US 11,989,891 B2
System and method for 3D multi-object tracking in LiDAR point clouds
Xiaodong Yang, Beijing (CN); and Chenxu Luo, Beijing (CN)
Assigned to Beijing Qingzhouzhihang Technology Co., LTD., Beijing (CN)
Filed by Beijing Qingzhouzhihang Technology Co., LTD., Beijing (CN)
Filed on Aug. 6, 2021, as Appl. No. 17/395,626.
Prior Publication US 2023/0043061 A1, Feb. 9, 2023
Int. Cl. G06T 7/246 (2017.01); G01S 17/89 (2020.01); G06V 20/56 (2022.01)
CPC G06T 7/248 (2017.01) [G01S 17/89 (2013.01); G06V 20/56 (2022.01); G06T 2207/10028 (2013.01); G06T 2207/20081 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for multi-object tracking, comprising:
determining a hybrid-time position map of a current point cloud fragment, wherein the hybrid-time position map represents a first appearance position of a valid object in the current point cloud fragment, and the valid object is an object other than a dead object in the current point cloud fragment;
converting a tracked position map of a previous point cloud fragment into a temporary tracked position map of the current point cloud fragment based on a predicted motion field and an ego-motion of a local device, wherein the previous point cloud fragment is a point cloud fragment before the current point cloud fragment, part of point cloud data of the previous point cloud fragment is the same as that of the current point cloud fragment, and the tracked position map comprises position information and an object identity of each object; and
averaging the hybrid-time position map and the temporary tracked position map of the current point cloud fragment, to generate a tracked position map of the current point cloud fragment, wherein the tracked position map of the current point cloud fragment inherits an object identity in the tracked position map of the previous point cloud fragment.