An Improved Algorithm for the Estimation of Multibody Motion

The problem of estimating ego motion and eoru motions through a vehicle mounted with a camera is related in this paper. Localization of multiple moving objects and estimating their motion are crucial for autonomous vehicles, but it is not that much successful in estimating the motion of the moving vehicles and objects. Ego motion can be calculated only by conventional localization and mapping techniques. The framework for estimation of multiple motions in addition to the camera ego motion is presented. The video is processed through MATLAB for pre-processing. The video is then segmented into frames, and then the framework is done to estimate the multibody motion through different algorithms. The algorithms like block matching algorithms, corner detection, and background subtraction algorithm are used to estimate the multibody motions of the moving objects. From this, we can detect and estimate the motion and speed of the object in the frame. Then, it is processed in hardware (raspberry pi3) using the same algorithms, so that we can effectively use it in any autonomous cars.