A Shield Segment Bolt Detection and Localization Method Based on Ellipse Detection

The automatic assembly of shield tunnel segments can significantly reduce labor intensity and improve tunneling efficiency. However, during the automatic segment gripping process, traditional methods of setting targets introduce manual errors and complicate operations. Moreover, single-object detection methods only capture the bolt area, which makes precise bolt positioning challenging. Consequently, this research introduces an approach to detect and pinpoint the position of segment bolts by utilizing the circular imagery of the bolt head’s surface. The technique uses the elliptical shape of the bolt head in images to accurately determine bolt locations through ellipse detection methods. In response to the issue of decreased elliptical detection accuracy due to circular arc defects, an arc-cluster-based elliptical detection algorithm is further proposed, which encompasses key steps such as image preprocessing, arc clustering, and least squares fitting. Furthermore, in response to the challenge that monocular vision systems face in obtaining vertical data with point lasers during the gripping process, this study presents a monocular positioning strategy. This strategy is founded on the constant diameter of the bolt head’s circular form, facilitating the determination of the bolt’s spatial orientation without the need for additional height measurements. This strategy is based on the constant diameter of the bolt head’s circular shape, facilitating the determination of the bolt’s spatial orientation without the need for additional height measurements. Experimental results demonstrate that this method improves elliptical detection accuracy, and gripping experiments further validate its feasibility.