An Information Fusion-based Vision System for an Autonomous Vehicle

This paper presents an information fusion-based vision system for an autonomous vehicle. To perceive road environments, the autonomous vehicle uses multiple sensors such as cameras, laser radar sensors, global positioning system (GPS), and in-vehicle sensors. The cameras, one of the multiple sensors, are used for object detection by means of color segmentation, multi-scaled template matching, and AdaBoost algorithm. However, since the cameras are influenced by outdoor conditions such as sudden illumination changes and shadows, detection performance cannot be guaranteed. To overcome the problem, the information fusion is applied using the multiple sensors. The information fusion-based vision system performs the object detection with low false positives and fast computation time. Finally, the proposed approach is validated through on-the-road experiments with various environmental conditions.