Occlusion Handling by Depth-value Separation in Indoor Object Detection and Tracking for Range Sensor Device

摘要

Consistently identifying moving objects in object detection and tracking system when occlusion occurred is somewhat a great challenge when the information of the objects are insufficient. While no color information can be extracted from the range device, the depth information can be used in the similar manner as color information to track the moving objects in time. Moreover, by utilizing watershed algorithm for the depth information, occlusion handling by separating two or more objects via segmentation of the merged depth value can be achieved. Experimental results show that the separation is reliable to consistently identify objects even during occlusion.