For an object that is rotating in depth, the feature points near the leading boundary are always disappearing in subsequent frames. These feature points should not be tracked over consecutive frames because they correspond to different points on the object. Modify the Lucas-Kanade-Tomasi tracking method to filter out these feature points. Your algorithm should be able to track feature points over multiple frames that are not lying along the leading boundary. To do so, you need to first identify the boundary of the object...