Dynamic Background Modeling and Foreground Detection using Orthogonal Projection onto the Subspace of Moving Objects
B G Vishruth, M. T. Gopalakrishna, Megha J · 2023
Moving object detection and recognition is an important and very relevant topic for any video surveillance application. Object detection in video sequences is a challenging task due to several hindrances such as illumination variations, shadows, dynamic background, and clutter background. Existing methods are inadequate in addressing these challenges. Therefore, the author proposes a novel system for moving object detection in video based on Orthogonal Preserving Projection (OLPP). OLPP generates orthogonal basis functions which preserve locality better than Locality Preserving Projection (LPP), as it is non-orthogonal which makes it hard in reconstructing the data. Therefore, OLPP is deemed to pertain higher power of discrimination than LPP. The proposed method was tested on standard and the author's own dataset. The results obtained with the proposed system were compared with existing methods and the results are satisfactory. The proposed system is efficient and robust for detecting and recognizing moving objects in video surveillance applications.