文献类型：期刊文献

英文题名：Image-Based, 3D reconstruction for smart farming Applications: A review

作者：Wang, Haihua[1,2,3,4,5,6]; Zhao, Jiayin[1,6]; Zhang, Shulong[1,6]; Yao, Mingyuan[1,6]; Xue, Lin[7]; Li, Daoliang[1,2,3,4,5,6]

第一作者：Wang, Haihua

机构：[1] National Innovation Center for Digital Fishery, Beijing, 100083, China; [2] Key Laboratory of Smart Farming Technologies for Aquatic Animal and Livestock, Ministry of Agriculture and Rural Affairs, Beijing, 100083, China; [3] State Key Laboratory of Efficient Utilization of Agricultural Water Resources, China Agricultural University, Beijing, 100083, China; [4] Beijing Engineering and Technology Research Center for Internet of Things in Agriculture, Beijing, 100083, China; [5] Key Laboratory of Digital Fisheries of Shandong Province, Shandong, Yantai, 264670, China; [6] College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, China; [7] Smart City College, Beijing Union University, Beijing, 100101, China

第一机构：National Innovation Center for Digital Fishery, Beijing, 100083, China

通讯机构：[1]National Innovation Center for Digital Fishery, Beijing, 100083, China

年份：2026

卷号：248

外文期刊名：Computers and Electronics in Agriculture

收录：EI(收录号：20261820606650);Scopus(收录号：2-s2.0-105036997037)

语种：英文

外文关键词：3D reconstruction - Agricultural robots - Decision making - Deep learning - Image reconstruction - Precision agriculture - Stereo image processing - Three dimensional computer graphics

摘要：As a core driving force in the digital transformation of smart farming, 3D reconstruction technology has recently been shown to have substantial potential for agricultural environmental perception, precision operations, and decision-making support. This review summarizes recent advances in image-based 3D reconstruction from both methodological and application perspectives. Conventional methods, including voxel carving, Structure from Motion and Multi-View Stereo (SfM-MVS), and photometric stereo, are first reviewed, followed by recent developments in deep learning–based reconstruction, Neural Radiance Fields (NeRF), and 3D Gaussian Splatting (3DGS). These methods are discussed in relation to the major challenges of agricultural environments, including occlusion, heterogeneous illumination, and target motion. Representative applications in crop phenotyping, robotic perception and operation, and livestock monitoring are then analyzed, with comparison of technical systems in terms of accuracy, cost, and adaptability. Finally, current limitations related to real-time performance, robustness, and system integration are discussed, and future directions in lightweight modeling, multimodal data fusion, and edge deployment are outlined. ? 2026 Elsevier B.V.