In this paper we propose a coarse-to-fine method to detect pedestrians in video sequences. The detection process is divided into two stages: ROI (region of interest) generation stage and ROI classification stage. In the generation stage haar-like features are exploited to rapidly search the whole image and find interesting regions which may contain pedestrians. In the classification stage shapelet features are used to classify interesting regions into pedestrian region and non-pedestrian region. To evaluate the performance of our method, we test it on several video sequences taken from different scenes and compare it against the HOG-SVM pedestrian detector provided in OpenCV library. Experiment results show that our method achieves comparable performance to the HOG-SVM detector with an average 90% detection rate. But our method is about 50% faster than the HOG-SVM detector.
展开▼
