Efficient Seed Volume Measurement Framework

Abstract

Modern seed breeding programs require the ability to analyze seeds efficiently to be useful. Even simple measures such as volume and density can be challenging to compute efficiently with modest equipment. Accurately measuring seed volume becomes a highly under-constrained problem. Multiple images from different perspectives are required.This paper presents an efficient and affordable 3D single seed volume measurement system to extract image contours and compute volumes using a modified volume carving method in a controlled lab environment. The framework is constructed with a turntable, a stepper motor controlled by an Arduino microcontroller, three orthogonal cameras, and camera control via a modest computer used for data acquisition and processing. For testing, images are captured using only a side camera from different angles by rotating the turntable. Then, the framework processes the multiple images in parallel and reconstructs 3D seed objects to calculate the volume based on the voxel numbers. The proposed framework: (1) generates single seed 3D geometries for visualization, (2) calculates precise seed volumes within seconds, and (3) achieves less than a 3% error rate on a reference ceramic sphere.