Video stabilization and motion detection using Matlab video processing toolbox

This paper presents video stabilization and motion detection using Matlab Simulink. Simulink is a classroom learning tool that offers an environment for multi-domain simulation and model-based design for dynamic and embedded systems. The project utilizes that tool to solve not only a problem common to amateur videos but also a situational awareness issue. The stabilization process starts with computing the optical flow between successive frames. The motion vectors are computed per block of 4×4 pixels for a good representation of the jerkiness present in the video and faster processing. From the optical flow, there is an affine representation of the unwanted motions using four parameters. The four parameters represent the camera motion by rotation matrix and translation vector. The smoothing process of the jerky video uses the inverse rotation matrix and oppose translation vector. The motion detection is done by frame difference block provided by Simulink. A tracking function allows the localization of the moving object. Simulink processing blocks represent the different steps taken in implementing the project. The video for complete testing allows detection only after stabilization.