海岸带浅海水深高光谱遥感反演方法研究

近红外波段(760~900 nm)反射率对水深最为敏感,通过波段比值方法可以提高与水深的相关性,而711nm处反射率一阶微分值与水深的相关系数高达-0.87。对于近岸混浊度高的样本,单波段和比值模型反演效果不好,平均相对误差均高于30%;而光谱微分模型的精度较好,平均相对误差为17%。研究结果证明:水体反射率的一阶微分可以有效地削弱水质变化给水深反演带来的误差。

Study on Shallow Coastal Water Depth Derived Method Based on Hyperspectral Remote Sensing

The theoretical basis on the water depth retrieval by optical remote sensing was analyzed based on the research on radiative transfer process of light wave in water.After a summary of water depth retrieval methods of the predecessors,derivative spectra method for water depth retrieval was introduced.Water reflective spectra were collected by using ASD field spectroradiometer,and water depths were measured by a digital echo sounding system simultaneously at yancheng coastal zone,in Jiangsu Province of China.The turbidity was inhomogeneous in the test area and scattered signal from material in the water was also different in spatial distribution.The reflectance of near infrared band(760-900 nm) was most sensitive to water depth(R=-0.73).As simulated in TM band settings,the correlation between water depth and reflectance ratio between TM4 and TM1 is better than others(R=-0.81).The correlation between water depth and the first derivative of reflectance at 711nm is significant(R=-0.87).Single band method,ratio method and derivative spectra method were respectively applied to calculating water depths.The accuracy evaluation of three models showed that the accuracy of single band model and ratio model were low for the points near the shore,which average relative error was more than 30%.The accuracy of derivative spectra model was improved as to the same points,which average relative error was 17%.The results indicated that derivative spectra method was an effective tool for reducing the error bring by variety of water quality.