This paper presents an automatic method to detect and follow people on video streams. This method uses two techniques to determine the initial position of the person at the beginning of the video file: one based on optical flow and the other one based on Histogram of Oriented Gradients (HOG). After defining the initial bounding box, tracking is done using four different trackers: Median Flow tracker, TLD tracker, Mean Shift tracker and a modified version of the Mean Shift tracker using HSV color space. The results of the methods presented in this paper are then compared at the end of the paper.
This paper describes a complete pedestrian detection system based on sliding windows. Two feature vector extraction techniques are used: HOG (Histogram of Oriented Gradient) and CSS (Color Self Similarities), and to classify windows we use linear SVM (Support Vector Machines). Besides these techniques, we use mean shift and hierarchical clustering, to fuse multiple overlapping detections. The results we obtain on the dataset INRIA Person shows that the proposed system, using only HOG descriptors, achieves better results over similar systems, with a log average miss rate equal to 43%, against 46%, due to the cutting of final detections to better adapt them to the modified annotations. The addition of the modified CSS increases the efficiency of the system, leading to a log average miss rate equal to 39%.