Multi object detection techniques in video surveillance application

One of the important task in computer vision is to accurately detect the moving objects from dynamic scene. The most popular and repeatedly used technique for object detection is Background subtraction (BS) method. Background modeling is undoubtedly the first step in BS method. For accurate object detection, exact construction of background model is very important. Further steps of video surveillance system such as object recognition, tracking are greatly dependent on quality of detection. In this paper, we have implemented the object detection techniques such as Approximate Median filter(AMF), Gaussian Mixture Model (GMM) and Optical flow method (OF) and compare their results with the help of quantitative analysis. Experimental results gives the performances of implemented algorithms on various video sequences.