Research on apple point cloud segmentation techniques in natural environments
Due to the complexity of apple orchard scenes in natural environments and factors such as changes in illumination,there are challenges in precise positioning and spatial morphological evaluation of fruits for intelligent picking operations.This article proposed a method of apple point cloud segmentation based on multi-dimensional features.This method creates apple point cloud segmentation masks by integrating Euclidean distance,curvature analysis,and color features to segment the apple point cloud.It also introduced K-D Tree for clustering correction to obtain comprehensive spatial information of the final fruit after fitting.Experimental results showed that the segmentation purity rates of this method were 96.20％,97.67％,and 97.93％under backlight,frontlight and sidelight conditions in natural orchard,Compared with the single feature segmentation method based only on Euclidean distance or color features,the purity rate of this method was improved by 37.57％and 14.53％,respectively.And the problem of clustering misclassification has been effectively solved.This algorithm displayed good robustness and accuracy,which can provide technical support for the precision and reliability of apple intelligent picking operations.
binocular camerapoint cloud segmentationrecognition and localizationfeature fusion
万方数据
维普
