应用GASVM的城市景观水体富营养化多光谱遥感监反演v1.docVIP

应用GA-SVM的城市景观水体富营养化多光谱遥感监测反演 霍艾迪1王菊翠1 张学真1 （1长安大学环境科学与工程学院，陕西 西安 710018） 摘 要: 建立了基于支持向量机的遥感水质参数反演模型。基于多光谱遥感数据和水质实地监测数据，以西安市景观水体为例进行了遥感反演，并与水质遥感反演的多元线性回归法、神经网络法和GA优选SVM参数法进行了分析比较。实验结果表明，在样本数目有限的情况下, 提出的GA-SVM方法的反演结果比神经网络和传统的多元回归方法模拟结果均好, 且神经网络方法比多元回归法结果稍好。SVM 具有强的非线性映射能力,适合小样本情况, 由GA实现了模型参数的自动优选, 使GA-SVM用于解决回归问题表现出优势。因此，将机器学习和全局优化智能计算方法引入, 使得GA-SVM为城市景观水体遥感监测提供了一种新方法, 取得了较好的反演结果。 关键词: 支持向量机, 遗传算法, 水质参数, 遥感反演 中图分类号: X832/TP79 文献标识码: A Multispectral Remote Sensing Inversion for City Landscape Water Eutrophication based on GA – SVM Aidi HUO1 Jucui WANG1 Xue-zhen ZHANG1 (1 School of Environment Science and Engineering, Chang’an University, Xi’an 710054,China) Abstract: The paper establishes an model for Water Quality remote sensing Retrieving of the Urban Landscape Water based on Support Vector Machine (SVM). The model is experimented with monitoring water quality data and multispectral remote sensing data of Landscape Water in xi’an city, and is compared with water quality Retrieving methods of Multiple regression and BP neural network. The results demonstrate that GA- SVM (Genetic algorithm) method is better than the inversion results of the neural network and the traditional statistical regression method, and the method of multiple regression results better than a dollar regression results. SVM has strong nonlinear mapping ability, and SVM is suitable for the limited number of samples, model parameters has realized automatic optimization choice by GA, let GA - SVM show advantage to solve regression problems. The machine learning and global optimization intelligent computation methods is introduced, GA- SVM provides a new method for Remote Sensing Monitoring and Application of the Urban Landscape Water Eutrophication and has made the good inversion results Key words : SVM; Genetic algorithm; water quality; Remote sensing 1 引言 水体富营养化是指水中氮、磷浓度达到一定阈值，促使浮游性藻类大量繁殖、水色发暗、透明度降低、水体溶解氧下降、水质恶化、水体功能衰退的水环境污染现象[1]。 常规的水体富营养化监测是通过采集水样、过滤、萃取以及分光光度计分析得出，这种方法耗时、耗力、耗钱，且只能了解监测断