土壤含盐量BP神经网络反演模型  

Inversion Model of BP Neural Network for Soil Salinity

作　　者：刘丹丹[1] 王强[1] 

机构地区：[1]黑龙江工程学院,哈尔滨150050

出　　处：《东北林业大学学报》2009年第12期88-90,共3页Journal of Northeast Forestry University

基　　金：黑龙江省骨干教师基金项目(1153G029);黑龙江工程学院省级重点实验室(空间地理信息实验室)开放研究基金项目

摘　　要：运用Hyperion数据,以黑龙江省大庆市某一实验区为例,通过对图像预处理、特征提取、土壤含盐量、波段与土壤含盐量的相关性分析,建立BP神经网络模型(Back Propagation Network)、经验统计模型等进行研究,并开展了对土壤含盐量的定量提取研究,探讨Hyperion数据反演土壤含盐量的方法。结果表明:与传统的经验统计模型相比,BP神经网络模型具有不可比拟的优越性;同时,Hyperion数据为建立土壤含盐量模型提供了高维的输入样本,大大提高了反演的精度;土壤含盐量的反演模型的研究还有待于进一步深入。An experimental area in Daqing City, Heilongjiang Province is taken as an example to perform a quantitative reversion of soil salinity. The inversion method of soil salinity using Hyperion data is discussed by image preprocessing, feature extraction, laboratory analysis of soil salinity, relativity analysis of wave band and soil salinity, establishment of BP neural network model （Back Propagation Network） , empirical statistical model, and so on. Results show that BP neural network model has incomparable superiority compared with the traditional empirical statistical model. Moreover, Hyperion data provide a high-dimensional input sample for the establishment of soil salinity model, which greatly improves the accuracy of inversion.

关 键 词：土壤含盐量 HYPERION数据 反演 BP神经网络模型 

分 类 号：TP7[自动化与计算机技术—检测技术与自动化装置] S15[自动化与计算机技术—控制科学与工程]