Abandoned object detection using blind motion history analysis

International audienceIn this paper we propose a new approach for the detection of abandoned objects in video. A simple blind tracking of static objects centroid technique is applied to track the motion history of the candidates. Moving objects are detected using Gaussian mixture model based background subtraction method, next, connected component analysis is performed to delineate bounding boxes. Static objects are detected using a blind foreground objects motion history analysis. Ghosts left by removed scene objects are detected by comparing the foreground object edges structures in foreground mask and current frame, finally the abandoned object candidate is checked for being a still person . We have evaluated our approach on PETS2006 benchmark, and the obtained results are promisin