A fast and accurate approach to the extraction of leaf midribs from point clouds

摘要

Many studies have obtained high-quality plant point clouds but it remains difficult to extract the midrib from a leaf point cloud. In this study, a fast and accurate approach to the extraction of leaf midrib from point clouds was developed. The leaf was converted into its principal component analysis (PCA) coordinates to find the tip and base points of the leaf. A new search algorithm was proposed, which quickly searched for the approximate shortest curve between tip and base points on the leaf point cloud. The curve then was projected and fitted to eliminate the deviation between the curve and the leaf midrib. Two types of point cloud, generated using the structure-from-motion method and using a laser scanner, were obtained to verify our approach. As a result, the extracted curves were both in good agreement with the leaf midrib for the two kinds of point clouds. Compared with manual measurements of the leaf length, the root-mean-square error (RMSE) of the lengths of the two types of the extracted curve were 2.55 mm and 1.38 mm, respectively. The result shows that our method is robust and practical and may assist in the development of plant morphology measurements.