摘要
获取高分辨率的农作物种植结构图对保障粮食安全和优化农业政策具有重要意义。但在实际应用中,利用多源遥感数据提取作物种植结构时,面临光谱分辨率差异导致的星机协同性难题,以及卫星影像中混合像元对面积信息提取的干扰。本研究探索了基于无人机多光谱数据和Planet Scope卫星遥感数据提取作物种植结构的方法,以宝鸡峡灌区的小麦、玉米、葡萄和猕猴桃为研究对象,开展空间分布与面积信息提取研究。本研究首先通过计算卫星与无人机像元的波段反射率比值,校正卫星光谱波段,从而精细化作物分布的提取阈值;其次通过机器学习算法对无人机影像进行分类,获取卫星混合像元中的作物纯净面积比例;最后基于遗传算法优化的随机森林模型,构建混合像元中植被指数与面积权重的量化关系。结果表明,多源遥感方法获取的作物分布图(6 m分辨率)中,重合像元数量较仅使用卫星影像下降35.75%,这一结果说明多源遥感数据可有效缓解“异物同谱”问题。其中,小麦和玉米的面积信息提取效果最优,相较于单时相卫星影像,典型区域中小麦和玉米的面积相对误差率分别降低19.17%和38.49%;整体来看,小麦、玉米、葡萄、猕猴桃的面积相对误差率分别为-4.83%、0.51%、6.55%、8.79%。本研究提出的从无人机到卫星跨尺度协同观测方法,为灌区主要作物种植管理措施制定提供了技术与数据保障,为作物空间分布与面积信息提取从田块尺度向灌区尺度扩展提供新思路,有助于促进灌区的精准农业技术发展。
High-resolution mapping of crop planting structures is crucial for ensuring food security and optimizing agricultural policies.However,in practical applications,extracting crop planting structures using multi-source remote sensing data faces challenges such as satellite—UAV collaboration issues due to spectral resolution differences,and the interference of mixed pixels in satellite imagery during area information extraction.This study proposes a method for extracting crop planting structures based on UAV multispectral data with Planet Scope satellite imagery.Taking wheat,maize,grapes,and kiwifruit in the Baojixia Irrigation District as case crops,spatial distribution and area information were extracted.First,satellite spectral bands were corrected by calculating the reflectance ratio between satellite and UAV pixels,thereby refining the threshold for crop distribution extraction.Second,UAV images were classified using machine learning algorithms to estimate the proportion of pure crop areas within mixed satellite pixels.Finally,a genetic algorithm-optimized random forest model was employed to establish a quantitative relationship between vegetation indices and area weights in mixed pixels.The results showed that in the crop distribution map(6 m resolution)generated using the multi-source remote sensing approach,the number of overlapping pixels decreased by 35.75%compared to results from satellite imagery alone,demonstrating that integrating multi-source data can effectively mitigate the issue of"same spectrum,different objects".Among the crops,wheat and maize had the most accurate area extraction.Compared with single-temporal satellite images,the relative error rates for wheat and maize in representative regions decreased by 19.17%and 38.49%,respectively.Overall,the relative error rates for wheat,maize,grapes,and kiwifruit area estimates were—4.83%,0.51%,6.55%,and 8.79%,respectively.The cross-scale collaborative observation approach from UAV to satellite proposed in this study provides technical and data support for crop management strategies in irrigation districts.It offers new perspectives for extending crop spatial distribution and area extraction from the field scale to the irrigation district scale,and contributes to advancing precision agriculture technologies in irrigated regions.
作者
罗珍
杨妮
尚晓晖
余心城
朱婧怡
杨光
胡笑涛
LUO Zhen;YANG Ni;SHANG Xiao-Hui;YU Xin-Cheng;ZHU Jing-Yi;YANG Guang;HU Xiao-Tao(College of Water Resources and Architectural Engineering,Northwest A&F University,Yangling 712100,Shaanxi,China;Key Laboratory of Arid Land Agricultural Soil and Water Engineering,Ministry of Education,Northwest A&F University,Yangling 712100,Shaanxi,China)
出处
《作物学报》
北大核心
2025年第12期3317-3330,共14页
Acta Agronomica Sinica
基金
国家级创新训练项目(202410712267)
国家自然科学基金项目(U2243235)资助。
关键词
Planet
Scope卫星
无人机
多光谱数据
种植结构
监督分类
遗传算法优化随机森林
planet scope satellite
unmanned aerial vehicle(UAV)
multispectral data
planting structure
supervised classification:random forest optimized by genetic algorithm
