Multi-Information Fusing Based Railroad Object Detection

This paper proposes a railroad object detection algorithm based on multi information fusing, supplying the drawback of traditional surveillance methods in wild environment. Firstly, the railway is found in a given image by template matching. Then, integrity and consistency descriptors are computed to find out suspicious object. Finally, whether there exits an obstacle is judged in the railway via potential function after size revised, which is one of our main contributions. In addition, passing-by trains are detected via vibration signals. Our experiments in different kinds of wild railroad situations have showed better results than traditional methods.