System design for target recognition and position-pose measurement based on TOF camera

Abstract

An embedded system based on the TOF camera with laser as the driving light source is designed. The gray image and depth image are collected to detect and match the cooperative target in real time, and the pose information of the target object is obtained. The target used is circular, and the classical threshold algorithm is used to extract the centroid coordinate of the target and calculate the center-of-heart rate as the target recognition condition. The target matching is completed using the classical deterministic annealing algorithm, and two sets of three-dimensional coordinates are obtained. And the SVD method is used to solve the pose data. In the experiment, the position calculated by the TOF camera can achieve an accuracy of 0.13 degrees, and the position translation has an accuracy of 0.6%. It can accurately identify the target and calculate pose, which verifies the feasibility of the TOF camera for rendezvous and docking, SLAM and other fields.

Keywords

• TOF camera
• target recognition
• deterministic annealing
• pose solution