基于OLI多光谱遥感影像的八所港浅海水深反演

Shallow Water Depth Extraction from OLI Remote Sensing Image in Basuo Port

为了解决传统全局模型应用遥感影像反演水深的不足,提出地理自适应模型应用于八所港Landsat-8 OLI多光谱卫星影像。采用归一化水体指数进行水陆分离,在产生模拟数据时,为保证数据的合理性,对光谱灰度值进行自然对数求解。地理自适应模型将整个区域细分为5个小的区域单元,模型参数是自适应变化的,一般数学形式与传统全局模型一样。通过反演15m以浅的水深,发现光谱中对水体的敏感波段出现"红移";水深反演结果证明地理自适应模型有效地缓和了全局传统模型在底质类型和水体性质空间不均匀的问题,水深反演精度得到明显提高,并控制在1m以内。

In order to address the inadequacy of conventional global inversion models in remote sensing image, a geographically adaptive inversion model is proposed for retrieving bottom depth values from Landsat-8 OLI multispectral image over Basuo Port. Water-land separation is carried out by using the normalized difference water index.In the process of generating simulation data,in order to ensure the rationality of the data,the natural logarithmic solution of spectral gray value is used. The model subdivides the image scene into five geographical regions, and model parameters are adaptively changed, but the general mathematical form is same as the traditional global model. Through the inversion of the shallow water around 15m depth, it is found that the sensitive band of water in the spectrum is redshift.The result demonstrates that the adaptive model effectively mitigates the spatial heterogeneity problem of sediment and water quality, and the accuracy of water depth inversion is obviously improved to one meter level.