Motion Determination Using Non-uniform Sampling Based Density Clustering

Motion analysis in sequential images has many applications in both modern military and civilian environments. Motion determination based on frame difference is to determine the number, location and size of moving targets using the spatial clue. Density-based clustering algorithm DBSCAN is introduced into motion determination in this paper, which can discover clusters of arbitrary shape as well as to distinguish noise in spatial database, but it can only work well on data sets where cluster densities are similar. However, significant density difference usually occurs among different moving regions in frame difference images, and DBSCAN can not work well on them. This paper proposed a non-uniform sampling method to smooth the density difference among different moving regions, which makes it possible to obtain reasonable clustering results by DBSCAN. On the other hand, the sampling scheme can reduce the size of point data dramatically, which speeds up the clustering process for moving targets. Some experiments demonstrate the effectiveness of the proposed method.