Active Relearning for Robust On-Road Vehicle Detection and Tracking

This paper aims to introduce a novel robust real time system capable of rapidly detecting and tracking vehicles in a video stream using a monocular vision system. The framework used for this purpose is an actively relearned implementation of the Haar-Iike feature based Viola-Jones classifier capable of classifying image frame regions as a vehicle or non-vehicle. A passively trained supervised system (based on Adaboost) is initially built by cascading a set of weak classifiers working with Rectangular Haar-like features. An actively learned model is then generated from the initial passive classifier by querying misclassified instances when the model is evaluated on an independent dataset. This classifier is integrated with a Lucas-Kanade Optical Flow Tracker and an empirical distance estimation algorithm to evolve the system into a complete real-time detection and tracking system. The built model is then evaluated extensively on static as well as real world data and results are presented.