Partitioning Positional and Normal Space for Fast Occlusion Detection

Occlusion detection is a fundamental and important problem in optical sensor inspection planning. Many viewplanning algorithms have been developed for optical inspection, however, few of them explicitly develop practical algorithms for occlusion detection. This paper presents a hierarchical space partition approach that divides both positional and surface normal space of an object for fast occlusion detection. A k-d tree is used to represent this partition. A novel concept of δ – occlusion is introduced to detect occlusion for objects in an un-organized point cloud representation. Based on the δ – occlusion concept, several propositions regarding to a range search on a k-d tree have been developed for occlusion detection. Implementation of this approach demonstrated that significant time can be saved for occlusion detection using the partition of both positional and surface normal space.