An Algorithm for Automatic Detection of Banners in Surveillance Videos

In this paper we mainly focus on the effects of two different algorithms for automatic detection of banners in practical use. These two algorithms are scale-invariant feature transform (SIFT) and Hough transform-based contour feature extraction method, respectively. A new algorithm based on color threshold and corner detector is described. This algorithm can capture color features and corner of images effectively, especially for objects with certain color such as bright red banners. The experimental results indicate that we can obtain the most stable and effective structure element characteristic in images when applying the algorithm based on color threshold and corner detector.