本文提出的K-medoids噪点预处理算法实现步骤如图5所示：其中步骤

The implementation steps of the K-medoids noise preprocessing algorithm proposed in this paper are shown in Figure 5: Step 1 is data collection: adjust the camera pose to collect contour circle data; Step 2 is data grouping, and the contour data is divided into 3 groups in order and equally ; Step 3 is data conversion: the contour data is converted to the vicinity of the center of the circle by fitting a circle with three points; Step 4 is cluster analysis: clustering the converted data by K-medoids clustering algorithm; Step 5 is the selection of the cluster center : According to the number of clusters (the effective points are much larger than the noise points), the cluster center of the effective points can be determined, and the effective contour point data can be recorded by mapping to eliminate the contour noise points.

The implementation steps of k-medoids noise preprocessing algorithm proposed in this paper are shown in Figure 5: Step 1 is data acquisition: adjust the camera pose to collect contour circle data; Step 2 is data grouping, and the contour data are evenly divided into three groups in order; Step 3 is data conversion: convert the contour data near the center of the circle through three-point fitting circle; Step 4 is cluster analysis: cluster the transformed data by k-medoids clustering algorithm; Step 5 is the selection of cluster center: the cluster center of effective points can be determined according to the number of clusters (effective points are much larger than noise), and the effective contour point data can be recorded through mapping to eliminate contour noise.

The implementation steps of the K-medoids noise preprocessing algorithm proposed in this paper are shown in Figure 5: step 1 is data collection: adjusting the camera pose to collect contour circle data; Step 2, data grouping, namely, dividing contour data into three groups on average in sequence; Step 3 is data conversion: the contour data is converted to the vicinity of the center of the circle through three-point fitting circle; Step 4 is clustering analysis: clustering the converted data by K-medoids clustering algorithm; Step 5 is the selection of cluster centers: according to the number of clusters (the number of effective points is much larger than the noise), the cluster centers of effective points can be determined, and the data of effective contour points can be recorded by mapping to eliminate contour noise.