A Novel Method for Filled/Unfilled Grain Classification Based on Structured Light Imaging and Improved PointNet++

China is the largest producer and consumer of rice, classification filled/unfilled rice grains great significance for breeding genetic analysis. The traditional method grain identification was generally manual, which had disadvantages low efficiency, poor repeatability, precision. In this study, we have proposed a novel based on structured light imaging Improved PointNet++. Firstly, 3D point cloud data were obtained by imaging. And then specified processing algorithms developed single segmentation, enhancement with normal vector. Finally, PointNet++ network improved adding an additional Set Abstraction layer combining maximum pooling vectors to realize classification. To verify model performance, compared six machine learning methods, PointNet PointConv. results showed that optimal XGboost, accuracy 91.99%, while 98.50% outperforming 93.75% PointConv 92.25%. conclusion, study has demonstrated effective recognition.