A real-time estimation method of vehicle queue length based on roadside lidar, including:

Obtain all the position points scanned by the roadside lidar for vehicles on the road, and obtain the 3D point cloud data to be processed;

Perform background filtering on the 3D point cloud data to be processed in turn; perform clustering processing on the 3D point cloud data after background filtering, and divide all points belonging to the same object into one category;

Perform target recognition on the clustered 3D point cloud data to identify vehicles on the road;

Perform lane recognition based on the result of target recognition;

Estimate the speed of each vehicle on the road based on the 3D point cloud data of adjacent frames of the lidar;

Based on the speed of each vehicle on the road, the last vehicle in each lane is determined and the queue length in each lane is estimated.

In a second aspect, the present disclosure provides a real-time estimation device for vehicle queuing length based on roadside lidar;

A real-time estimation device for vehicle queue length based on roadside lidar, including:

an acquisition module, which is configured to: acquire all the position points scanned by the roadside lidar for vehicles on the road, and obtain the three-dimensional point cloud data to be processed;

The background filtering module is configured to: sequentially perform background filtering on the three-dimensional point cloud data to be processed;

a clustering module, which is configured to: perform clustering processing on the three-dimensional point cloud data after background filtering, and divide all points belonging to the same object into one category;

a target recognition module, which is configured to: perform target recognition on the clustered 3D point cloud data, and identify vehicles on the road;

a lane recognition module, which is configured to: perform lane recognition based on the result of the target recognition;

a vehicle speed estimation module, which is configured to: estimate the speed of each vehicle on the road based on the three-dimensional point cloud data of the adjacent frames of the lidar;

A queue length estimation module configured to: determine the last vehicle on each lane based on the speed of each vehicle on the road, thereby estimating the queue length on each lane.

In a third aspect, the present disclosure also provides an electronic device, including a memory, a processor, and computer instructions stored in the memory and executed on the processor, and when the computer instructions are executed by the processor, the first aspect is completed. Methods.

In a fourth aspect, the present disclosure further provides a computer-readable storage medium for storing computer instructions that, when executed by a processor, complete the method of the first aspect.

In a fifth aspect, the present disclosure provides a real-time estimation system for vehicle queue length based on roadside lidar;

A real-time estimation system of vehicle queue length based on roadside lidar;

A real-time estimation system for vehicle queue length based on roadside lidar, including:

A lidar module, a communication module, and the device for real-time estimation of vehicle queue length based on roadside lidar as described in Embodiment 2;

The lidar module is used to scan vehicles within the scanning range, obtain three-dimensional point cloud data, and upload the point cloud data to a real-time estimation device for vehicle queuing length based on roadside lidar; vehicle queuing based on roadside lidar The real-time length estimation device is used to receive the three-dimensional point cloud data collected by the lidar for processing; transmit the processing results to the traffic control terminal through the communication module;

The traffic control terminal implements new traffic control measures based on the processing results of the real-time estimation device for vehicle queuing length by roadside lidar.

Compared with the prior art, the beneficial effects of the present disclosure are:

Based on the processing and analysis of the point cloud data collected by the lidar, the estimation of the vehicle queuing length is completed on this basis, which solves the problem that the current vehicle queuing length cannot be estimated in real time. Based on the estimated queuing length, a certain Traffic control measures can reduce queue lengths in real time, ease traffic congestion, and improve road traffic efficiency.

Description of drawings

The accompanying drawings that constitute a part of the present disclosure are used to provide further understanding of the present disclosure, and the exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure.

FIG. 1 is a schematic diagram of a method for estimating vehicle queuing length in real time by using roadside lidar according to Embodiment 1 of the present disclosure;

FIG. 2 is a schematic diagram of a system for estimating vehicle queuing length in real time by using roadside lidar according to Embodiment 1 of the present disclosure;

3 is a schematic diagram of missing point cloud data in Embodiment 1 of the present disclosure;

4 is a schematic diagram of missing point cloud data due to occlusion in Embodiment 1 of the present disclosure;

5 is a schematic diagram of missing point cloud data due to packet loss according to Embodiment 1 of the present disclosure;

Fig. 6 is the confirmation of the vehicle at the end of the fleet according to the first embodiment of the present disclosure, case 1;

Fig. 7 is the confirmation of the vehicle at the end of the convoy according to the first embodiment of the present disclosure, case 2;

FIG. 8 is a schematic diagram of the installation layout of the roadside laser radar according to the first embodiment of the present disclosure.

Detailed ways

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the present disclosure. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.

It should be noted that the terminology used herein is for the purpose of describing specific embodiments only, and is not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural as well, furthermore, it is to be understood that when the terms "comprising" and/or "including" are used in this specification, it indicates that There are features, steps, operations, devices, components and/or combinations thereof.

Embodiment 1, this embodiment provides a real-time estimation method of vehicle queue length based on roadside lidar;

As shown in Figure 1, the real-time estimation method of vehicle queue length based on roadside lidar includes:

S1: Obtain all the position points scanned by the roadside lidar for vehicles on the road, and obtain the 3D point cloud data to be processed;

S2: perform background filtering in sequence on the 3D point cloud data to be processed; perform clustering processing on the 3D point cloud data after background filtering, and divide all points belonging to the same object into one category;

Perform target recognition on the clustered 3D point cloud data to identify vehicles on the road;

Perform lane recognition based on the result of target recognition;

S3: Estimate the speed of each vehicle on the road based on the 3D point cloud data of the adjacent frames of the lidar;

Based on the speed of each vehicle on the road, the last vehicle in each lane is determined and the queue length in each lane is estimated.

Further, the lidar is installed on the utility pole beside the road and is 3-5 meters away from the ground.

Further, the three-dimensional coordinates of each position point are three-dimensional coordinates x, y, and z with the position where the lidar is located as the origin of the Cartesian coordinate system.

Further, the data file obtained after the lidar scans the object is further explained: the file is in csv format, which contains countless points and the coordinates (x, y, z) corresponding to the point, and the lidar runs within the time. Each frame of , will have such a data file, that is, all scanned objects will be re-rendered in the form of point clouds.

As one or more embodiments, the specific steps of sequentially performing background filtering on the 3D point cloud data to be processed include:

S201: Acquire 3D point cloud data when there is no vehicle on the road within a set time period;

S202: Segment the vehicle-free 3D point cloud data with cubes of the same size, that is, perform rasterization processing, set a point cloud density threshold, and select the cubes whose point cloud density is greater than the point cloud density threshold as background point cubes ; store the points contained in all the selected background point cubes into a matrix, and the matrix is regarded as a background matrix;

S203: Subtract the background matrix from the three-dimensional point cloud data to be processed to obtain three-dimensional point cloud data after background filtering.

It should be understood that the beneficial effect of the background filtering is to filter out the point cloud (background point) data of objects other than road users.

As one or more embodiments, the clustering process is performed on the 3D point cloud data after background filtering, and all points belonging to the same object are divided into one category, and the specific steps include:

Based on the DBSCAN algorithm, the 3D point cloud data after background filtering is clustered, and all points belonging to the same object are clustered together.

It should be understood that the point cloud in the lidar data belongs to the disordered point cloud, that is to say, the points belonging to the same object are not clustered together. Clustering points that belong to the same object and naming these points as a unified ID is helpful for subsequent data processing.

As one or more embodiments, performing target recognition on the clustered 3D point cloud data to identify vehicles on the road; specific steps include:

Extract the characteristics of each clustered object from the clustered 3D point cloud data;

The features of each clustered object are input into the pre-trained random forest classifier, and the current object category is output; according to the current object category and the clustering results, the body length of each vehicle on the road is obtained.

Further, the feature of extracting each clustered object includes:

The length, height, ratio of length and height of each clustered object, the distance between the current clustered object and the lidar, the number of points included in the current clustered object, and the outline of the current clustered object are extracted.

Further, the training process of the pre-trained random forest model includes:

Extract the features of the known object category labels, input the features of the known object category labels into the random forest model for training, and obtain a trained random forest model.

It should be understood that after point cloud clustering, there are various road users (cars, bicycles, electric vehicles, pedestrians, etc.) on the road. Classification. By selecting the features of 6 objects (target length, target height, difference between target height and target length, distance from lidar, number of points, and target contour), a target classifier is established (this disclosure adopts An RF (Random Forest) classifier, trained through machine learning), followed by object classification.

As one or more embodiments, the specific steps of performing lane recognition based on the result of target recognition include:

The point cloud density threshold is set, and the area in the 3D point cloud data whose point cloud density is greater than the point cloud density threshold is regarded as a lane.

After the target classification step is completed, there is only one road user on the road. It needs to be clear that the estimated vehicle queue length is for a certain lane, so it is necessary to distinguish and identify the lanes within the scanning range of the lidar. In the lane recognition step, it is assumed that the vehicle will not change lanes when approaching the intersection, that is, the vehicles are driving in a straight line, so the density of the point cloud in each lane is higher than the point on the dividing line between the lane and the lane. Cloud density. Based on this, similar to the method of finding background points, the lane is identified by comparing the density of the point cloud.

As one or more embodiments, the speed of each vehicle on the road is estimated based on the three-dimensional point cloud data of the adjacent frames of the lidar, and the specific steps include:

In several frames of data scanned by lidar, the same vehicle is tracked, and a point in the point cloud of the tracked vehicle that is closest to the lidar is selected, and the speed of the vehicle can be obtained by using the change of the coordinates of the same point in adjacent frames.

It should be understood that when selecting a point in the tracked vehicle point cloud that is closest to the lidar, the speed of the vehicle can be calculated by using the change in the coordinates of the same point in adjacent frames. The specific formula is as follows:

Among them, V represents the speed of the vehicle, F represents the rotation frequency of the lidar, unit: HZ, X _i represents the abscissa of a point in the vehicle point cloud in the i-th frame closest to the lidar, and Y _i represents the vehicle point in the i-th frame. The vertical coordinate of a point in the cloud that is closest to the lidar, Z _i represents the vertical coordinate of a point in the vehicle point cloud in the i-th frame that is closest to the lidar, and X _i-1 represents the vehicle point cloud in the i-1th frame. The abscissa of the point closest to the lidar, Y _i-1 represents the ordinate of the point closest to the lidar in the vehicle point cloud in the i-1th frame, and Z _i-1 represents the vehicle point in the i-1th frame The vertical coordinate of a point in the cloud that is closest to the lidar.

Further, in several frames of data scanned by the lidar, the same vehicle is tracked, and the global nearest neighbor algorithm GNN is used.

It should be understood that the information of the vehicle speed is extremely critical for finding the vehicle at the end of the convoy, because the speed of the vehicle at the end of the convoy is very low or even stationary, so the vehicle at the end of the convoy can be determined according to the information of the vehicle speed. Therefore, it is necessary to continuously track the same vehicle between different frames (the present disclosure adopts the global nearest neighbor algorithm GNN for vehicle tracking). ) to get a rough estimate of its speed.

As one or more embodiments, determining the last vehicle on each lane based on the speed of each vehicle on the road, and then estimating the queue length on each lane, the specific steps include:

Identify the last vehicle in each lane;

Determine the length of the last vehicle body on each lane;

The queue length in each lane is calculated based on the last vehicle in each lane, the body length of the last vehicle, and the body lengths of other vehicles in the current lane.

Further, the specific steps for determining the last vehicle on each lane include:

When the _nth vehicle appears in the i-th lidar scan image and also appears in the i+1-th lidar scan image, calculate the speed V of the n-th vehicle at the i-th frame and the i+1-th frame If V'<V and V'<5km/h, the vehicle is considered to be at the end of the convoy, otherwise, the n-1th vehicle is considered to be at the end of the convoy.

When the nth vehicle appears in the i-th lidar scanning image and disappears in the i+1-th lidar scanning image, the distance between the n-1th vehicle and the n-2th vehicle is obtained, Indirectly estimate the distance between the nth vehicle and the n-1th vehicle; assuming that the distance between the nth vehicle and the n-1th vehicle is equal to the distance between the n-1th vehicle and the n-2th vehicle spacing between;

Calculate the speed V of the nth vehicle in the ith frame and the speed V' of the nth vehicle in the i+1th frame. When V'<V and V'<5km/h, it is considered that the nth vehicle is at the end of the convoy , otherwise the n-1th car is considered to be at the end of the convoy.

When the n-th vehicle appears in the i-th lidar scan image and disappears in the i+1-th lidar scan image, it is considered that the n-th vehicle is occluded, or the data is missing due to packet loss.

After the previous background filtering, clustering and other operations, the point clouds related to each vehicle on the lane are classified into one category, so the length of each vehicle and the distance between the vehicles can be known. The distance between vehicles refers to the previous The distance between the rear of a vehicle and the front of the following vehicle.

The reason why it is necessary to determine which car is the last vehicle is because the farther away from the lidar, the fewer point clouds, the incomplete detection of the vehicle, and the inability to accurately determine the length of the vehicle body. Assuming that the respective vehicle body lengths (except the last one) are x1, x2...xn-1, plus the determined vehicle body length xn at the end, the fleet length can be calculated.

As shown in Figure 5, assuming that there are n vehicles in this lane, when the nth vehicle appears in the ith frame and disappears in the i+1th frame, it can be considered that the vehicle is at the end of the convoy. At this time, about the nth vehicle The distance between the car and the n-1th car can be obtained by the backward method, that is, the distance between the n-1th car and the n-2th car is used to indirectly estimate the nth car and the n-th car. The distance between n-1 vehicles, so that the speed at the i-th frame and the i+1-th frame can be calculated (respectively denoted as V and V'), when V'<V and V'<5km/h, it is considered that The car is at the end of the convoy, as shown in Figure 6, otherwise it is considered that the n-1th car is at the end of the convoy.

It should be understood that determining the vehicle at the end of the convoy is a key factor in estimating the length of the vehicle queue, because when the last vehicle in the convoy is determined, the queue length can also be determined.

It should be understood that in determining the position of the end vehicle on each lane, the following two situations are considered:

(1) As shown in Figure 3, large trucks will block small vehicles in their adjacent lanes;

(2) As shown in Figure 4, the connection between the lidar and the computer for data processing is not stable enough to cause data packet loss. Both of the situations described result in a sectorial defect in the point cloud, which in turn makes it impossible to see some vehicles in the convoy.

Further, to determine the vehicle body length at the end of the fleet, the specific steps include:

When the measured length of the vehicle at the end of the fleet is greater than 6m, the measured length is taken as the length of the vehicle at the end of the fleet. If the measured length of the vehicle at the end of the fleet is less than 6m, the default vehicle length at the end of the fleet is 6m.

The reason for determining the length of the vehicle at the end of the fleet is because, as shown in Figure 7, when the road is blocked and the vehicle queue is too long, the last vehicle in the fleet may not be fully detected (as the distance from the lidar increases, the density of the point cloud increases). The lower it is, the point cloud cannot show the full picture of the car).

Therefore, the following choices can be made for the measured data of vehicles at the end of the fleet:

(1) Only when the speed of the car is below 5km/h can it be considered that the car is at the end of the convoy;

(2) It is easy to know that the length of a small car is less than 6m. When the measured body length is greater than 6m, the measured length is taken as the body length. If the measured length is less than 6m, then the default body length is 6m, which is summarized as follows:

A method and system for real-time estimation of vehicle queuing length using roadside lidar proposed in this embodiment, by acquiring vehicle information on the road in real time, generating point cloud data with high precision, and performing a series of processing on these point cloud data , the vehicle queuing length is estimated, and the traffic control terminal dynamically implements new traffic control measures based on this, thereby alleviating road traffic congestion and improving road traffic efficiency. At present, there is no effective real-time method and means for obtaining the vehicle queuing length, and the embodiments of the present invention can make up for this application gap.

Embodiment 2, this embodiment provides a real-time estimation device for vehicle queuing length based on roadside lidar;

A real-time estimation device for vehicle queue length based on roadside lidar, including:

an acquisition module, which is configured to: acquire all the position points scanned by the roadside lidar for vehicles on the road, and obtain the three-dimensional point cloud data to be processed;

The background filtering module is configured to: sequentially perform background filtering on the three-dimensional point cloud data to be processed;

a clustering module, which is configured to: perform clustering processing on the three-dimensional point cloud data after background filtering, and divide all points belonging to the same object into one category;

a target recognition module, which is configured to: perform target recognition on the clustered 3D point cloud data, and identify vehicles on the road;

a lane recognition module, which is configured to: perform lane recognition based on the result of the target recognition;

a vehicle speed estimation module, which is configured to: estimate the speed of each vehicle on the road based on the three-dimensional point cloud data of the adjacent frames of the lidar;

A queue length estimation module configured to: determine the last vehicle on each lane based on the speed of each vehicle on the road, thereby estimating the queue length on each lane.

Embodiment 3, this embodiment also provides an electronic device, including a memory, a processor, and computer instructions stored in the memory and run on the processor, and when the computer instructions are run by the processor, the first embodiment is completed. method described.

In Embodiment 4, this embodiment further provides a computer-readable storage medium for storing computer instructions, and when the computer instructions are executed by a processor, the method described in Embodiment 1 is completed.

Embodiment 5, this embodiment provides a real-time estimation system for vehicle queuing length based on roadside lidar; as shown in FIG. 2 , a real-time estimation system for vehicle queuing length based on roadside lidar;

A real-time estimation system for vehicle queue length based on roadside lidar, including:

A lidar module, a communication module, and the real-time estimation device for vehicle queuing length based on roadside lidar as described in Embodiment 2;

The lidar module is used to scan vehicles within the scanning range, obtain three-dimensional point cloud data, and upload the point cloud data to a real-time estimation device for vehicle queuing length based on roadside lidar; vehicle queuing based on roadside lidar The real-time length estimation device is used to receive the three-dimensional point cloud data collected by the lidar for processing; transmit the processing results to the traffic control terminal through the communication module;

The traffic control terminal implements new traffic control measures based on the processing results of the real-time estimation device for vehicle queue length by roadside lidar.

The laser radar module includes: a laser radar, a processor and a data transmission line connected in sequence.

Further, as shown in Figure 8, the lidar (either rotating lidar, solid-state lidar, etc.) is installed on the roadside telephone pole or signal pole, 3 to 5 meters away from the ground, and can be installed by means of lift trucks, ladders, etc. , the specific method is selected according to the actual situation. At the lower position of the pole, about 1m from the ground, a processor with data processing algorithms is installed. The two are connected by a data transmission line and covered with an iron box to avoid being affected by the surrounding environment. Both the lidar and the processor need power supply, so the power supply needs can be met by routing inside the rod. In order to transmit the results processed by the processor, it is necessary to connect the optical fiber to the processor.

The traffic control terminal receives the results of the point cloud data collected and processed by the drive test lidar module, analyzes the results (the data analysis unit works), and executes new traffic control measures in real time. In this embodiment, taking a signal light as an example, the duration of the traffic light at the intersection can be reconfigured in real time (the control unit works), thereby reducing the queue length of vehicles, so as to achieve the purpose of alleviating traffic congestion and reducing parking time.

The information communication module is used to transmit the information of the lidar module and the traffic control terminal.

The above descriptions are only preferred embodiments of the present disclosure, and are not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included within the protection scope of the present disclosure.

Claims (10)

1. A real-time estimation method for vehicle queuing length based on roadside lidar, characterized in that it includes: Obtain all the position points scanned by the roadside lidar for vehicles on the road, and obtain the 3D point cloud data to be processed; Perform background filtering on the 3D point cloud data to be processed in turn; perform clustering processing on the 3D point cloud data after background filtering, and divide all points belonging to the same object into one category; Perform target recognition on the clustered 3D point cloud data to identify vehicles on the road; Perform lane recognition based on the result of target recognition; Estimate the speed of each vehicle on the road based on the 3D point cloud data of adjacent frames of the lidar; Based on the speed of each vehicle on the road, the last vehicle in each lane is determined and the queue length in each lane is estimated. 2. The method according to claim 1, wherein the concrete steps of sequentially performing background filtering on the three-dimensional point cloud data to be processed include: Obtain 3D point cloud data when there are no vehicles on the road within a set time period; The vehicle-free 3D point cloud data is divided by cubes of the same size, that is, rasterization is performed, the point cloud density threshold is set, and the cube whose point cloud density is greater than the point cloud density threshold is selected as the background point cube; The points contained in all the selected background point cubes are stored in a matrix, and the matrix is regarded as a background matrix; The background matrix is subtracted from the 3D point cloud data to be processed to obtain the 3D point cloud data after background filtering. 3. The method according to claim 1, wherein the three-dimensional point cloud data after background filtering is subjected to clustering processing, and all points belonging to the same object are divided into one class, and the specific steps include: Based on the DBSCAN algorithm, the 3D point cloud data after background filtering is clustered, and all points belonging to the same object are clustered together. 4. The method according to claim 1, wherein the three-dimensional point cloud data after the clustering process is subjected to target recognition to identify vehicles on the road; the specific steps include: Extract the characteristics of each clustered object from the clustered 3D point cloud data; Input the features of each clustered object into the pre-trained random forest classifier, and output the category of the current object; according to the category of the current object and the result of the clustering, the body length of each vehicle on the road is obtained; or, The feature of extracting each clustered object includes: Extract the length, height, ratio of length and height of each clustered object, the distance between the current clustered object and the lidar, the number of points included in the current clustered object, and the outline of the current clustered object; or, The training process of the pre-trained random forest model includes: Extract the features of the known object category labels, input the features of the known object category labels into the random forest model for training, and obtain a trained random forest model. 5. The method according to claim 1, wherein the lane recognition is performed based on the result of the target recognition, and the specific steps include: Set the point cloud density threshold, and regard the area where the point cloud density is greater than the point cloud density threshold in the 3D point cloud data as a lane; or, The speed of each vehicle on the road is estimated based on the three-dimensional point cloud data of the adjacent frames of the lidar, and the specific steps include: In several frames of data scanned by lidar, the same vehicle is tracked, and a point in the point cloud of the tracked vehicle that is closest to the lidar is selected, and the speed of the vehicle can be obtained by using the change of the coordinates of the same point in adjacent frames. 6. The method according to claim 1, wherein, based on the speed of each vehicle on the road, determining the last vehicle on each lane, and then estimating the queue length on each lane, the specific steps include: Identify the last vehicle in each lane; Determine the length of the last vehicle body on each lane; Calculate the queue length in each lane according to the last vehicle in each lane, the body length of the last vehicle and the body lengths of other vehicles in the current lane; or, The specific steps for determining the end vehicle on each lane include: When the nth vehicle appears in the i-th lidar scan image and also appears in the i+1-th lidar scan image, calculate the speed V of the n-th vehicle at the i-th frame and the i+1-th frame speed and V ^′ , if V ^′ < and V ^′ < 5km/h, the vehicle is considered to be at the end of the convoy, otherwise, the n-1th vehicle is considered to be at the end of the convoy; When the nth vehicle appears in the i-th lidar scanning image and disappears in the i+1-th lidar scanning image, the distance between the n-1th vehicle and the n-2th vehicle is obtained, Indirectly estimate the distance between the nth vehicle and the n-1th vehicle; assuming that the distance between the nth vehicle and the n-1th vehicle is equal to the difference between the n-1th vehicle and the n-2th vehicle distance between; calculate the speed V of the nth vehicle in the ith frame and the speed V ^′ of the nth vehicle in the i+1th frame, when V ^′ < and V ^′ <5km/h, it is considered that the nth vehicle is in the At the end of the convoy, on the contrary, the n-1th car is considered to be at the end of the convoy; or, To determine the length of the vehicle at the end of the fleet, the specific steps include: When the measured length of the vehicle at the end of the fleet is greater than 6m, the measured length is used as the length of the vehicle at the end of the fleet; if the measured length of the vehicle at the end of the fleet is less than 6m, the default vehicle length at the end of the fleet is 6m. 7. A real-time estimation device for vehicle queuing length based on roadside lidar, characterized in that it includes: an acquisition module, which is configured to: acquire all the position points scanned by the roadside lidar for vehicles on the road, and obtain the three-dimensional point cloud data to be processed; The background filtering module is configured to: sequentially perform background filtering on the three-dimensional point cloud data to be processed; a clustering module, which is configured to: perform clustering processing on the three-dimensional point cloud data after background filtering, and divide all points belonging to the same object into one category; a target recognition module, which is configured to: perform target recognition on the clustered 3D point cloud data, and identify vehicles on the road; a lane recognition module, which is configured to: perform lane recognition based on the result of the target recognition; a vehicle speed estimation module, which is configured to: estimate the speed of each vehicle on the road based on the three-dimensional point cloud data of the adjacent frames of the lidar; A queue length estimation module configured to: determine the last vehicle on each lane based on the speed of each vehicle on the road, thereby estimating the queue length on each lane. 8. an electronic device, is characterized in that, comprises memory and processor and the computer instruction that is stored on memory and runs on processor, when described computer instruction is run by processor, completes any one of claim 1-6 the method described. 9 . A computer-readable storage medium, characterized in that it is used for storing computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1-6 is completed. 10 . 10. A real-time estimation system for vehicle queuing length based on roadside lidar, characterized in that it includes: A lidar module, a communication module, and a real-time estimation device for vehicle queuing length based on roadside lidar as claimed in claim 7; The lidar module is used to scan vehicles within the scanning range, obtain three-dimensional point cloud data, and upload the point cloud data to a real-time estimation device for vehicle queuing length based on roadside lidar; vehicle queuing based on roadside lidar The real-time length estimation device is used to receive the three-dimensional point cloud data collected by the lidar for processing; transmit the processing results to the traffic control terminal through the communication module; The traffic control terminal implements new traffic control measures based on the processing results of the real-time estimation device for vehicle queuing length by roadside lidar.

Images