Moving object recognition and detection using background subtraction

Abstract

Motion detection and object recognition algorithms are a significant research area in computer vision and involve building blocks of numerous high-level methods in video scrutiny. In this paper, a methodology to identify a moving object with the use of a motion-based segmentation algorithm, i.e. background subtraction, is explained. First, take a video as an input and to extract the foreground from the background apply a Gaussian mixture model. Then apply morphological operations to enhance the quality of the video because during capture the quality of a video is degraded due to environmental conditions and other factors. Along with this, a Kalman filter is used to detect and recognize the object. Finally, vehicle counting is complete. This method produces a better result for object recognition and detection.