基于深度学习的农业区土地利用无人机监测分类

Rapid monitoring and classification of land use in agricultural areas by UAV based on deep learning method

农业种植区土地利用快速监测与分类对政府部门制定规划、土地资源管理、生态环境保护规划与决策以及 农业旱情与旱灾动态监测评估具有重要意义。本研究以东雷二期抽黄灌区具有下垫面代表性的小区域为研究区, 利用卷积神经网络深度学习方法,针对较高空间分辨率的无人机航片影像,开展了农业区土地利用监测分类研 究,并与最大似然法进行比较,探究该方法对于农业区土地利用监测分类的适用性。结果表明,该方法优于最大 似然法,其总体分类精度达93%以上,Kappa系数为0.9以上,能够更清晰地识别提取出地物边界,分类效果较 好。本研究有助于提升应急抗旱减灾工作对农业区土地利用的快速监测与分类能力,为旱情与旱灾快速监测评 估、决策提供技术支持,同时能够及时为政府、土地资源管理以及生态环境保护规划等部门提供基础数据。

The rapid monitoring and classification of land use in agricultural areas is important for land re?source management,ecological environmental protection planning and decision-making,government depart?ment planning,the dynamic monitoring and assessment of agricultural drought.Based on the remote sensing images of the unmanned aerial vehicle(UAV)with higher spatial resolution,we chose small regions with the underlying surface representation of the Donglei Irrigated District(PhaseⅡ)as the research area to carry out the monitoring of land use classification research in agriculture areas using the convolutional neural network(CNN)deep learning methods.In addition,the method is compared with the maximum like?lihood method to explore the applicability of the method for land use monitoring classification in agricultur?al areas.The results show that the classification accuracy of this method is better than the maximum likeli?hood method.The overall classification accuracy is over 93%,and the Kappa coefficient is above 0.9.The boundary of the extracted features can be clearly identified and the classification effect is better.This study can help improve the rapid monitoring and classification of land use in agricultural areas during emergency drought and mitigation,and provide technical support for rapid monitoring and assessment of drought,which will provide basic data for the government,land resource management and ecological environmental protection planning in a timely manner.