ection, some unremoved husks were counted as seeds by the counting machine despite manual efforts to separate seeds from husks. We expect the effect on the ground truth to be small. The stereo images, camera poses, human-labeled seed segmentations, panicle weights, and human-counted seed counts can be found in our dataset 3 3 3 https://labs.ri.cmu.edu/aiira/resources/ . Figure 8: (a) 100 sorghum panicles from 10 different sorghum species. (b) Our data collection system, a stereo camera attached to the UR5 robot arm. (c) Seeds were manually stripped and (d) counted using a seed counting machine. IV-B 3D Reconstruction Quality We assess the effectiveness of our approach with ablation tests usi
Open resource ↗lines:141-165Paper record
3D Reconstruction-Based Seed Counting of Sorghum Panicles for Agricultural Inspection
arXiv · 14 Nov 2022 · 10.48550/arxiv.2211.07748
Abstract
In this paper, we present a method for creating high-quality 3D models of sorghum panicles for phenotyping in breeding experiments. This is achieved with a novel reconstruction approach that uses seeds as semantic landmarks in both 2D and 3D. To evaluate the performance, we develop a new metric for assessing the quality of reconstructed point clouds without having a ground-truth point cloud. Finally, a counting method is presented where the density of seed centers in the 3D model allows 2D counts from multiple views to be effectively combined into a whole-panicle count. We demonstrate that using this method to estimate seed count and weight for sorghum outperforms count extrapolation from 2D images, an approach used in most state of the art methods for seeds and grains of comparable size.
Code and data availability