Vehicle-mounted millimeter wave radar self-calibration method based on target tracking

Abstract

The invention discloses a vehicle-mounted millimeter-wave radar self-calibration method based on target tracking, and the main conception is that a calibration process is automatically activated according to the driving speed of a vehicle and the tracking effect of a front vehicle, and the trajectory point cloud direction of a single lane or multiple lanes is fitted based on the linear features formed by the trajectory point cloud of the front tracked vehicle. Therefore, the millimeter-wave radar deviation value of the vehicle is judged. According to the method, the characteristic of the millimeter wave radar good at dynamic target detection is effectively utilized, and radar calibration is realized by tracking the vehicle running in front, so that the problem of static target detection is effectively avoided. On one hand, the method does not need to depend on any fixed facility, enables the calibration process not to be limited by a specific environment or facility, and can smoothly carry out the calibration task in various complex and changeable roads and traffic environments. And through flexible and abundant front vehicle tracking scenes, the calibration process is more visual and easy to trigger, the computing power is saved, and the operation convenience and the intelligent level are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving sensor calibration, and in particular to a vehicle-mounted millimeter-wave radar self-calibration method based on target tracking. Background Art

[0002] The three major coordinate systems in autonomous driving include: the world coordinate system, the vehicle coordinate system, and the coordinate systems of various sensors, such as the coordinate system of 4D millimeter-wave radar (4D millimeter-wave radar is an advanced form of millimeter-wave radar, which can not only measure the distance, direction and speed of the target, but also measure the height information of the target, that is, the fourth dimension).

[0003] The target information detected by the radar is in the radar coordinate system, while the various intelligent driving functions in autonomous driving are completed in the vehicle coordinate system. These are two different coordinate systems. The conversion parameters between the two coordinate systems must be determined through radar calibration in order to map the target information detected by the radar to the vehicle coordinate system. At the same time, radar calibration is also an important parameter for multi-sensor spatiotemporal fusion, which directly affects the output effect of the multi-source sensor perception fusion algorithm.

[0004] From the perspective of calibration scenarios, there are currently three calibration methods for millimeter-wave radar:

[0005] (1) Sensor production line calibration (EOL calibration): refers to the process of evaluating whether the installation tolerance meets the requirements before the vehicle rolls off the production line, measuring and storing the actual installation position and angle of the sensor. Generally speaking, sensor production line calibration adopts a static calibration solution and is calibrated in a static calibration room.

[0006] (2) Sensor 4S shop calibration: refers to the process of evaluating whether the installation tolerance meets the requirements after the vehicle's ADS sensor is replaced or adjusted at the 4S shop, and measuring and storing the actual installation position and angle of the sensor. Generally speaking, sensor 4S shop calibration adopts a combination of static target + open road dynamic calibration.

[0007] (3) Online automatic calibration: refers to the process of monitoring the sensor angle in the background while the vehicle is in use. Through online monitoring, the owner will be able to obtain timely information from the HMI whether the vehicle needs to be recalibrated / returned to the store for repair. The online automatic calibration function can mainly solve the problem of slight changes in the installation angle of the radar caused by temperature changes, bracket aging, bracket deformation caused by collision, etc. during the use of the vehicle.

[0008] Based on this background technology, the self-calibration scheme for vehicle-mounted radar proposed in the industry was analyzed and experimentally practiced. It is believed that this technology still has many shortcomings:

[0009] 1. When the radar detects a target, the returned speed is the radial speed from the target to the vehicle. Therefore, when there is a pedestrian or object walking sideways in front, the radar will identify the second speed of the target as 0, thereby falsely detecting more stationary targets and causing errors.

[0010] 2. In a high-density traffic environment, there are many other vehicles and pedestrians, which will interfere with the radar's measurement and make it impossible to detect stationary targets.

[0011] 3. Guardrails are usually used as road boundary markers. Their position and shape may be affected by factors such as road design, construction and maintenance. Therefore, they are not suitable as the main reference object for millimeter-wave radar calibration.

[0012] 4. In some scenarios, such as urban roads or complex terrain, guardrails may not exist or cannot serve as effective calibration reference objects, which limits the application scope of millimeter-wave radar self-calibration. Summary of the invention

[0013] In view of the above, the present invention aims to provide a vehicle-mounted millimeter-wave radar self-calibration method based on target tracking, which realizes automatic calibration of the radar by tracking the vehicle ahead on the road.

[0014] The technical solution adopted by the present invention is as follows:

[0015] The present invention provides a vehicle-mounted millimeter-wave radar self-calibration method based on target tracking, which includes:

[0016] During the driving process, the radar self-calibration mode is automatically activated according to the vehicle speed and the movement information of the target vehicle in front of the vehicle.

[0017] In the self-calibration mode, the point cloud data of the moving target vehicle is continuously acquired;

[0018] Get the straight line features of point cloud data;

[0019] Fit the final overall trajectory direction based on the straight line features;

[0020] According to the overall trajectory direction, a deviation value of the millimeter-wave radar of the vehicle is determined.

[0021] In at least one possible implementation, the self-calibration method further includes processing the acquired point cloud data as follows:

[0022] Clustering the point cloud data;

[0023] Based on the clustering results, the point cloud data of the same target vehicle are associated and matched to obtain the driving trajectory corresponding to the current target vehicle;

[0024] Multi-frame data fusion is performed in the above manner to obtain the complete trajectory point cloud of the target vehicle.

[0025] In at least one possible implementation manner, obtaining straight line features of point cloud data includes:

[0026] The trajectory point cloud is filtered according to established rules to obtain point cloud data of several single frames and their numbers, wherein the established rules include: according to the trajectory point cloud of the target vehicle, a first minimum circumscribed rectangle is obtained, and a first central axis angle of the first minimum circumscribed rectangle is output; the trajectory point clouds of multiple consecutive frames are divided into a group frame by frame, and the second minimum circumscribed rectangle of each group as a whole is obtained, and the second central axis angle of the second minimum circumscribed rectangle is output; if the angle deviation between the first central axis angle and the second central axis angle exceeds a preset angle threshold, the trajectory point cloud of the current group is discarded; the above steps are repeated until all the trajectory point clouds of the target vehicle are traversed, and the point cloud data of several single frames retained after the screening is satisfied and their numbers are output.

[0027] The selected trajectory point cloud data are grouped according to the continuity of the numbering sequence;

[0028] The third minimum circumscribed rectangle of each group is calculated, and the third central axis of the third minimum circumscribed rectangle is determined.

[0029] In at least one possible implementation, if the amount of data retained after screening of the current trajectory point cloud is less than a predetermined standard, another front vehicle is reselected as the target vehicle, and tracking and point cloud data acquisition are performed.

[0030] In at least one possible implementation, the fitting of the overall trajectory direction of the target vehicle includes:

[0031] Calculating a translation vector of the target vehicle according to the spacings between the plurality of third central axes;

[0032] A straight line is refitted based on the translation vector as the overall trajectory direction of the target vehicle.

[0033] In at least one possible implementation method, the method for processing point cloud data also includes: before performing trajectory point cloud computing, selecting a target point cloud related to the movement of the target vehicle from the point cloud data based on the speed information of the point cloud, and removing outliers in the point cloud data.

[0034] Compared with the prior art, the main design concept of the present invention is to automatically activate the calibration process according to the driving speed of the vehicle and the tracking effect of the vehicle in front, and fit the straight line characteristics formed by the trajectory point cloud of the vehicle being tracked in front to the direction of the trajectory point cloud of a single lane or multiple lanes, so as to judge the deviation value of the millimeter-wave radar of the vehicle. The present invention effectively utilizes the characteristics of millimeter-wave radar that is good at detecting dynamic targets, and realizes radar calibration by tracking the vehicle in front, thereby effectively avoiding the problem of static target detection.

[0035] Compared with the traditional method, the advantages of the present invention are as follows:

[0036] Most traditional methods rely on fixed road facilities, such as road guardrails, which greatly limits the selection and application scope of calibration scenarios; the present invention does not need to rely on any additional fixed facilities, so that the calibration process is no longer limited to a specific environment or facility, so that calibration tasks can be carried out smoothly in various complex and changeable road and traffic environments, thereby demonstrating the extremely high adaptability and flexibility of the present invention.

[0037] In addition, the traditional radar self-calibration algorithm needs to detect stationary targets outside the road, which increases the complexity of the algorithm to a certain extent; while the present invention makes the calibration process more intuitive, easy to trigger, and saves computing power through flexible and rich front vehicle tracking scenarios, and the calibration trigger mechanism does not require manual intervention or additional activation, greatly improving the convenience and intelligence of operation. In addition, the solution of the present invention is easy to deploy in the existing intelligent driving system, thereby reducing the difficulty and cost of system development and integration. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described below with reference to the accompanying drawings, in which:

[0039] Figure 1 A schematic diagram of a vehicle-mounted millimeter-wave radar self-calibration method based on target tracking provided by an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of fusing multi-frame point clouds provided by an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of intercepting a trajectory point cloud provided by an embodiment of the present invention;

[0042] Figure 4 A schematic diagram of meeting the screening conditions and obtaining the central axis provided by an embodiment of the present invention;

[0043] Figure 5 A schematic diagram of a final fitting straight line provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] Embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.

[0045] The present invention proposes an embodiment of a vehicle-mounted millimeter-wave radar self-calibration method based on target tracking. Specifically, Figure 1 shown, including:

[0046] Step S1: during the driving process of the vehicle, according to the vehicle speed and the motion information of the target vehicle in front of the vehicle being tracked, the radar self-calibration mode is automatically activated;

[0047] For example, the millimeter-wave radar self-calibration mode can be started when the vehicle speed is stable within a preset range and the speed and position relationship between the target vehicle in front of the vehicle and the vehicle is within a predetermined controllable range for a certain period of time. The target vehicle is not limited to being directly in front of the vehicle in the same lane, but can also refer to the vehicle in front of the side; and the target vehicle is preferably also in a relatively stable driving state, rather than speeding up and slowing down or appearing and disappearing in the field of vision.

[0048] In actual operation, users can pre-start the automatic calibration process by opening "Assisted Driving-ADS Sensor Calibration-Radar Calibration" through the vehicle HMI portal. After that, the radar will continue to run the self-calibration standby state. Once the vehicle's driving speed and the tracking effect of the vehicle in front meet the calibration conditions, the calibration process can be carried out in time.

[0049] Step S2: in the self-calibration mode, continuously acquiring point cloud data of the target vehicle in motion;

[0050] Furthermore, the acquired point cloud data can be processed as follows:

[0051] Step S21, clustering the point cloud data;

[0052] For example, a density-based clustering method such as DBSCAN can be used. Compared with other clustering algorithms, it does not require predetermining the number of clusters to be clustered, and is more suitable for clustering an unknown number of target point clouds that the present invention is concerned with, thereby achieving target type conversion. Therefore, in some embodiments of the present invention, a multi-target tracking algorithm based on intra-frame DBSCAN clustering is preferably used.

[0053] Step S22: Based on the clustering result, the point cloud data of the same target vehicle are associated and matched to obtain the driving trajectory corresponding to the current target vehicle;

[0054] For example, the Hungarian algorithm can be used to associate and match the point clouds of the same target in consecutive frames to construct its approximate motion trajectory in a single lane or different lanes.

[0055] Step S23: perform multi-frame data fusion in the above manner to obtain a complete trajectory point cloud of the target vehicle.

[0056] After the above processing method of the motion trajectory point cloud data, the multi-frame radar point cloud data are fused, such as Figure 2 As shown in FIG. 1 , a relatively complete expression of the target vehicle trajectory point cloud is obtained by merging.

[0057] In addition, the processing method for point cloud data may also include, when performing trajectory point cloud computing, first selecting a target point cloud related to the movement of the target vehicle from the point cloud data based on the speed information of the point cloud, and using an outlier filter to remove outliers in the point cloud data, thereby achieving preprocessing of the point cloud data.

[0058] Step S3, obtaining straight line features of point cloud data;

[0059] Specifically, the straight line features of the point cloud data are obtained as follows:

[0060] Step S31, filtering the trajectory point cloud according to a predetermined rule to obtain point cloud data of a plurality of single frames and their serial numbers, wherein the predetermined filtering rule is as follows:

[0061] According to the trajectory point cloud of the target vehicle, a first minimum circumscribed rectangle Rectangle_1 is obtained, and a first central axis angle Angle_1 of the first minimum circumscribed rectangle is output;

[0062] like Figure 3 In other words, the trajectory point clouds of three consecutive frames (not limited) are grouped one by one, the second minimum enclosing rectangle Rectangle_2 of the group as a whole is obtained, and the second central axis angle Angle_2 of the second minimum enclosing rectangle is output;

[0063] If the angle deviation between Angle_2 and Angle_1 exceeds the preset angle threshold (such as 10°), the trajectory point cloud of the current group is discarded;

[0064] Repeat steps ~ , until all the trajectory point clouds of the target vehicle are traversed, and the output satisfies Point cloud data and their serial numbers of several single frames under this condition. Figure 3As shown, after the above screening process, it can be determined that the point cloud of the target vehicle in the lane merging process does not meet the conditions, so it is not used as the data basis for subsequent processing;

[0065] It can also be supplemented by this concept that if the data that does not meet the conditions after screening of the current trajectory point cloud is greater than the established standard, then return to the first step, select other front vehicles as target vehicles and track them and obtain point cloud data.

[0066] Step S32, grouping the selected trajectory point cloud data according to the continuity of the numbering sequence (this method can be more suitable for the real driving scene and realize the accurate division of the tracked target vehicle in different lanes, that is, more suitable for the real driving tracking condition);

[0067] Step S33: Calculate the third minimum circumscribed rectangle of each group, and determine the third central axis of the third minimum circumscribed rectangle (e.g. Figure 4 Schematic diagram).

[0068] Step S4, fitting the final overall trajectory direction based on the straight line features;

[0069] In combination with the above specific embodiments, the overall trajectory direction of the target vehicle can be fitted as follows: according to the spacing between the third central axes, the translation vector of the target vehicle is calculated; according to the translation vector, a straight line is refitted as the overall trajectory direction of the target vehicle, thereby converting all the trajectory point clouds of the target vehicle to the same straight line direction, such as Figure 5 It shows the refitted straight line based on two driving trajectory lines (if the target vehicle point cloud is only in a single lane, it can be processed in the same way).

[0070] Step S5: Determine the deviation value of the millimeter-wave radar of the vehicle according to the overall trajectory direction.

[0071] That is, by combining the overall trajectory direction of the refitting with the coordinate system of the vehicle-mounted millimeter-wave radar, the radar yaw angle θ (the inclination of the final fitting straight line) is calculated, and the radar calibration parameters are modified accordingly. It can also be added here that the self-calibration method also includes: judging whether the calibration parameters to be modified based on the deviation value are within the established installation permission range, and if so, avoiding errors caused by environmental changes and outputting online calibration values; if not, reporting to the system and giving an alarm, indicating that the deviation is too large to modify the calibration parameters.

[0072] In summary, the main design concept of the present invention is to automatically activate the calibration process according to the driving speed of the vehicle and the tracking effect of the vehicle in front, and fit the straight line characteristics formed by the trajectory point cloud of the vehicle being tracked in front in the direction of the trajectory point cloud of a single lane or multiple lanes to determine the deviation value of the millimeter-wave radar of the vehicle. The present invention effectively utilizes the characteristics of millimeter-wave radar that is good at detecting dynamic targets, and realizes radar calibration by tracking the vehicle in front, thereby effectively avoiding the problem of static target detection.

[0073] Compared with the traditional method, the advantages of the present invention are as follows:

[0074] There is no need to rely on any additional fixed facilities, so that the calibration process is no longer limited to a specific environment or facility, so that the calibration task can be smoothly carried out in various complex and changeable road and traffic environments, thereby demonstrating the extremely high adaptability and flexibility of the solution of the present invention.

[0075] In addition, through flexible and rich front vehicle tracking scenarios, the calibration process is more intuitive, easy to trigger, and saves computing power. The calibration trigger mechanism does not require manual intervention or additional activation, which greatly improves the convenience and intelligence of operation. In addition, the solution of the present invention is easy to deploy in the existing intelligent driving system, thereby reducing the difficulty and cost of system development and integration.

[0076] If the expressions expressing orientation are mentioned in the embodiments of the present invention, they are relative concepts based on the embodiments. In addition, "at least one" means one or more, and "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, wherein a, b, c can be single or multiple.

[0077] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings, but the above is only a preferred embodiment of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched into a variety of equivalent schemes by those skilled in the art without departing from or changing the design ideas and technical effects of the present invention; therefore, the present invention is not limited to the scope of implementation shown in the drawings, and all changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the specification and drawings, should be within the protection scope of the present invention.

Claims

1. A vehicle-mounted millimeter-wave radar self-calibration method based on target tracking, characterized in that: include: During the driving process, the radar self-calibration mode is automatically activated according to the vehicle speed and the movement information of the target vehicle in front of the vehicle. In the self-calibration mode, the point cloud data of the moving target vehicle is continuously acquired; Get the straight line features of point cloud data; Fit the final overall trajectory direction based on the straight line features; According to the overall trajectory direction, a deviation value of the millimeter-wave radar of the vehicle is determined.

2. The vehicle-mounted millimeter-wave radar self-calibration method based on target tracking according to claim 1 is characterized in that: The self-calibration method further includes performing the following processing on the acquired point cloud data: Clustering the point cloud data; Based on the clustering results, the point cloud data of the same target vehicle are associated and matched to obtain the driving trajectory corresponding to the current target vehicle; Multi-frame data fusion is performed in the above manner to obtain the complete trajectory point cloud of the target vehicle.

3. The vehicle-mounted millimeter-wave radar self-calibration method based on target tracking according to claim 2 is characterized in that: The straight line features of the point cloud data are obtained as follows: The trajectory point cloud is screened according to the established rules to obtain point cloud data of several single frames and their serial numbers, wherein the established rules include: according to the trajectory point cloud of the target vehicle, a first minimum circumscribed rectangle is obtained, and a first central axis angle of the first minimum circumscribed rectangle is output; the trajectory point clouds of multiple consecutive frames are grouped together frame by frame, a second minimum circumscribed rectangle of each group as a whole is obtained, and a second central axis angle of the second minimum circumscribed rectangle is output; if the angle deviation between the first central axis angle and the second central axis angle exceeds a preset angle threshold, the trajectory point cloud of the current group is discarded; the above steps are repeated until all the trajectory point clouds of the target vehicle are traversed, and the point cloud data of several single frames retained after the screening is satisfied and their serial numbers are output; The selected trajectory point cloud data are grouped according to the continuity of the numbering sequence; The third minimum circumscribed rectangle of each group is calculated, and the third central axis of the third minimum circumscribed rectangle is determined.

4. The vehicle-mounted millimeter-wave radar self-calibration method based on target tracking according to claim 3 is characterized in that: If the amount of data retained after screening of the current trajectory point cloud is less than the established standard, another vehicle ahead is reselected as the target vehicle and tracked and point cloud data is acquired.

5. The vehicle-mounted millimeter-wave radar self-calibration method based on target tracking according to claim 3 is characterized in that: The overall trajectory direction of the fitted target vehicle includes: Calculating a translation vector of the target vehicle according to the spacings between the plurality of third central axes; A straight line is refitted based on the translation vector as the overall trajectory direction of the target vehicle.

6. The vehicle-mounted millimeter-wave radar self-calibration method based on target tracking according to claim 2 is characterized in that: The method for processing the point cloud data also includes: before performing trajectory point cloud computing, selecting a target point cloud related to the movement of the target vehicle from the point cloud data according to the speed information of the point cloud, and removing outliers in the point cloud data.

Images