Overcoming camera instability problem for detecting and tracking moving objects in video using reduced data

Abstract

Moving objects detection is a vital field of study in various applications. Many of such applications may have to capture and process a lot of data, then such these data need to be reduced as much as possible in order to have a reasonable and suitable system for achieving the desired aims efficiently. The proposed algorithm utilizes singular value decomposition (SVD) and Bayer pattern filter for their good properties in producing very representative reduced data. This data is then handled by frame difference objects detection, which in turn is an approach that doesn’t need to handle much data. The camera shaking which can be caused by a windy weather in the case of the outdoor static camera may introduce a frame difference with imprecise moving objects detection, hence frames compensation is conducted utilizing a transformation based on speed up robust feature transform (SURF) detected key points.