基于LM算法的土壤表层含水率遥感监测  

Remote Sensing Monitoring of Soil Surface Moisture Content Based on LM Algorithm

作　　者：许景辉 王雷[2] 王一琛[2] 赵钟声 韩文霆 XU Jinghui;WANG Lei;WANG Yichen;ZHAO Zhongsheng;HAN Wenting(Key Laboratory of Agricultural Soil and Water Engineering in Arid Areas,Ministry of Education,Northwest A&F University,Yangling,Shaanxi 712100,China;College of Water Resources and Architectural Engineering,Northwest A&F University,Yangling,Shaanxi 712100,China;Institute of Water-saving Agriculture in Arid Areas of China,Northwest A&F University,Yangling,Shaanxi 712100,China)

机构地区：[1]西北农林科技大学旱区农业水土工程教育部重点实验室,陕西杨凌712100 [2]西北农林科技大学水利与建筑工程学院,陕西杨凌712100 [3]西北农林科技大学中国旱区节水农业研究院,陕西杨凌712100

出　　处：《农业机械学报》2019年第6期233-240,共8页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家重点研发计划项目(2017YFC0403203);陕西省水利科技计划项目(2014slkj-18);中央高校基本科研业务费专项资金项目(2452015050)

摘　　要：为探讨数据挖掘技术中LM(Levenberg-Marquardt)算法在土壤表层(约1 cm)含水率遥感监测中的应用,选取黄绵土、粘黄土、红土为试验材料,配制含水率分别为0、6%、10%、14%、18%、22%的土壤样本,在09:00—10:00和15:00—16:00时间段进行可见光采样,并对图像亮度进行梯度处理,以此模拟全天光线变化。采用样本实测含水率及图像RGB三阶颜色矩数据作为数据集,对上午、下午样本和两时间段混合样本采用LM算法建立含水率回归模型,并与BP(Back propagation)算法和分类回归树(Classification and regression trees,CART)算法进行比较。结果表明,基于土壤表层RGB颜色矩的LM算法具有较好的应用效果,混合样本不同土样回归模型决定系数 R^2分别为0.958、0.943、0.949,均方根误差(RMSE)分别为1.6%、2.0%、1.9%,相对分析误差(RPD)分别为4.873、4.183、4.440。不同光照时的混合样品分析结果表明,LM算法适用于不同光线采集样品的土壤含水率监测,适用于土壤表层(约1 cm)含水率的监测。Aiming to research the data mining technology in remote sensing monitoring. The LM algorithm was used in the soil surface layer( about 1 cm) of soil moisture measurement( soil moisture content,SMC). Three kinds of soil,including yellow spongy,loess and red clay were selected. The soil water content samples of 0,6%,10%,14%,18% and 22% were prepared respectively. Visible light images were taken during 09: 00-10: 00 and 15: 00-16: 00,and image brightness was gradiently processed for simulating the change of light throughout the day. LM algorithm was compared with back propagation( BP) algorithm and classification and regression trees( CART) algorithm to verify the practical effect of LM. It was showed that the LM algorithm had a good application effect for the data mining based on the RGB color moment of soil pictures. The determination coefficient of the regression model for yellow spongy,loess and red clay was 0. 958,0. 943 and 0. 949,root mean square error( RMSE) was 1. 6%,2. 0% and 1. 9%,and the relative analysis error( RPD) was 4. 873,4. 183 and4. 440,respectively. By the study of pictures at different intensities,LM algorithm can be used for monitoring soil moisture content of samples. It can be used for the soil surface( about 1 cm) moisture content measurement.

关 键 词：土壤含水率 数据挖掘 LM算法 遥感 监测 

分 类 号：S152.7[农业科学—土壤学]