Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract. Tree architecture is indicative of the apple tree growth and apple crop load potential. While many studies have widely used 2D imaging to measure architectural traits of tree crops, the measurement accuracy was largely limited by occlusion and scene ambiguity due to a single perspective projection in 2D images. The goal of this study was to develop a 3D point cloud-based approach for apple tree segmentation and architecture characterization. Specifically, a biological-constrained graph-based algorithm was developed to segment the trunk and branch and characterize the architectural phenotypic traits of apple trees. A Laplacian-based skeletonization algorithm was used to extract a curve skeleton from the 3D point cloud and the topological structure of the skeleton was optimized by a weighted minimum spanning tree algorithm. The trunk was segmented from the skeleton by searching for an optimal path starting from the root point and branches were divided into groups by 3D DBSCAN clustering algorithm. Tree height and trunk diameter were estimated based on the geometric information of the tree trunk. Quantitative branch patterns were revealed by estimating the branch length, branch diameter, and branch angles. Point cloud data of 9 apple trees were collected under the field condition using a terrestrial LiDAR at Cornell Orchard in Ithaca, NY, USA. Preliminary experiments showed that the proposed algorithm achieved a higher accuracy for measuring morphological traits. The extracted traits can be used for monitoring apple tree growth and apple crop load estimation.