Beijing Qingzhouzhihang Intelligent Technology Co., Ltd

Beijing Qingzhouzhihang Technology Co., Ltd.

Chenxu Luo

Xiaodong Yang

A method and a device for multi-object tracking, and an electronic device are provided. The method includes: determining a hybrid-time position map of a 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; 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. With the method and the device for multi-object tracking, and the electronic device, the hybrid-time position map and temporary tracked position map of the current point cloud fragment are averaged, so that not only the tracked position map of the current point cloud fragment is accurately generated, but also an object ID is inherited. Based on the object ID, the same object in different point cloud fragments are associated, so that multi-object tracking is implemented without an association step in the conventional solutions. It is unnecessary to set additional hyper-parameters, and strong versatility is achieved.

System and method for 3D multi-object tracking in LiDAR point clouds

A method and a device for self-supervised learning, a storage medium, and an electronic device are provided. The method includes: organizing real points in one column along a vertical direction into a pillar; determining a predicted point in a next frame; determining a first loss term based on a minimum distance among distances between predicted points in the next frame and real points in the next frame, and generating a loss function including the first loss term; and performing self-supervised learning processing based on the loss function. A pillar motion parameter representing motion of a real point is determined with the pillar as a unit, so as to enhance correlation between point clouds. Self-supervised learning can be realized in a case of no precise correspondence between the predicted point and the real point, and training is performed based on a large number of unlabeled point clouds.

Growing community of inventors

Beijing, China

Chenxu Luo