Note: An object detection method for active camera

Abstract

To solve the problems caused by a changing background during object detection in active camera, this paper proposes a new method based on SURF (speeded up robust features) and data clustering. The SURF feature points of each image are extracted, and each cluster center is calculated by processing the data clustering of k adjacent frames. Templates for each class are obtained by calculating the histograms within the regions around the center points of the clustering classes. The window of the moving object can be located by finding the region that satisfies the histogram matching result between adjacent frames. Experimental results demonstrate that the proposed method can improve the effectiveness of object detection.