Object Detection and Tracking in Discriminant Subspace

Abstract

Detection and tracking of moving objects in video is essential for many computer vision applications and it is considered as a challenging research issue due to dynamic changes in background, illumination, object size and shape. Many traditional algorithms fails to detect and track the moving objects accurately, this paper proposes a robust method, to detect and track moving objects based on the combination of background subtraction and Orthogonalized Fisher’s Discriminant (OFD). Background subtraction detects the foreground objects on subtracting frame by frame basis and updating the background model recursively. Orthogonalized Fisher’s Discriminant projects high dimensional data onto a one dimensional space with the highest recognizability, which speedup the detection and tracking process and also preserves the structure of the objects resulting high accuracy. The proposed method is tested on standard datasets with complex environments and experimental results obtained are encouraging.