Bolt 3D Point Cloud Segmentation and Measurement Based on DBSCAN Clustering

In complex railway conditions, the segmentation and height measurement of bolt in RGB image is easily affected by background. Although deep learning based method may complete the task, a large quantity of images for training are required. In this paper, an approach by using the traditional DBSCAN clustering in 3D point cloud is proposed to achieve the segmentation and height measurement of bolts. Specially, with the slice clustering along the height direction and the compression clustering on the certain dimensions implemented on the 3D point cloud while considering the information in the designed drawing, all the bolts can be segmented and then their heights are measured with high precision. The real 3D point of a platform with multiple bolts is obtained by the depth camera and the effectiveness of the proposed approach is demonstrated, showing the stronger anti-interference ability.