Recognition and localization of ratoon rice rolled stubble rows based on monocular vision and model fusion

• DOI: 10.3389/fpls.2025.1533206
• Publication Date: 2025-01-31T06:44:06Z
• Authors: Yuanrui Li; Liping Xiao; Zhaopeng Liu; Muhua Liu; Peng Fang; Xiongfei Chen; Jiajia Yu; Jinlong Lin; Jinping Cai
• Keywords: Monocular

Ratoon rice, as a high-efficiency rice cultivation mode, is widely applied around the world. Mechanical righting of rolled stubble can significantly improve yield in regeneration season, but lack automation has become an important factor restricting its further promotion. In order to realize automatic navigation machine, method fusing instance segmentation model and monocular depth prediction was used localization rows this study. To achieve prediction, estimation trained on training set we made, absolute relative error validation only 7.2%. address problem degradation model's performance when migrated other cameras, based law input image's influence output results, two optimization methods adjusting inputs outputs were that decreased from 91.9% 8.8%. After that, carried out fusion experiments, which showed CD (chamfer distance) between predicted 3D coordinates points obtained by results models labels 0.0990. The point cloud label 0.0174.