农业机械学报2025，Vol.56Issue(10)：429-436,8.DOI:10.6041/j.issn.1000-1298.2025.10.037
基于多源遥感数据的林农间作种植结构精细分类
Fine Classification under Pattern of Forest and Crops Intercropping Based on Multi-source Remote Sensing Data
摘要
Abstract
The pattern of forest and crops intercropping,as a characteristic planting model in Southern Xinjiang,is of great significance for improving agricultural production efficiency and optimizing resource utilization.Focusing on the precise classification of forest-crop intercropping land in Lopu County,Hotan Region,Xinjiang,based on multi-source and high-resolution remote sensing images,the hybrid task cascade(HTC)instance segmentation model was applied to extract the precise boundaries of farmland parcels.Meanwhile,multi-temporal Sentinel-2 remote sensing data were utilized to calculate spectral characteristic indices of normalized difference vegetation index(NDVI)and normalized difference red-edge 1(NDre1)during the key reproductive phases of crops.The Transformer temporal model then accurately extracted the information of the forest and crops intercropping planting structure.The distribution of typical forest-food intercropping(walnut and maize)and forest-vegetable intercropping(walnut and radish)patterns in Luopu County was analyzed.The results showed that the overall accuracy(OA)of crop planting structure classification with parcel as the basic unit reached 83.2％.Among them,the forest-food intercropping pattern in Luopu County was dominant,with a classification precision of 78.4％and a total planting area of 274.85 km2,accounting for 64.5％of the total area of extracted farmland parcels.The forest-vegetable intercropping pattern was only 15.55 km2,and its classification accuracy was as high as 96.5％,which was usually manifested as scattered small parcels.The research result can provide a method for the fine identification of forest-crop intercropping planting structures,which was of great significance in guiding the precision management of agriculture in Southern Xinjiang.关键词
林农间作/种植结构/精细分类/HTC实例分割/Transformer模型/多源遥感数据Key words
forest and crops intercropping/planting structure/fine classification/HTC instance segmentation/Transformer model/multi-source romote sensing data分类
农业科学引用本文复制引用
张依涵,沈占锋,寇雯齐,王浩宇,张驰,马于博..基于多源遥感数据的林农间作种植结构精细分类[J].农业机械学报,2025,56(10):429-436,8.基金项目
新疆第三次科学考察项目(2021xjkk1403)和新疆维吾尔自治区重点研发任务专项(2022B03001-3) （2021xjkk1403）
