Paper Details

ABSTRACT

— Object detection and tracking are important and challenging tasks in many computer vision applications such as surveillance, vehicle navigation, and autonomous robot navigation. Object tracking becomes even more challenging in the presence of variable illumination conditions, background motion, complex shaped object, partial and full object occlusions, etc. Furthermore, detecting and tracking of the human body are very important to understand the human activity. Several human detection and tracking systems have been developed which use a new class of local image features that are invariant to image scaling, translation and rotation. In addition, these features are partially invariant to illumination changes, affine or 3D projection. In any surveillance system, the camera location is stationary, but there is the change in ambience, orientation, and scaling of the human body. This paper describes scale invariant feature extraction methods for human detection and tracking from surveillance video, which originally consists of two components: feature extraction for human detection and tracking. A detailed study of these methods is presented in the paper along with the advantages and disadvantages of each of these methods.