Obstruction Detection Based on Light Field Image Processing

Detection of foreign object intrusion on railway track is a critical research topic to railway information systems. Current imaging systems often suffer from a lack of spatial scene information due to the reliance on single visible-light sensors. A light-field imaging-based scheme for foreign object intrusion detection is proposed. By extracting saliency features from light field data, it leverages the capability of light field cameras to simultaneously capture both RGB image information and depth information. Firstly, traditional manifold sorting (MR) algorithm is employed to improve the accuracy of RGB image detection. Then the salient features of depth images are extracted as a supplement to RGB image processing. Finally, an effective data fusion method is proposed to further highlight the object contour and suppress the background, so as to improve the detection performance. The experimental results show that the proposed light-field-based scheme has the application ability to detect foreign objects in rail transit environment efficiently.