Lucas Kanade based Optical Flow for Vehicle Motion Tracking and Velocity Estimation

Abstract

Optical flow is a powerful application of image processing that is used in a variety of applications, primarily in object tracking and motion estimation. In this paper, we implement a system for vehicle motion tracking and velocity estimation using Lucas-Kanade (L-K) algorithm based optical flow method. The work includes two applications of optical flow: tracking the movement of the vehicle in the case of a fixed camera and velocity estimation of a vehicle with a camera mounted on it. Pre-processing steps include gaussian smoothing, and computing spatial and temporal gradients. This is followed by the further formulation of Lucas kanade equation in the form of matrices. The system of equations is then solved using the least square error criteria, and the flow vectors are obtained. Processes such as segmentation, blob analysis, camera calibration, and thresholding are further done which are used for velocity estimation as well as motion tracking. The functionality was tested and verified on video sequences obtained from the lab testing scenarios and real-world camera visuals taken from various sources.