Contour extraction of artificial ski runs based on composite supervoxels

Abstract

Increasing numbers of people have taken up skiing in recent years due to the strong promotion of the Beijing Winter Olympics in 2022 by We Media. Nowadays, the establishment of three-dimensional digital twin snow fields has become an important way to effectively manage and maintain snow fields. However, due to the large altitude difference and the presence of many trees and rocks in ski resorts, traditional methods face difficulties such as low segmentation accuracy and low merging efficiency of segmentation blocks when performing semantic segmentation of ski runs. Consequently, this paper proposes a contour extraction method of artificial ski runs using composite supervoxels based on the characteristics of artificial ski resorts. To begin with, the point cloud data set of ski resort is segmented to get supervoxels; secondly, the difference in elevation between the seed supervoxel and the adjacent connecting block is calculated to determine whether the merging plane is the ground or another plane; then, according to the normal vector angle threshold and the orthogonal distance threshold, the similarity between the current clustering region and adjacent blocks is evaluated; and finally, the region growth algorithm is optimized based on the point cloud supervoxels of ski resorts, in order to reap the benefits of ski track semantic segmentation. And experiments have shown that the proposed method is superior to the other two in terms of segmentation accuracy, efficiency, and robustness, and is suitable for the segmentation and extraction of ski tracks in complex scenes, such as artificial snow fields.