A mesh-divide-based region of interest clustering and forecasting in video frames based on the background/foreground construction

Abstract

Image processing and security surveillance system has more and more widely used in recent society such as bank surveillance and pedestrian tracking. The detection of Region of Interest (RoI) is always been regarded as the most significant in tracking system. One of the algorithm which can be used in RoI detecting is “Density-Based Spatial Clustering of Application with Noise” (DBSCAN). But because of its structure, the runtime consuming costs too much when handling large spatial dataset. Considering the features of image processing, a mesh-divide and Kalman Filter forecasting method is proposed combing DBSCAN for RoI detection and forecasting of image processing. The DBSCAN can be used in the RoI detection and position forecast at the next frame in surveillance system to decrease the runtime cost and improve the accuracy at the same time.