Integrating Object Detection and Tracking in Outdoor Environment

摘要

Object tracking is one of the key technologies in smart visual surveillance. However, due to the existing of varying sunlight, shadows, plants swaying, the difficulty of object tracking is greatly increased in outdoor environment. To tackle this problem, a new algorithm based on the augmented particle filter is proposed in this paper. It is able to combine object detection and tracking. Here, we employ Gaussian mixture model to model the background of the monitored environment. Furthermore, we extract the binary image of object by background subtraction as the corresponding observation. In the subsequent tracking phase, we use Kalman filter to introduce the most recent observations into particle filter and produce the suboptimal Gaussian proposal distribution. Experimental results demonstrate the proposed algorithm can handle a certain degree of complexity in outdoor environment.