Motion Based Object Detection using Background Subtraction Technique for Smart Video Surveillance

Abstract

newlineIn video surveillance our main aim is to detect the moving object from video frames. Generally, the objects of interest are moving humans, vehicle, animal, robot, etc. It is also known as foreground. In this thesis work, we have used a background subtraction technique which is a widely used technique for detecting moving objects in video. According to the literature, the video can be captured through the following situations. newlinei. The camera is static position and the background is static. newlineii. The camera is in static position and the background is in dynamic. newlineIn real time, lots of challenges exists in the background of video frames due to various issues such as bootstrapping, camouflage, shadowing, camera hardware noise, illumination changes , slow leafy movement, rippling or spouting water in the background. For resolving some of the above issues, this thesis work is used background subtraction technique. newlineThis thesis aims to show the performance of efficient detection of object in both cases colored and thermal video sequences that are captured from a fixed camera. newlineIn the first objective we explore an efficient method that enhances the existing method by using some image processing techniques in order to improve detection quality of classified pixels and compare them against existing methods newlineIn the second objective of this thesis work, we developed a histogram based background subtraction technique. This works performed over colored video frames. Here, the background model is developed using initial few frames and then difference is computed with the current frame. newlineThe third objective of this thesis work investigates a Fishers ratio based threshold for pixel classification using background subtraction technique. This method also depicts better results and minimizes the illumination variation and dynamic nature of the background. This thesis work has explored the effectiveness of the moving object detection in both colored and thermal video frames for real-time video surveillance applications. newline newline newline newline newline newline newline newline newline newline newline newline newline newline