Comparison of 3D Rice Organs Point Cloud Classification Techniques

The performance of phenotypic evaluation is a hurdle in the rice breeding process. Because of manual evaluation, the traditional evaluation procedure is very slow. Therefore, this is the primary impediment to the development of new rice varieties. The focus of this research is to compare techniques for categorizing plant organs using 3D point clouds so that plant morphology and structure can be analyzed automatically. A 3D laser scanner was used to scan the rice plant and create the rice plant point cloud. The point cloud of the rice plant contains the rich information due to the enormous points in the cloud cluster. Therefore, the number of points in each cloud must be minimized via skeleton preprocessing technique before can be used for training the network. The skeleton approach is used to thin the cloud while keeping the vital information intact. We suggested two approaches for classifying rice organs in this study. Tensor-based classification is the first method. The second technique is PointNet. In comparison, the PointNet was the most successful for classifying rice organs and spent the least amount of time.