Many technical applications require precise awareness of the position and orientation of moving objects in 3D space. 2D image data lack depth information, so it is difficult to estimate the 3D pose of an object or to reconstruct its 3D shape. Instead, laser scanners allow the acquisition of accumulated 3D point clouds by successive and point-wise scanning of the scene. However, if laser scanners are used to detect objects that are moving unpredictably, these objects appear misplaced and distorted within the scanned data. In contrast to 2D imaging as well as classical 3D laser scanning, matrix LiDAR systems like the flash LiDAR sensor enable 3D acquisition of the complete field of view with a single laser pulse. This sensor principle results in undistorted 3D range images, where the range computation is performed at the same time for each pixel of the detector matrix (here: 128x128 pixels).
In order to demonstrate the advantages of a matrix LiDAR system for 3D object tracking, a person tracker has been developed exemplarily. For this purpose, a flash LiDAR sensor was mounted to a motorized pan&tilt head. The data acquired by the system are converted to a 3D point cloud in real-time, and a simple model (cylindrical shape of reasonable size) is used to detect and track persons in the scene. If a person is detected, the pan&tilt head is used to direct the sensor towards the detected position, such that the moving person is centered in the field of view of the flash LiDAR camera.
Video: 3D person detection and tracking with a flash LiDAR camera.
- Hammer, M., Hebel, M., Arens, M., 2016. Automated object detection and tracking with a flash LiDAR system. Proceedings of SPIE 9988, Electro-Optical Remote Sensing X, 998803. [DOI: 10.1117/12.2240640]