基于多光谱遥感图像信息的水质污染监测研究

Research on Water Quality Monitoring Based on Multi-spectral Remote Sensing Imagery

在国内陆水资源保护与污染治理事业中,水资源污染监测是其中的核心环节。传统水质监测方法存在需要消耗大量人力物力和时空局限性的问题。随着遥感技术和成像光谱仪的发展,以及机器学习相关领域的快速发展,提出一种基于极限学习机的高分图像水质监测方法。对高分一号卫星所检测到的多光谱水域图像数据信息进行处理,并结合部分时间段的水质检测站实测数据对处理后的WFV数据进行标记,搭建基于极限学习机模型对水域进行异常检测,定性地判断该水域是否污染,通过对太湖兰山嘴和洪泽湖盱眙淮河大桥区域的水质污染判定,结果表明该方法对水质监测具有一定的可行性和有效性。

In China’s inland water resources protection and pollution control business,water pollution monitoring is one of the core links.Traditional water quality monitoring methods have the problem of consuming a great deal of manpower,material resources and space-time limitations.With the development of remote sensing technology and imaging spectrometers,and the rapid development of machine learning related fields,we propose a GF-1 image water quality monitoring method based on extreme learning machine.The processing of multi-spectral water image data detected by GF-1 satellite is carried out,and the processed WFV data is marked by using the measured data of water quality detection station in some time periods,and then the water quality is determined based on extreme learning machine model.The results show that this method is feasible and effective for water quality monitoring of Taihu Lanshanzui and Hongzehu Xuyi Huaihe River Bridge.