Abstract: In computer vision, the fundamental task of background modelling entails removing the static background from a scene in order to extract the foreground objects. Several computer vision applications, including object tracking, motion detection, and video surveillance, require this process as a prerequisite. For background modelling, a variety of techniques have been put forth, from straightforward threshold-based strategies to complex deep learning models. This paper presents a method that includes the K.M.M. baseline model pipeline followed by two pre-processing techniques that address the varying illumination problem. We also go over the difficulties associated with background modelling, including lighting variations, camera jitter, and PTZ, and we highlight some potential future research directions in this area. Finally, we compare the various methods based on their computational complexity, robustness, and MIOU score, and we offer some guidelines for picking the best method for a particular application.

Keywords: Background Subtraction, Foreground Detection, OpenCV, KNN

PDF | DOI: 10.17148/IJARCCE.2023.12654

Open chat
Chat with IJARCCE