An Algorithm Based on ACF and SRG for Abnormality Detection of Bolts on the Subway

Bolt detection is an essential part of the train inspection system. This paper presents a novel method of wear and looseness detection of the bolt on the subway, based on the depth and intensity maps captured by a 3D line-scan camera. First, this paper adopts the aggregate channel features algorithm to locate bolts by two steps on the intensity map. Second, by combining the seeded region growing and the random sample consensus algorithm(RANSAC), the method achieves region segmentation in complex environments. RANSAC fits the plane of upper and lower surfaces of the bolt. The difference between the average depth values of inliers with the twice RANSAC is defined as the nut thickness. Third, the wear and looseness detection is realized by comparing it with the historical data. Finally, the 3D image is reconstructed to facilitate human observation. Algorithm verification has been carried out on different trains and locations of bolts. In dynamic operation, the detection success rate is above 94.44%. The relative detection error of the same vehicle with the latest data is within 0.3mm, proving the effectiveness and robustness of the proposed detection scheme.