Vehicle Detection and Tracking from Airborne Images

Abstract

For road traffic monitoring from aerial videos, vehicle detection and tracking are the basic tasks. However, background motions, scene complexity, and congested road conditions increase the difficulty. To address these issues, a system for vehicle detection and tracking was developed. The Viola-Jones object detection scheme is applied to vehicle detection, and vehicle tracking is proposed based on rectangular region prediction and similarity of gray-level histogram. Experiments in different scenarios showed that the system can achieve a high detection rate and tracking rate with a low false alarm rate.