作者: Esther Resendiz , John Hart , Narendra Ahuja
DOI: 10.1109/TITS.2012.2236555
关键词:
摘要: Thousands of miles railroad track must be inspected twice weekly by a human inspector to maintain safety standards. A computer vision system, consisting field-acquired video and subsequent analysis, could improve the efficiency current methods. Such system is prototyped, following challenges are addressed: detection, segmentation, defect assessment components whose appearance vary across different tracks identification inspection special areas such as turnouts. An algorithm that utilizes periodic manner in which repeat an developed. Spectral estimation signal-processing methods used provide robust detection periodically occurring components. Results demonstrated on images video.