Traditional plannar images lack depth information and are challenging for accurate measurement of morphological variables such as volume and deformity index of apples. This study presents a new evaluation of a custom-assembled binocular structured light system, combined with a three-dimensional (3D) reconstruction approach, for accurate estimations of the morphological variables of apples. The system was mainly composed of a digital projector for generating structured patterns and two cameras in a symmetric arrangement. The 3D reconstruction accuracy was verified by standard samples, with the relative errors concentrating within 3%, and most were within 1% (depth error of 0.5 mm). Multi-view registration based on a rotary table was used for restoring more complete 3D information of apples. A method based on the normal vector of the point cloud was presented to complement the bottom of the apple and extract the shoulders. 3D reconstruction of 82 apples with different morphologies was performed, from which morphological variables of deformity index, the maximum diameter, fruit shape index, volume and mass were estimated. The values between the actual measurements and the results obtained from the 3D point cloud were 0.9941, 0.9594, 0.8015, 0.9915 and 0.9960, respectively. This study showed that high-accuracy measurement of apple morphological variables can be achieved by binocular structured light-based 3D reconstruction, which would be beneficial for subsequent apple grading and phenotyping analysis.

中文翻译：

---

基于双目结构光的 3D 重建用于苹果形态测量

传统的平面图像缺乏深度信息，对于准确测量苹果的体积和畸形指数等形态变量具有挑战性。这项研究提出了对定制组装的双目结构光系统的新评估，结合三维 (3D) 重建方法，可准确估计苹果的形态变量。该系统主要由一个用于生成结构化图案的数字投影仪和两个对称排列的摄像机组成。 3D重建精度经标准样品验证，相对误差集中在3%以内，大部分在1%以内（深度误差0.5毫米）。采用基于转台的多视图配准来恢复更完整的苹果3D信息。提出了一种基于点云法向量的方法来补充苹果的底部并提取肩部。对82个不同形态的苹果进行3D重建，估算畸形指数、最大直径、果形指数、体积和质量等形态变量。实际测量与 3D 点云结果之间的值分别为 0.9941、0.9594、0.8015、0.9915 和 0.9960。本研究表明，基于双目结构光的3D重建可以实现苹果形态变量的高精度测量，这将有利于后续苹果的分级和表型分析。