Kalman Filter Based Motion Estimation for ADAS Applications

Autonomous cars are no longer in the sphere of just science fiction but they are out in reality. The accelerated rise of disruptive technologies like diverse mobility, autonomous driving, electrification, and connectivity are the forces behind this revolutionary innovation. One of the crucial parts of the driverless cars lies behind the accurate results in detection and tracking of obstacles, especially pedestrians.This paper focuses on developing a solution to detect and track pedestrians and provide full-fledged analysis under various real-time conditions like occlusions, non-linear motion and multiple pedestrian. Frame extraction, collection of Meta data, HoG descriptors and non-maximum suppression make up a major part of detection. The tracking is done using Kalman filter, a linear, Gaussian filtering procedure that tracks the pedestrian using iterative measurements observed over a definite period of time.The results obtained are compared with the reported works.