Multi-Object Tracking Based on a Novel Feature Image With Multi-Modal Information

Multi-object tracking technology plays a crucial role in many applications, such as autonomous vehicles and security monitoring. This paper proposes multi-object framework based on the multi-modal information of 3D point clouds color images. At each sampling instant, cloud image acquired by LiDAR camera are fused into cloud, where objects detected Point-GNN method. And, novel height-intensity-density (HID) is constructed from bird's eye view. The HID truly reflects shapes materials effectively avoids influence object occlusion, which helpful to tracking. In two sequential images, new rotation kernel correlation filter proposed predict objects. Furthermore, an retention module re-recognition developed overcome matching failure in-between frames. method takes full advantage data achieves complementation improve accuracy experiments with KITTI dataset show that has best performance among existing traditional methods.