CSCA-Based Expectivity Indices for LIDAR Computer Vision

Continuum-Shape Constraint Analysis (CSCA) is a shape analysis approach applicable to pose estimation tasks in computer vision. A variety of useful measures (indices) which predict the accuracy of pose estimation can be derived from CSCA. Conceived for computer-vision assisted spacecraft rendezvous analysis, the approach was developed for blanket or localized scanning by LIDAR or similar range-finding scanner that samples non-specific points from the object across the area observed from a single view. The application problem addressed in this paper is the question of what view of an object can be expected to lead to the lowest pose estimation error computed via the Iterative Closest-Point Algorithm (ICP), or conversely, what level of error can be expected for a particular scan view. Based on CSCA, different forms of indices are developed for this purpose and demonstrated in both numerical and experimental studies using the Stanford Bunny and a cuboid shape. The continuum nature of the CSCA formulation produces metrics, including the Expectivity Index, that are pure shape properties an object. Key-Words: Computer Vision, LIDAR, ICP, Pose Estimation, Expectivity Index