Real-Time Front Vehicle Detection Algorithm for an Asynchronous Binocular System

摘要

This paper describes a multi-resolution stereovision system for detecting the front-vehicle in advanced safety vehicles (ASVs). The two asynchronous CMOS cameras in the proposed system are mounted on a platform that can be easily clamped to the rear-view mirror of a vehicle for detecting vehicles ahead. The asynchronous binocular platform provides a small low-cost obstacle detection system for practical ASVs that is easy to set up. The system uses a stereovision vehicle detection algorithm for real-time matching because the exposure times of the CMOS cameras are not synchronous. The algorithm uses a line segment matching module to match the extreme points of the horizontal and vertical edge segments at different resolutions to decrease the search area and computing complexity. As the distance of each matched segment can be calculated from the disparity value, each vehicle can be detected by clustering the segments that have similar distances in a searching and distance estimation module. The system was evaluated using static and dynamic analyses. Experimental results show that the proposed system can robustly and accurately detect the front-vehicles in real time under different illumination and road conditions.