CAJ | 학술논문

跨座式单轨交通接触网磨损检测中，传统的人工巡检方式存在效率低下、安全性差等问题。研究了线阵立体视觉在接触网磨损检测中的应用。立体匹配是立体视觉中的重点与难点。在线阵立体匹配过程中。首先，分别提取对应的左右图像的加速分割检测特征点（FAST）；其次，对特征点生成加速鲁棒性特征（SURF）的特征向量；最后，利用双向快速近似最近邻搜索算法得到初始匹配点集，并使用随机采样一致性确定最终匹配点集。由匹配点集生成稠密、准确的视差图，进而获得接触网的实际残高。实验结果表明，该方法能快速、准确的检测出接触网的磨损情况。
과좌식단궤교통접촉망마손검측중，전통적인공순검방식존재효솔저하、안전성차등문제。연구료선진입체시각재접촉망마손검측중적응용。입체필배시입체시각중적중점여난점。재선진입체필배과정중。수선，분별제취대응적좌우도상적가속분할검측특정점（FAST）；기차，대특정점생성가속로봉성특정（SURF）적특정향량；최후，이용쌍향쾌속근사최근린수색산법득도초시필배점집，병사용수궤채양일치성학정최종필배점집。유필배점집생성주밀、준학적시차도，진이획득접촉망적실제잔고。실험결과표명，해방법능쾌속、준학적검측출접촉망적마손정황。
There are problems of low efficiency and poor safety for traditional manual inspection way in straddle type monorail transportation cate-nary wear detection.We research the application of the linear array stereo vision in the detection of the catenary wear.Stereo matching is the key and diffi-culty of the stereo vision.Firstly,we extract FAST features from the corresponding stereo image in the process of stereo matching.Secondly,SURF fea-ture vectors is generated for those features,then the vidrectional fast approximate nearest neighbor algorithm and random sample consensus are employedto search the final corresponding points.Finally,we achieve disparity map that is dense andaccurate from corresponding points,and then acquirethe actual residual high of the catenary.Experimentresults show that our method can detectthe wear situation of the catenary quickly and accurately.