REAL TIME OBJECT TRACKING IN VIDEO SURVEILLANCE USING PRIOR CONTEXT AND IMAGE PATCHING

Abstract

Here we discuss a fast and a robust algorithm for tracking an object in video surveillance, there are various challenges when it comes to dealing with video tracking, such as image illumination, motion object, occlusion, scaling etc. we present our algorithm that handle such measure, we start with generating the weighted map for the image, then we track the object in form of patches, whose combined effect is used as a vote map. Then using image signature and spatiotemporal computation we generate confidence map to predict the next position. Our algorithm doesn’t require training, we just provide the initial position of the object which need to be track. Numerous experiment result show that our algorithm achieve significant improvement our SPC tracker (Cong Ma, 2017).