Target Localization from 3D data for On-Orbit Autonomous Rendezvous & Docking

Neptec has developed a vision system for autonomous on-orbit rendezvous and docking that does not require the use of cooperative markers on the target spacecraft. The system uses an active TriDAR 3D sensor and efficient model based tracking algorithms to provide 6 degree of freedom (6DOF) relative pose information in realtime. The TriDAR (triangulation + LIDAR) sensing technology combines triangulation and time-of-flight (ToF) active ranging techniques within the same optical path. This configuration takes advantage of the complementary nature of these two imaging technologies and allows the system to provide fast and accurate 3 dimensional data at both short and long range. In partnership with the Canadian Space Agency (CSA), Neptec has developed a novel object localization algorithm that calculates the relative pose of a target spacecraft without requiring an initial estimate of the pose. This algorithm will be used to automatically initiate the model based tracking process and recover if tracking is lost. The technique was specifically designed for real-time operations in space where a target spacecraft could be tumbling and processing power is limited. Most traditional approaches to object recognition and pose estimation algorithms require high resolution data arranged in a grid such that convolution based operators can be used. This generally means that data acquisition is slow and real-time data processing requires a powerful computer. Neptec's approach follows the more information less data (MILD) paradigm. It requires only a unorganized sparse 3D point cloud of the target, keeping the data acquisition time to a minimum. This paper presents an overview of the overall tracking system as well as results from testing the localization algorithm. The algorithm was tested with both simulated and TriDAR sensor data.