On-road vehicle detection based on appearance features for autonomous vehicles

In this paper, we propose a monocular camera-based vehicle detection system for use in autonomous vehicles. In order to accurately and rapidly detect a vehicle on the real road, we have designed a vehicle detection system that follows two basic steps namely; Hypothesis Generation and Hypothesis Verification. In the hypothesis generation step, a candidate region of vehicles is set by using the shadow properties of the vehicle. In the hypothesis verification step, based on the candidate regions, we are able to distinguish between the vehicle and the non-vehicle. For the hypothesis verification, we use histograms of oriented gradients (HOG) feature and support vector machine (SVM) classifier. To fit the vehicle detection system, detailed settings of the HOG such as the cell, block and bin were selected.