Abstract:

To solve the problem of detecting abandoned objects in complex environments, we propose an efficient method. Firstly, the candidate areas of the objects are obtained by comparing the foregrounds derived from a Gaussian mixture model based on partial updating and from an improved threeframedifference method.And the blobs of temporarily static objects are segmented by applying shadow elimination and the connected component analysis. Then the blob whose static duration reaches the threshold is labeled as the abandoned object after ruling out the possibility of a static human by applying Histograms of Oriented Gradients (HOG) pedestrian detection. Finally,to overcome pedestrian occlusion and illumination changes, the Features from Accelerated Segment Test (FAST) local feature matching is applied to the detected abandoned objects. Experimental results show that the proposed method has high accuracy and fast processing speed , and a good antiinterference performance is achieved in complex environments.

Key words: abandoned object detection;Gaussian mixture model;shadow elimination;pedestrian detection;local feature matching