earticle

영어

An approach for analyzing abnormal action of pedestrian based on video sequence was proposed in this paper. Firstly, the binary image of pedestrian was detected using self‐adaptive background updating algorithm. Then, the Radon transform was used to describe geometrical shape of moving person and the energy curves were protracted. Finally, the feature curves of different training action samples were protected, and abnormal action was detected by Euclidean distance. The experimental results show that the abnormal action of pedestrian are exactly analyzed by the proposed method, and it is reasonably robust in background noise and the analysis of moving person which lost detail information.