Research On Water Body Extraction Method Of Domestic Gao Fen Series Remote Sensing Images

As one of the important features on the surface of the earth,water body play a vital role in the development of society and economy.Therefore,the use of domestic Gao Fen satellite images to grasp and acquire water body information in a timely manner has become one of the hot research topics of the domestic Gao Fen satellites.At present,the research on water extraction in the existing literature is mainly carried out on foreign medium and high resolution images,and most of the research methods are specific to specific images and specific regions,and the extraction methods are not universal.With the successive launch of domestic Gao Fen series of remote sensing satellites,the systematic research on the extraction of water information from domestic Gao Fen satellite images is relatively lagging behind.The main reasons are:(1)Some domestic Gao Fen satellite imaging physical models are not perfect.There are few imaging bands and insufficient spectral information.(2)In the multi-spectral image,the mid-infrared band commonly used in water body modeling is lacking,and the water body information is easily interfered by ground shadows,dark features,et al.In view of the above problems,this paper uses the threshold method and the classifier to carry out the extraction experiment of the water information of the two selected research areas,using the domestic GF-1,GF-2,GF-4 image data.The results are analyzed and compared.For the GF-2 image,an enhanced Shadow Water Index(ESWI)model is proposed and a decision tree method is constructed by combining the model with the single-band threshold method.The method can effectively distinguish the shadow and water body information and improve the water body extraction accuracy.Finally,the effects of the extraction accuracy and stability of water bodies in different regions are analyzed and compared,which is caused by the selection of the threshold value in the GF-2 image threshold method.The main conclusions of the thesis research are as follows:(1)It is quick and simple to use the threshold method to extract water.However,it is different threshold methods,different regions,different resolution images and so on that will lead to differences in water extraction threshold and extraction results,and the same extraction method will have different effects in different research areas.The classifier rule requires a certain amount of processing time,and the process is relatively complicated.(2)The extraction effect of different resolution images is different,and the effect of water body information presented by images with high spatial resolution is not better.The image with high spatial resolution has more noise than the image with low spatial resolution,and the details of the background object will be enlarged.The shadow blocking effect of the mountain and the building is more obvious,which will have certain interference on the water information extraction.In practical applications,different extraction methods and images with different resolutions should be selected according to the different uses of water extraction.(3)An enhanced shadow water index(ESWI)model is proposed for GF-2 images,and a decision tree method is constructed by combining this model with the single-band threshold method.Experiments show that the method can effectively distinguish the shadow and water body information,and to some extent,can identify the water body information of the shaded area.Compared with water bodies in complex terrain areas,this method is more suitable for areas with single objects and large water bodies.(4)The selection of threshold values for different water body extraction models has different effects on water extraction accuracy and stability.The accuracy of the water body in the complex terrain area of the GF-2 image is NIR,NDWI,MSWI,NDVI,SWI from high to low,that is when the threshold values of the above models are changed by the same amplitude,the NIR method has the least precision change,and the SWI method has the larger precision change.In the region where the ground object is single and the water body area is large,the stability from high to low is NIR,MSWI,NDWI,SWI,NDVI.