现代农业装备2025，Vol.46Issue(6)：31-38,8.DOI:10.3969/j.issn.1673-2154.2025.06.005
非结构化农林环境下的三维语义地图构建方法
Robust 3D Semantic Mapping for Unconstructed Outdoor Agricultural and Forestry Environment
摘要
Abstract
Mobile manipulation tasks in unstructured outdoor agricultural and forestry environments—such as transportation,fruit picking,weeding and spraying—are increasingly recognized as requiring environmental models that extend beyond conventional geometric and appearance-based representations.Instead,3D models enriched with semantic information are required to support autonomous task planning and precise execution.However,most existing agricultural mobile robots have been developed with a strong reliance on low-level geometric features or single-modality perception,leaving systematic semantic modeling capabilities largely absent and robust perception under complex and dynamic conditions unachieved.To overcome these limitations,a vision-LiDAR fusion-based 3D point cloud segmentation method was proposed.In this approach,visual semantic segmentation results were projected onto point clouds,and semantic labels are automatically propagated in 3D space through geometric consistency.Building on this method,a 3D semantic mapping system tailored for large-scale dynamic environments was constructed,enabling the integration of semantic perception with spatial geometric modeling.Field experiments demonstrated that robust performance were maintained under unstructured conditions and dynamic interference,so as the real-time semantic segmentation,pose estimation,and mapping had been achieved,thereby validating the effectiveness and application potential of the proposed method in complex outdoor agricultural and forestry scenarios.关键词
语义地图构建/自主导航/非结构化环境Key words
semantic mapping/autonomous navigation/unconstructed environment分类
信息技术与安全科学引用本文复制引用
ZENG Jun,XIE Zhaopai,ZHANG Hongwei,ZHANG Linlin,LI Mutong..非结构化农林环境下的三维语义地图构建方法[J].现代农业装备,2025,46(6):31-38,8.基金项目
广东省教育厅普通高校重点领域专项(服务"百千万工程")(2024ZDZX4103) （服务"百千万工程"）
广州市教育局产学研结合项目(202235425) （202235425）
