摘要
获取实时、精准的耕地分布信息是现代土地资源管理和农业高质量发展中至关重要的一项任务。随着卫星技术的迅猛发展,遥感监测逐渐成为当前耕地信息提取的重要手段,同时,深度学习技术迅速崛起,并逐渐成为遥感影像耕地提取的关键技术。本文整理了国内外近期耕地提取的相关研究成果,阐述了传统提取算法的不足及高分辨率遥感图像对耕地提取的积极意义、耕地提取的基本流程、耕地提取算法发展的主要过程和研究策略,归纳了耕地提取算法的主要优化方法以及多任务网络模型的应用,最后结合现有深度学习算法存在的不足对未来耕地提取技术发展趋势进行了展望。
Obtaining real-time and accurate information on cropland distribution is a critical task in modern land resource management and high-quality agricultural development.With the rapid advancement of satellite technology,remote sensing monitoring has gradually become a key method for extracting cropland information.At the same time,deep learning technology has risen rapidly and is increasingly becoming a pivotal technique for cropland extraction from remote sensing imagery.This paper consolidates recent research findings on cropland extraction both domestically and internationally,elaborates on the limitations of traditional extraction algorithms,and highlights the positive significance of high-resolution remote sensing images for cropland extraction.It outlines the fundamental processes of cropland extraction,reviews the primary development stages and research strategies of cropland extraction algorithms,summarizes the main optimization methods for such algorithms,and explores the application of multi-task network models.Finally,it discusses the limitations of current deep learning algorithms and anticipates future trends in the development of cropland extraction technologies.
作者
巫志雄
李乔宇
王宗良
曾世伟
Wu Zhixiong;Li Qiaoyu;Wang Zongliang;Zeng Shiwei(School of Physical Science and Information Technology,Liaocheng University,Liaocheng 252000,China;Institute of Information and Economic Research,Shandong Academy of Agricultural Sciences,Jinan 250100,China)
出处
《山东农业科学》
北大核心
2024年第12期163-170,共8页
Shandong Agricultural Sciences
基金
国家重点研发计划项目(2021YFB3901300)
山东省农业科学院农业科技创新工程项目(CXGC2023D02)。
关键词
耕地提取
深度学习
语义分割
高分辨率影像
Cultivated land extraction
Deep learning
Semantic segmentation
High resolution image
