Detecting moving objects in video frames

Abstract

Object detection and tracking are challenging critical works in many machine vision applications like traffic control, video surveillance, and person tracking.For detecting moving objects in a video, we use object detection algorithms which compares a static background frame at the pixel level with the current frame. In this thesis, we use the combination of kernel density estimation and modified running average method for efficient motion detection. In this method we use single fixed camera with static background for video. Further we use edge detection algorithm together with morphological operation to improve object detection technique and then combine the output of KDE and MRA method. By using adaptive value in modified running average method for detecting the object, output is less affected by motion.