Classification of rice seed variety using point cloud data combined with deep learning

Rice variety selection and quality inspection are key links in rice planting. Compared with two-dimensional images, three-dimensional information on seeds shows the appearance characteristics of more comprehensively accurately. This study proposed a classification method using point cloud data surface combined deep learning network to achieve rapid accurate identification varieties. First, collection platform was set up Raytrix light field camera as core collect seeds; then, collected filled, filtered smoothed; after that, segmentation is based RANSAC algorithm, downsampling combination random sampling algorithm voxel grid filtering algorithm. Finally, processed input improved PointNet for feature extraction species classification. The added cross-level connection structure, made full use features at different levels, better extracted structure seeds. After testing, model had an average accuracy 89.4% eight varieties rice, which 1.2% higher than that model. this efficient Keywords: seed, classification, data, learning, DOI: 10.25165/j.ijabe.20211405.5902 Citation: Qian Y, Xu Q J, Yang Y Lu H, Li Feng X B, et al. Classification seed learning. Int J Agric & Biol Eng, 2021; 14(5): 206–212.