Phase synchronization method for multi-lidar front fusion based on vehicle-road collaboration
Through vehicle-road collaboration technology, the point cloud data of lidar on multiple roadside units is gathered, position labeling and time deviation calculation are performed, and the phase of lidar is adjusted, which solves the problem of insufficient perception of single lidar in complex intersection scenarios, and realizes phase synchronization and data splicing of multi-lidars, improving the accuracy and confidence of object detection.
Patent Information
- Application Number
- CN202111674118.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-31
- Publication Date
- 2025-08-19
- Estimated Expiration
- 2041-12-31
AI Technical Summary
In the case of complex traffic intersections, the lack of edge sensing texture information of single lidar, resulting in increased uncertainty in object detection, and the fusion of multiple sensors fails to fully utilize the accuracy, and the MEMS lidar cannot perform phase locking, affecting the object detection performance of the perception algorithm.
Through the wireless communication technology of vehicle-road collaborative whole-area network, the point cloud data of multi-roadside unit lidar is collected, position labeling and time deviation calculation are performed, and the time offset of the phase of mechanical and MEMS lidar and the false main clock source is adjusted to realize the front fusion and phase synchronization of multi-lidar.
It improves the lack of perceived texture information in complex intersection scenarios, reduces the object detection distortion rate, enhances the detection confidence of point cloud data, and fully utilizes the perception ability of multiple sensors to form a super-visual sensor.
Abstract
Description
Technical Field
[0001] The present invention relates to the field of road traffic technology, and in particular to a phase synchronization method for front fusion of multiple laser radars based on vehicle-road collaboration. Background Art
[0002] At present, under the vehicle-road cooperative full-area network, the phase adjustment technology of lidar is mainly as follows: the roadside unit and the computing unit form a perception suite, independently perceive the object output, and then adopt a post-fusion strategy to realize object tracking under the full-area network; multiple types of perception devices only implement redundant solutions in the post-fusion of perception algorithms; roadside units and roadside units perform pre-fusion of data, and obtain the timestamp information of each point in the point cloud as the data compensation input of the pre-fusion algorithm; mechanical lidar has a phase-locked configuration, but MEMS lidar does not have a phase-locked configuration.
[0003] The above process has the following disadvantages:
[0004] In complex traffic intersection scenarios, post-fusion perception algorithms lose feature information when processing edge information in point cloud data, increasing the uncertainty of object detection.
[0005] The post-fusion algorithm utilizes the ability of multiple sensor devices to independently detect objects, which only increases the redundancy of object tracking detection and fails to achieve the accuracy of object detection under the pre-fusion method.
[0006] The pre-fusion perception algorithm only uses the radar's position calibration information to fuse point cloud data, then selects a main radar as the filtering standard for point cloud stitching, and relies on the compensation algorithm for object detection and filtering.
[0007] Mechanical lidar can achieve phase calibration before splicing and fusion of multiple mechanical lidars through phase locking. However, when performing phase calibration with MEMS lidar and mechanical radar, the phase of the mechanical radar can only be passively adjusted by relying on the overlapping scanning area of the mechanical radar and MEMS in the scene. Summary of the Invention
[0008] The present invention addresses the above technical issues, overcomes the shortcomings of the prior art, and provides a phase synchronization method for multi-lidar front fusion based on vehicle-road collaboration, comprising:
[0009] (1) Aggregate lidar point cloud data from multiple roadside units;
[0010] (2) Splicing point cloud data and labeling their sources according to the position coordinates of each lidar;
[0011] (3) Calculate and count the time deviation of the point cloud data of the splicing edge and the overlapping side;
[0012] (4) According to the correspondence between the time deviation and angle of each lidar, the phase of the mechanical lidar and the time offset of the pseudo master clock source of the MEMS lidar are adjusted.
[0013] The technical solution further defined in the present invention is:
[0014] In the aforementioned phase synchronization method of multi-lidar front fusion based on vehicle-road collaboration, multiple roadside units use the wireless communication technology of the vehicle-road collaborative global network to realize the aggregation of point cloud data, input the position calibration parameters of the lidar, and splice the point cloud data of multiple lidars together.
[0015] The aforementioned phase synchronization method of multi-lidar front fusion based on vehicle-road collaboration, the synchronous stitching of lidar point clouds of multiple roadside units: the lidar point cloud data of multiple roadside units are aggregated, and the point cloud data are stitched together according to the position calibration parameters of the lidar. Then, the point cloud source is input, and the time difference of the point cloud data from different sources is calculated relative to the emission rotation center. Then, the phase lock value is obtained by averaging and input into the driving configuration of the lidar to complete the phase lock of the lidar.
[0016] The phase synchronization method of the multi-lidar front fusion based on vehicle-road collaboration mentioned above is plus or minus 5 degrees relative to the emission rotation center.
[0017] The aforementioned phase synchronization method of multi-lidar front fusion based on vehicle-road collaboration uses the MEMS lidar to calibrate the micro-vibration mirror every second, uses time deception technology to adjust the scanning phase of the MEMS lidar, and then, at the data layer, realizes data packet segmentation and grouping according to the angle of time offset, and finally performs time compensation to realize the phase calibration of the MEMS lidar.
[0018] The aforementioned phase synchronization method of multi-lidar front fusion based on vehicle-road collaboration and phase locking of MEMS lidar: input the MEMS laser time sequence table to obtain the correspondence between angle and time offset, and then perform port protection on the switch side. The MEMS lidar and the fake master clock source are constrained in the sub-LAN, and the time deviation of the fake master clock source is adjusted to adjust the emission timing of the first laser point of the MEMS. Finally, the data stream is time compensated at the data layer, and then framing is achieved based on the cutting angle.
[0019] The beneficial effects of the present invention are:
[0020] (1) The present invention can improve the insufficient perception of a single sensor. It relies on phase locking to converge and synchronously stitch data from multiple laser radars to form the perception results of the beyond-horizon sensor. This can make up for the insufficient performance of the perception algorithm in effectively detecting objects due to the lack of texture information in edge perception of a single laser radar in complex intersection scenes. At the same time, it can reduce the computing load of the perception algorithm in processing the object detection distortion rate caused by the time deviation of the point cloud data. In addition, it can also enhance the confidence of point cloud data detection.
[0021] (2) The present invention leverages the advanced wireless communication technology of the vehicle-road collaborative global network to achieve point cloud data aggregation, stitching together point cloud data from multiple lidars to fully restore the global feature information of complex traffic scenes;
[0022] (3) The present invention uses the front fusion algorithm to give full play to the perception capabilities of various types of lidars and aggregate multiple sensors into an ultra-high-definition super sensor;
[0023] (4) The present invention calculates the time difference of the spliced laser radar point cloud information, adjusts the phase of multiple mechanical laser radars, realizes the phase calibration of multiple types of laser radars, reduces the object detection distortion rate caused by the time difference in the spliced point cloud data processed by the perception algorithm, and thus reduces the weight of the compensation algorithm. DETAILED DESCRIPTION
[0024] This embodiment provides a phase synchronization method for multi-lidar front fusion based on vehicle-road collaboration, including:
[0025] (1) Aggregate lidar point cloud data from multiple roadside units;
[0026] (2) Splicing point cloud data and labeling their sources according to the position coordinates of each lidar;
[0027] (3) Calculate and count the time deviation of the point cloud data of the splicing edge and the overlapping side;
[0028] (4) According to the correspondence between the time deviation and angle of each lidar, the phase of the mechanical lidar and the time offset of the pseudo master clock source of the MEMS lidar are adjusted.
[0029] Synchronous stitching of lidar point clouds from multiple roadside units: The lidar point cloud data from multiple roadside units is aggregated and stitched together based on the lidar's position calibration parameters. The point cloud source is then input, along with the time difference relative to the launch rotation center (plus or minus 5 degrees). The time difference between point cloud data from different sources is calculated and averaged to obtain a phase lock value. This value is then input into the lidar's drive configuration to complete the lidar's phase lock.
[0030] The MEMS LiDAR calibrates the micro-mirror every second, employing time deception technology to adjust the scanning phase of the MEMS LiDAR. Then, at the data layer, data packets are segmented and grouped according to the time offset angle. Finally, time compensation is performed to achieve phase calibration of the MEMS LiDAR. Phase locking of the MEMS LiDAR: The MEMS laser time sequence table is input to obtain the corresponding relationship between angle and time offset. Port protection is then implemented on the switch side. The MEMS LiDAR and the fake master clock source are constrained to the sub-LAN. Time deviation adjustment is performed on the fake master clock source to adjust the emission timing of the MEMS's first laser point. Finally, time compensation is performed on the data stream at the data layer, and framing is achieved based on the cutting angle.
[0031] In a vehicle-road collaborative global network, complex intersection dynamic traffic information scenarios involve the deployment of multiple roadside units. Before the LiDARs of these multiple roadside units are spliced together, the LiDAR phase can be adjusted based on the LiDAR scanning principle and LiDAR position calibration information. This invention addresses the inability of MEMS LiDARs to achieve phase lock. By using time synchronization technology to deceive the MEMS LiDAR time, the initial point scanning time is locked. Then, time compensation is performed to achieve phase adjustment of the MEMS LiDAR.
[0032] In addition to the above embodiments, the present invention may also have other implementations. Any technical solution formed by equivalent replacement or equivalent transformation falls within the scope of protection required by the present invention.
Claims
1. A phase synchronization method for multi-lidar front fusion based on vehicle-road collaboration, characterized by: include: (1) Aggregate lidar point cloud data from multiple roadside units; (2) Splicing point cloud data and labeling their sources according to the position coordinates of each lidar; (3) Calculate and count the time deviation of the point cloud data of the splicing edge and the overlapping side; (4) According to the correspondence between the time deviation and angle of each laser radar, the phase of the mechanical laser radar and the time offset of the pseudo master clock source of the MEMS laser radar are adjusted, and the data packets are segmented and grouped according to the angle of the time offset. Finally, time compensation is performed to realize the phase calibration of the MEMS laser radar.
2. The phase synchronization method for multi-lidar front fusion based on vehicle-road collaboration according to claim 1 is characterized by: Multiple roadside units use the wireless communication technology of the vehicle-road cooperative full-area network to aggregate point cloud data, input the position calibration parameters of the lidar, and splice the point cloud data of multiple lidars together.
3. The phase synchronization method for multi-lidar front fusion based on vehicle-road collaboration according to claim 2 is characterized by: Synchronous stitching of lidar point clouds from multiple roadside units: The lidar point cloud data from multiple roadside units is aggregated and stitched together based on the lidar's position calibration parameters. The point cloud source is then input, and the time difference between the point cloud data from different sources relative to the emission rotation center is calculated. This is then averaged to obtain a phase lock value, which is input into the lidar's drive configuration to complete the lidar's phase lock.
4. The phase synchronization method for multi-lidar front fusion based on vehicle-road collaboration according to claim 3 is characterized by: ±5 degrees relative to the launch rotation center.
5. The phase synchronization method for multi-lidar front fusion based on vehicle-road collaboration according to claim 1 is characterized by: With the help of MEMS laser radar, the micro-vibration mirror will be calibrated every second, and the time deception technology will be used to adjust the scanning phase of the MEMS laser radar. Then, in the data layer, the data packets will be segmented and grouped according to the angle of time offset. Finally, time compensation will be performed to realize the phase calibration of the MEMS laser radar.
6. The phase synchronization method for multi-lidar front fusion based on vehicle-road collaboration according to claim 5 is characterized by: Phase locking of MEMS lidar: Input the MEMS laser time sequence table to obtain the correspondence between angle and time offset. Then, perform port protection on the switch side, constrain the MEMS lidar and the fake master clock source to the sub-LAN, and adjust the time deviation of the fake master clock source to adjust the emission timing of the first MEMS laser point. Finally, perform time compensation on the data stream at the data layer, and then implement framing based on the cutting angle.