Intelligent traffic-oriented in-tunnel high-precision radar speed measurement method

Abstract

The invention relates to the technical field of radar measurement, in particular to an in-tunnel high-precision radar speed measurement method for intelligent traffic. According to the method, initial point cloud data of different radars are matched by using a matching method, an information weight corresponding to each radar is determined by combining point cloud quality, and then effective point cloud data is determined. And further performing weighted fusion on the point cloud speed and the Doppler speed to obtain movement information of each vehicle area. The interference degree of the vehicle area at each moment is determined, and then the influence degree of each historical moment can be determined by combining the information between the historical moment and the real-time moment. Predicated movement information is determined according to the influence degree, and adjustment is conducted through movement information of the front vehicle. According to the method, the influence of the multipath effect in the tunnel is eliminated, and effective speed measurement and accurate prediction of each vehicle are realized.

Description

Technical Field

[0001] The present invention relates to the field of radar measurement technology, and in particular to a high-precision radar speed measurement method in a tunnel for smart transportation. Background Art

[0002] Millimeter-wave radar is often used to measure vehicle speeds in tunnels. The speed measurement results are uploaded to the cloud using IoT technology, enabling intelligent operation and management of tunnel traffic. Due to the complex tunnel structure and vehicle environment, millimeter-wave signals can be affected by multipath effects, leading to errors caused by signal reflection and refraction, reducing measurement accuracy. For example, when millimeter-wave radar is used to measure speed in a tunnel, the concrete and metal structures on the tunnel walls reflect the radar signal, causing the electromagnetic wave from the same target to return to the receiving device through multiple paths, including direct reflection. This results in multipath interference in the measured results, interfering with the accuracy of the speed measurement.

[0003] When conducting multipath reflection interference on millimeter-wave radar inside a tunnel, existing technologies usually distinguish the measured targets based on the distance, speed, azimuth, and pitch angle of the targets measured by the radar, thereby achieving speed measurement of moving targets. However, the distinction accuracy of the above method is affected in complex environments. For example, when the traffic volume in the tunnel is large and there are large vehicles blocking the view, the multipath interference phenomenon may be aggravated, which is not conducive to accurate speed measurement of moving vehicles. Summary of the Invention

[0004] In order to solve the technical problem that the existing technology cannot effectively analyze the data collected by radar and cannot accurately measure the speed of vehicles in tunnels due to the complex vehicle environment inside the tunnel, the purpose of the present invention is to provide a high-precision radar speed measurement method in tunnels for smart transportation. The technical solution adopted is as follows: The present invention proposes a high-precision radar speed measurement method in a tunnel for smart transportation, the method comprising: Multiple radars are installed at monitoring locations within the tunnel to obtain initial point cloud data acquired by each radar at the time of acquisition. Based on the degree of matching between the initial point clouds and the point cloud quality of each initial point cloud, an information weight for each initial point cloud data is obtained. All initial point cloud data are registered using the information weight to obtain valid point cloud data. Obtain the vehicle area in the valid point cloud data, obtain the point cloud velocity of each vehicle area at each moment through the movement changes of the vehicle area at adjacent moments, perform weight analysis using the point cloud density, and perform weighted fusion of the point cloud velocity and Doppler velocity to obtain the movement information of each vehicle area at real time; For each vehicle area, the interference degree of each vehicle area is obtained based on the Doppler velocity stability of the vehicle area at the corresponding moment; the influence degree of each historical moment is obtained based on the difference in movement information between each historical moment and the real-time moment, the interference degree of the vehicle area at the historical moment, and the distance between the historical moment and the real-time moment. The movement information of the historical moment is weightedly analyzed using the said influence degree to obtain the predicted movement information of each vehicle area, and the predicted movement information of each vehicle area is adjusted using the movement information of the preceding vehicle at the real-time moment.

[0005] Furthermore, the method for obtaining the matching degree includes: For any point cloud in any initial point cloud data, the point cloud is matched with other point clouds in other initial point cloud data within a preset neighborhood range, and whether the point cloud is a stable feature point is determined based on the matching result; the proportion of stable feature points in the initial point cloud data is used as the matching degree of each initial point cloud data.

[0006] Furthermore, judging whether the point cloud is a stable feature point according to the matching result includes: Any initial point cloud data is used as the target initial point cloud data, any point cloud in the target initial point cloud data is used as the target point cloud, and other point clouds of other initial point cloud data within the preset neighborhood of the target point cloud are used as point clouds to be matched; the cosine similarity of the FPFH vector between the target point cloud and the point cloud to be matched is obtained, and the matching index between the target point cloud and the point cloud to be matched is obtained by combining the Euclidean distance between the target point cloud and the point cloud to be matched, as well as the difference in Doppler velocity; the point cloud to be matched with a matching index greater than a preset threshold is used as the matching point cloud of the target point cloud; if there is a matching point cloud of the target point cloud in all other initial point cloud data, the target point cloud is a stable feature point.

[0007] Furthermore, the method for obtaining the point cloud quality includes: The product of point cloud density and average echo intensity is taken as the point cloud quality of an initial point cloud data.

[0008] Furthermore, the method for obtaining the vehicle area includes: The DBSCAN clustering algorithm is used to cluster the point clouds in the valid point cloud data, and each cluster obtained is a vehicle area.

[0009] Furthermore, the method for obtaining the interference degree includes: The Doppler velocity range difference between all point clouds in each vehicle area is obtained, and the ratio of the Doppler velocity range difference to the number of point clouds in the vehicle area is normalized to obtain the interference degree.

[0010] Furthermore, obtaining the real-time movement information of each vehicle area includes: Normalizing the point cloud density of the valid point cloud data to obtain a point cloud density weight; using the point cloud density weight as the weight of the point cloud velocity; subtracting the point cloud density weight from the positive integer 1 as the weight of the Doppler velocity; performing weighted fusion on the point cloud velocity and the Doppler velocity to obtain the moving velocity of each vehicle area in the moving information at the real time; The point cloud movement direction of the vehicle area between adjacent moments is used as the movement direction in the movement information.

[0011] Furthermore, the method for obtaining the degree of influence includes: For each historical moment, the directional similarity in the movement direction of the vehicle area between the historical moment and the real-time moment is obtained; as well as the movement speed difference between the movement information, the unreliability of the historical moment is obtained based on the movement speed difference, the temporal distance between the historical moment and the real-time moment, and the degree of interference of the vehicle area at the historical moment; the ratio of the directional similarity to the unreliability is used as the degree of influence.

[0012] Furthermore, the method for obtaining the predicted movement information includes: For each historical moment, convert the movement direction of the vehicle area at each historical moment into a unit vector, and use the influence degree to weight the unit vectors of all historical moments to obtain the predicted movement direction in the predicted movement information; The moving speed in the movement information at the real time is adjusted according to the moving speed increment at the real time relative to the previous time to obtain the predicted moving speed in the predicted movement information.

[0013] Furthermore, the use of the real-time movement information of the preceding vehicle to adjust the predicted movement information of each vehicle area includes: If there is no preceding vehicle in the vehicle area at the real time, the predicted movement information of the vehicle area will not be adjusted; If there is a preceding vehicle in the vehicle area at the real time moment and the distance is less than the safety distance specified by the tunnel, the predicted moving speed in the predicted movement information is reduced based on the distance between the vehicle area and the preceding vehicle at the real time moment and the difference between the predicted moving speed of the vehicle area and the moving speed of the preceding vehicle.

[0014] The present invention has the following beneficial effects: The present invention obtains the initial point cloud data collected by different radars at each monitoring position in the tunnel. Taking into account the problem of reduced reference value of the initial point cloud data of a certain radar due to the multipath effect, the present invention uses a matching method to match the initial point cloud data of different radars, and then determines the information weight corresponding to each radar in combination with the point cloud quality, and then determines the effective point cloud data. In order to further achieve accurate speed measurement of the vehicle, after obtaining the effective point cloud data, the point cloud velocity and Doppler velocity are further weighted and fused to obtain the movement information of each vehicle area, reducing the influence of the multipath effect and reflecting real and effective movement information. The present invention further takes into account the need for data prediction in smart transportation, first determines the degree of interference of the vehicle area at each moment, and then combines the information between the historical moment and the real time moment to determine the degree of influence of each historical moment. The degree of influence is used to characterize the reference degree of the information at the historical moment to the real time moment information, and then the predicted movement information can be determined in combination with the degree of influence, and the movement information of the preceding vehicle can be further adjusted to achieve effective speed measurement and accurate prediction of each vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A flow chart of a high-precision radar speed measurement method in a tunnel for smart transportation provided by one embodiment of the present invention; Figure 2 A schematic diagram of tunnel radar deployment provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0017] To further illustrate the technical means and effects of the present invention to achieve the intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effects of a high-precision radar speed measurement method in a tunnel for smart transportation proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable form.

[0018] Unless defined otherwise, 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 invention belongs.

[0019] The following describes in detail a specific scheme of a high-precision radar speed measurement method in a tunnel for smart transportation provided by the present invention with reference to the accompanying drawings.

[0020] See also Figure 1 , which shows a flow chart of a high-precision radar speed measurement method in a tunnel for smart transportation provided by one embodiment of the present invention, the method comprising: Step S1: Multiple radars are installed at monitoring locations in the tunnel, and the initial point cloud data obtained by each radar at the time of acquisition is obtained; based on the degree of matching between the initial point clouds and the point cloud quality of each initial point cloud, the information weight of each initial point cloud data is obtained; and all the initial point cloud data are aligned using the information weight to obtain valid point cloud data.

[0021] In the embodiment of the present invention, in order to reduce the interference of multipath effect, a multi-radar information fusion method is used to improve the information reference. Therefore, multiple radars are installed at each monitoring position in the tunnel, and each radar can obtain the corresponding initial point cloud data at each sampling time. That is, the initial point cloud data is the obtained point cloud set. Figure 2 , which shows a schematic diagram of a tunnel radar deployment provided by an embodiment of the present invention. In this embodiment of the present invention, multiple monitoring positions are set at fixed intervals in the tunnel. Each monitoring position is a monitoring section. Three millimeter-wave radars are deployed on the monitoring section to acquire the initial point cloud. In this embodiment of the present invention, the fixed interval between monitoring sections is set to 500 meters, and millimeter-wave radars are deployed directly above each monitoring section and at a position offset by 15° relative to the directly above. It should be noted that this embodiment of the present invention takes into account the different sensitivities of radars in different frequency bands to interference. Selecting radars in multiple frequency bands for joint monitoring can reduce the risk of detection failure caused by interference in a single frequency band. Radars in the same section correspond to different frequency bands. For example, in this embodiment of the present invention, 4D millimeter-wave radars in the range of 76-79GHz, 24.05-24.25GHz, and 92-96GHz are selected respectively. In this embodiment of the present invention, the data acquisition frequency is set to 30 Hz. After the data acquisition of each radar is completed, the coordinates of the initial point cloud data can be calibrated and mapped to the same coordinate system to facilitate data analysis.

[0022] It should be noted that after the initial point cloud data is obtained in the embodiment of the present invention, in order to avoid interference from non-vehicle information, points with a Doppler velocity of 0 are screened out. The initial point cloud data after preprocessing is a point cloud set that only contains vehicle information point clouds.

[0023] In a tunnel, the detection data of a single radar may be incomplete due to obstruction by large vehicles, or may be distorted by multipath interference caused by factors such as metal interference and tunnel wall reflection. Therefore, the embodiment of the present invention combines the initial point cloud data collected by multiple radars to obtain effective point cloud data with strong reference and less interference.

[0024] Due to the complexity of the tunnel environment, heavy traffic flow, and obstruction by large vehicles, individual radars are affected to varying degrees at different times, making it impossible to calibrate the information validity of each radar using a fixed weight. Therefore, the present invention dynamically calculates the data information weight of each initial point cloud based on the degree of matching between the initial point clouds and the point cloud quality of each initial point cloud. Specifically, for an initial point cloud, the closer it matches other initial point cloud data and the higher its point cloud quality, the less interference the corresponding radar is experiencing, the less data missing there is, the greater the information reference value of the initial point cloud data, and the greater its information weight.

[0025] Once the information weights are obtained, the initial point cloud data collected by all radars at each moment can be registered to obtain the valid point cloud data at that moment. In the embodiment of the present invention, the ICP algorithm can be used to register the initial point cloud data collected by radars within the same monitoring cross section using the information weights. The specific details are well known to those skilled in the art and will not be detailed here.

[0026] Preferably, in an embodiment of the present invention, the method for obtaining the matching degree includes: For any point cloud in any initial point cloud data, the point cloud is matched with other point clouds in other initial point cloud data within a preset neighborhood. Based on the matching results, the point cloud is judged to be a stable feature point. For an initial point cloud, the more stable feature points there are, the less affected the initial point cloud data is by multipath effects and the less severe the effects of data loss. Therefore, the proportion of stable feature points in the initial point cloud data can be used as the matching degree of each initial point cloud data.

[0027] Furthermore, in an embodiment of the present invention, during the matching process between point clouds, the more similar the Doppler velocities between the two point clouds and the closer the spatial distance, the higher the matching degree between the two points. Therefore, judging whether a point cloud is a stable feature point based on the matching result includes: Any initial point cloud data is used as the target initial point cloud data, any point cloud in the target initial point cloud data is used as the target point cloud, and other point clouds of other initial point cloud data within the preset neighborhood of the target point cloud are used as point clouds to be matched.

[0028] The cosine similarity of the Fast Point Feature Histograms (FPFH) vectors between the target point cloud and the point cloud to be matched is obtained. This is combined with the Euclidean distance and Doppler velocity difference between the two points to determine the matching index. Specifically, the smaller the Euclidean distance, the smaller the Doppler velocity difference, and the greater the cosine similarity, the closer the match between the two point clouds, and the greater the matching index.

[0029] The matching point cloud with a matching index greater than a preset threshold is used as the matching point cloud of the target point cloud; if the matching point cloud of the target point cloud exists in all other initial point cloud data, the target point cloud is a stable feature point. Similarly, all stable feature points can be determined in each initial point cloud.

[0030] In this embodiment of the present invention, the radius of the preset neighborhood range is set to 0.3 meters in the real coordinate system. That is, the neighborhood range is a range with a radius of 0.3 meters centered on the target point cloud. After the matching index is normalized, the threshold is set to 0.7. The normalization method uses range normalization to limit the value to between 0 and 1.

[0031] As an example, in an embodiment of the present invention, the method for obtaining the matching index is as follows: the absolute value of the difference in Doppler velocity between point clouds is used as the difference in Doppler velocity, and the Euclidean distance and the difference in Doppler velocity are multiplied together to form the difference index. The matching index can be obtained by performing negative correlation mapping on the difference index and then multiplying it with the cosine similarity. It should be noted that the negative correlation mapping method is a basic mathematical method well known to those skilled in the art, and specifically, methods such as the inverse form and function mapping can be used. In the embodiment of the present invention, the inverse form is used. At the same time, in order to avoid the denominator being 0, the inverse of the difference index after adding 0.01 is used as the negative correlation mapping result.

[0032] Preferably, in the embodiment of the present invention, the product of the point cloud density and the average echo intensity is used as the point cloud quality of initial point cloud data.

[0033] In the embodiment of the present invention, after both the point cloud quality and the matching degree are quantified, the product of the two is used as the information weight.

[0034] Step S2: Obtain the vehicle area in the valid point cloud data, obtain the point cloud velocity of each vehicle area at each moment through the movement changes of the vehicle area at adjacent moments, use the point cloud density to perform weighted analysis, and perform weighted fusion of the point cloud velocity and Doppler velocity to obtain the movement information of each vehicle area at real time.

[0035] The effective point cloud data obtained in step S1 is point cloud data with the multipath effect reduced. However, in order to further obtain effective vehicle motion information, the embodiment of the present invention does not directly use the Doppler velocity reflected in the effective point cloud data as the true velocity information. The Doppler velocity refers to the radial velocity of the target object relative to the observation device measured using the Doppler effect. Its core principle is based on the frequency change caused by the relative motion between the wave source and the receiver. Therefore, it is necessary to further combine the motion of the vehicle point cloud between adjacent moments and fuse and verify the two types of motion information to obtain effective movement information.

[0036] First, the present embodiment obtains vehicle regions from valid point cloud data. Because vehicles have fixed shapes, tunnels typically only move vehicles, with no other moving objects such as pedestrians. Therefore, the point cloud obtained after preprocessing consists solely of vehicle point clouds. Therefore, vehicle regions can be effectively filtered from the valid point cloud data based on the distance between point clouds. By matching vehicle regions at different times, the location and point cloud information of each vehicle region at different times can be determined. The matching process is similar to the point cloud data matching method in step S1: For a target vehicle region at real time, the target vehicle region is matched with each vehicle region at a historical time. The vehicle region at the historical time to be matched is the vehicle region to be matched. A matching index is calculated between the target point cloud in the target vehicle region and the point cloud to be matched in the vehicle region to be matched. If a matching point cloud has a matching index greater than a preset threshold, the match is considered successful. The target point cloud is modified, and the proportion of successfully matched point clouds in the target vehicle region is used as the matching degree between the target vehicle region and the vehicle region to be matched. Vehicle regions with matching degrees greater than 0.8 at historical times are selected as the same vehicle region as the target vehicle region.

[0037] It should be noted that, considering that the radar has a monitoring range, when matching vehicle areas between different times, the historical time range is set to the ratio of the spacing of the monitoring sections to the historical maximum vehicle speed.

[0038] At this point, the point cloud velocity of each vehicle zone at each moment can be derived from the changes in the movement of the vehicle zone at adjacent moments. Specifically, for the real-time moment, the center of mass of the vehicle zone is selected as the reference point, and the distance traveled between the real-time moment and the previous moment is calculated. The ratio of this distance to the time interval is the point cloud velocity. Point cloud velocity is velocity information determined by point cloud motion. For a point cloud, a greater point cloud density indicates a greater reference value for the point cloud velocity; conversely, a smaller point cloud density indicates a lesser reference value for the point cloud velocity, requiring more reference Doppler velocity. Therefore, point cloud density can be used for weighted analysis, combining the point cloud velocity and Doppler velocity for a weighted fusion to obtain the real-time movement information for each vehicle zone.

[0039] It should be noted that, because the effective point cloud data is the registered point cloud data, the Doppler velocity of the vehicle area is the average of the Doppler velocity components of all points in the area in the common coordinate system.

[0040] Preferably, in an embodiment of the present invention, the method for obtaining a vehicle area includes: The valid point cloud data is clustered using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. Each resulting cluster represents a vehicle region. In this embodiment of the present invention, the neighborhood radius in the DBSCAN clustering algorithm is set to 0.3 meters in the real-world coordinate system, and the minimum number of samples is set to 5.

[0041] Preferably, in an embodiment of the present invention, obtaining the real-time movement information of each vehicle area includes: Normalize the point cloud density of the valid point cloud data to obtain a point cloud density weight. This point cloud density weight is used as the point cloud velocity weight. Subtracting the point cloud density weight from the positive integer 1 is used as the Doppler velocity weight. The point cloud velocity and Doppler velocity are weighted and fused to obtain the real-time velocity of each vehicle region in the movement information. The point cloud movement direction of the vehicle region between adjacent moments is used as the movement direction in the movement information.

[0042] It should be noted that the normalization algorithms in the embodiments of the present invention may all adopt range normalization, that is, normalization is performed using the maximum and minimum values in their respective dimensions, and details will not be repeated here.

[0043] Step S3: For each vehicle area, the interference degree of each vehicle area is obtained according to the Doppler velocity stability of the vehicle area at the corresponding moment; the influence degree of each historical moment is obtained according to the difference in movement information between each historical moment and the real-time moment, the interference degree of the vehicle area at the historical moment, and the distance between the historical moment and the real-time moment, and the movement information of the historical moment is weightedly analyzed using the influence degree to obtain the predicted movement information of each vehicle area, and the movement information of the preceding vehicle at the real-time moment is used to adjust the predicted movement information of each vehicle area.

[0044] Smart transportation requires not only effective collection of vehicle movement information within tunnels but also effective prediction of future movement trends to avoid communication latency and enable terminals to respond promptly to management issues. Information prediction can be combined with historical information for statistical integration to analyze future movement trends. However, the impact of historical moments on real-time movement information varies, influenced by factors such as point cloud signal quality, the difference in movement information between two moments, and the distance between moments. Therefore, the present invention first determines the degree of interference for each vehicle zone based on the Doppler velocity stability of the vehicle zone at the corresponding moment. The impact level for each historical moment is then determined based on the difference in movement information between each historical moment and the real-time moment, the degree of interference for the vehicle zone at the historical moment, and the distance between the historical moment and the real-time moment. The impact level represents the degree of reference of the historical moment's movement information relative to the current moment. Therefore, using the impact level to weight the movement information at the historical moment, predicted movement information for each vehicle zone is obtained. Furthermore, considering that the distance between the vehicle area and the preceding vehicle at the real time will affect the vehicle's movement speed at the future time, the predicted movement information of each vehicle area can be adjusted in combination with the preceding vehicle's movement information at the real time, thereby obtaining accurate and effective prediction information.

[0045] Preferably, the method for obtaining the interference degree in the embodiment of the present invention includes: The Doppler velocity range between all point clouds in each vehicle region is obtained, and the ratio of this Doppler velocity range to the number of point clouds in the vehicle region is normalized to obtain the degree of interference. Specifically, at a given moment, the more uneven the Doppler velocity distribution of the point cloud data in the vehicle region, the greater the volatility of the point cloud detection results for the same vehicle, and thus the greater the degree of interference. The smaller the number of point clouds, the lower the point cloud quality and the greater the degree of interference.

[0046] Preferably, in an embodiment of the present invention, the method for obtaining the degree of influence includes: For each historical moment, the directional similarity of the vehicle area's movement direction between the historical moment and the real-time moment is determined; as well as the difference in movement speed between the movement information. The unreliability of the historical moment is determined based on the movement speed difference, the temporal distance between the historical moment and the real-time moment, and the degree of interference with the vehicle area at the historical moment. The greater the temporal distance, the greater the degree of interference, and the greater the difference in movement speed, the less relevant the information at that historical moment is, and the greater the unreliability. The ratio of the directional similarity to the unreliability is used as the degree of influence.

[0047] In this embodiment of the present invention, similar to the method for obtaining the matching index, the difference in movement speed, the temporal distance between the historical moment and the real time, and the degree of interference in the vehicle area at the historical moment are multiplied together and then negatively correlated. This is then multiplied by the directional similarity to obtain the degree of influence. The specific negative correlation mapping method is not further described.

[0048] It should be noted that directional similarity can be obtained using the cosine similarity method. The movement directions of the vehicle area at the historical moment and the real time are converted into unit vectors, and the cosine similarity between the unit vectors is calculated to obtain directional similarity. The speed difference is the absolute value of the difference in movement speed.

[0049] Preferably, in an embodiment of the present invention, the method for obtaining predicted movement information includes: For each historical moment, the movement direction of the vehicle area at each historical moment is converted to a unit vector. The unit vectors of all historical moments are weighted and fused using the influence degree to obtain the predicted movement direction in the predicted movement information. The direction of the weighted fused vector is used as the predicted movement direction.

[0050] The speed in the real-time movement information is adjusted based on the speed increment relative to the previous moment to obtain the predicted speed in the predicted movement information. The predicted speed is the sum of the real-time speed and the speed increment. The speed increment is the difference between the real-time speed and the previous moment's speed.

[0051] Preferably, in an embodiment of the present invention, using the real-time movement information of the preceding vehicle to adjust the predicted movement information of each vehicle area includes: If there is no preceding vehicle in the vehicle area at the real time, the predicted movement information of the vehicle area will not be adjusted; If there is a preceding vehicle in the vehicle zone at the moment, and the distance is less than the tunnel's specified safety distance, the predicted speed in the predicted movement information is reduced based on the distance between the vehicle zone and the preceding vehicle at the moment, as well as the difference in speed between the vehicle zone and the preceding vehicle in their movement information. In this embodiment of the present invention, the distance between the vehicle zone and the preceding vehicle is subtracted from the tunnel's specified safety distance to obtain the excess distance. The difference between the predicted speed of the vehicle zone and the preceding vehicle's speed is obtained, multiplied by the excess distance, and then normalized to obtain the degree of danger. The difference between the positive integer 1 and the degree of danger is used as an adjustment coefficient, and the product of the adjustment coefficient and the predicted speed is used as the adjusted predicted speed.

[0052] In summary, the embodiments of the present invention utilize a matching method to match the initial point cloud data from different radars, determine the information weight corresponding to each radar based on the point cloud quality, and thus determine the valid point cloud data. The point cloud velocity and Doppler velocity are further weighted and fused to obtain the movement information of each vehicle area. The degree of interference in the vehicle area at each moment is determined, and the degree of influence of each historical moment can be determined by combining the information between the historical moment and the real-time moment. The predicted movement information is determined based on the degree of influence, and the movement information of the preceding vehicle is used for adjustment. The present invention eliminates the influence of multipath effects in tunnels, achieving effective speed measurement and accurate prediction for each vehicle.

[0053] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0054] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A high-precision radar speed measurement method in a tunnel for smart transportation, characterized in that: The method comprises: Multiple radars are installed at monitoring locations within the tunnel to obtain initial point cloud data acquired by each radar at the time of acquisition. Based on the degree of matching between the initial point clouds and the point cloud quality of each initial point cloud, an information weight for each initial point cloud data is obtained. All initial point cloud data are registered using the information weight to obtain valid point cloud data. Obtain the vehicle area in the valid point cloud data, obtain the point cloud velocity of each vehicle area at each moment through the movement changes of the vehicle area at adjacent moments, perform weight analysis using the point cloud density, and perform weighted fusion of the point cloud velocity and Doppler velocity to obtain the movement information of each vehicle area at real time; For each vehicle area, the interference degree of each vehicle area is obtained based on the Doppler velocity stability of the vehicle area at the corresponding moment; the influence degree of each historical moment is obtained based on the difference in movement information between each historical moment and the real-time moment, the interference degree of the vehicle area at the historical moment, and the distance between the historical moment and the real-time moment. The movement information of the historical moment is weightedly analyzed using the said influence degree to obtain the predicted movement information of each vehicle area, and the predicted movement information of each vehicle area is adjusted using the movement information of the preceding vehicle at the real-time moment.

2. The high-precision radar speed measurement method in a tunnel for smart transportation according to claim 1 is characterized in that: The method for obtaining the matching degree includes: For any point cloud in any initial point cloud data, the point cloud is matched with other point clouds in other initial point cloud data within a preset neighborhood range, and whether the point cloud is a stable feature point is determined based on the matching result; the proportion of stable feature points in the initial point cloud data is used as the matching degree of each initial point cloud data.

3. The high-precision radar speed measurement method in a tunnel for smart transportation according to claim 2 is characterized in that: The step of determining whether the point cloud is a stable feature point according to the matching result includes: Any initial point cloud data is used as the target initial point cloud data, any point cloud in the target initial point cloud data is used as the target point cloud, and other point clouds of other initial point cloud data within the preset neighborhood of the target point cloud are used as point clouds to be matched; the cosine similarity of the FPFH vector between the target point cloud and the point cloud to be matched is obtained, and the matching index between the target point cloud and the point cloud to be matched is obtained by combining the Euclidean distance between the target point cloud and the point cloud to be matched, as well as the difference in Doppler velocity; the point cloud to be matched with a matching index greater than a preset threshold is used as the matching point cloud of the target point cloud; if there is a matching point cloud of the target point cloud in all other initial point cloud data, the target point cloud is a stable feature point.

4. The high-precision radar speed measurement method in a tunnel for smart transportation according to claim 1 is characterized in that: The method for obtaining the point cloud quality includes: The product of point cloud density and average echo intensity is taken as the point cloud quality of an initial point cloud data.

5. The high-precision radar speed measurement method in a tunnel for smart transportation according to claim 1 is characterized in that: The method for obtaining the vehicle area includes: The DBSCAN clustering algorithm is used to cluster the point clouds in the valid point cloud data, and each cluster obtained is a vehicle area.

6. The high-precision radar speed measurement method in a tunnel for smart transportation according to claim 1 is characterized in that: The method for obtaining the interference degree includes: The Doppler velocity range difference between all point clouds in each vehicle area is obtained, and the ratio of the Doppler velocity range difference to the number of point clouds in the vehicle area is normalized to obtain the interference degree.

7. The high-precision radar speed measurement method in a tunnel for smart transportation according to claim 1, characterized in that: The obtaining of the real-time movement information of each vehicle area includes: Normalizing the point cloud density of the valid point cloud data to obtain a point cloud density weight; using the point cloud density weight as the weight of the point cloud velocity; subtracting the point cloud density weight from the positive integer 1 as the weight of the Doppler velocity; performing weighted fusion on the point cloud velocity and the Doppler velocity to obtain the moving velocity of each vehicle area in the moving information at the real time; The point cloud movement direction of the vehicle area between adjacent moments is used as the movement direction in the movement information.

8. The high-precision radar speed measurement method in a tunnel for smart transportation according to claim 1 is characterized in that: The method for obtaining the impact degree includes: For each historical moment, the directional similarity in the movement direction of the vehicle area between the historical moment and the real-time moment is obtained; as well as the movement speed difference between the movement information, the unreliability of the historical moment is obtained based on the movement speed difference, the temporal distance between the historical moment and the real-time moment, and the degree of interference of the vehicle area at the historical moment; the ratio of the directional similarity to the unreliability is used as the degree of influence.

9. The high-precision radar speed measurement method in a tunnel for smart transportation according to claim 1, characterized in that: The method for obtaining predicted movement information includes: For each historical moment, convert the movement direction of the vehicle area at each historical moment into a unit vector, and use the influence degree to weight the unit vectors of all historical moments to obtain the predicted movement direction in the predicted movement information; The moving speed in the movement information at the real time is adjusted according to the moving speed increment at the real time relative to the previous time to obtain the predicted moving speed in the predicted movement information.

10. The high-precision radar speed measurement method in a tunnel for smart transportation according to claim 1, characterized in that: The method of adjusting the predicted movement information of each vehicle area using the real-time movement information of the preceding vehicle includes: If there is no preceding vehicle in the vehicle area at the real time, the predicted movement information of the vehicle area will not be adjusted; If there is a preceding vehicle in the vehicle area at the real time moment and the distance is less than the safety distance specified by the tunnel, the predicted moving speed in the predicted movement information is reduced based on the distance between the vehicle area and the preceding vehicle at the real time moment and the difference between the predicted moving speed of the vehicle area and the moving speed of the preceding vehicle.

Images