Measurement of Potato Volume With Laser Triangulation and Three-Dimensional Reconstruction

Abstract

Potato grading is related to weight. Three-dimensional (3D) reconstruction can provide highly accurate volume measurements of potatoes, which can help farmers to analyze their phenotypic characteristics and grade them. Considering their low cost and the required accuracy, a monocular camera and line laser were used to build a potato phenotype determination scanning device. The system obtains coordinates along the surface of a potato, collects laser light reflected from the surface in real time, and completes the coordinate calculation of the original points using the triangulation method. However, the original point clouds lose large areas of point clouds at the top and bottom of the potato. Point cloud repair is carried out by interpolation of points. In addition, the surface point cloud is smoothed. Finally, the generated point cloud is used for 3D reconstruction and volume calculation. In a volume error analysis test, potatoes are divided into calibration and verification groups. First, linear regression is used to relate the real and measured potato volume, and then the density of the potato is calculated. The volume and mass of potatoes in the verification group are measured by the device, and the standard volume and mass are measured manually. The results show that the average relative error in measured volume is −0.08%, and the average relative error in estimated mass is 0.48%. These results indicate that the combination of a line laser and a single camera provides accurate measurements of potato volume that can be used for yield estimation and potato grading.