Research On Theory And Method Of Sub-pixel Coastline Extraction

A shoreline is defined as the intersection between coastal land and water surface and features seawater edge movements as tides rise and fall,which is the baseline for dividing the administrative region of the ocean and land surface,and is the boundary between the water and the land in the territorial sea.As one of the most important parameters in the marine field,the shoreline plays an important role in marine fisheries,routes,urban expansion,management and other fields.Therefore,the shoreline location dynamic change of the rapid and accurate determination is a technical activity,which studies the interaction between sea and land,coastal reclamation,port development,and is also important content of coastal zone survey and management.Remote sensing has played an important role in the shoreline extraction,remote sensing technology can provide a large range of dynamic monitoring of the shoreline,but the extracted coastline by remote sensing images is the instant boundary of water and land at the time of satellite observation.A variety of remote sensing data sources are available to detect coastline,such as SAR images,LiDAR,and multispectral/ hyperspectral images.However,speckle noise sea surface characteristics cause to have low accuracy about coastline extraction from SAR images.Coastline extracted using LiDAR data source is generally limited in its temporal and spatial availability because of high cost.During coastline extraction multispectral images generate a discrete signal in only a limited number of broadband spectra that contain less spatial information.In contrast with the other remote sensing image types,hyperspectral images contain nearly continuous spectral information and abundant spatial information,giving them a huge potential for distinguishing seawater from other coastal objects;thus,hyperspectral images can greatly improve the accuracy of coastline extraction.However,traditional hard classification methods are performed only at the pixel-level.Each pixel in the image is allocated to the class with which it has the greatest spectral similarity,and the effect of this allocation process is to constrain the prediction of water boundaries,which leads to classification errors and cannot truly reflect coastal surface types of water,vegetation,impervious surfaces and soil,and has the low accuracy of extracting coastline.In order improve extraction accuracy,we used the sub-pixel unmixing method,which allows a pixel to have multiple and partial class memberships and gives an accurate and realistic representation of water,vegetation,impervious surfaces,and soil,thus obtaining a more accurate of the coastline extraction than hard classification techniques.Based on the above research,this paper proposed an automatic sub-pixel coastline extraction(ASPCE)method,which is applied to the research areas of South China Sea,East China Sea,and Bohai Sea,China,respectively.And finally,the results of the coastline extraction by ASPCE method is compared and analyzed based on the original pixel-level and sub-pixel-level contrast methods.The paper has the main contents as the following parts:1.This paper in detail introduced improved W-V-I-S(Water-VegetationImpervious-Soil)model,which was proposed based on the following four indexes: Normalized Difference Water Index(NDWI),Normalized Difference Vegetation Index(NDVI),Normalized Difference Built-Up Index(NDBI),and Normalized Difference Soil Index(NDSI).And the discriminant criterion of extracting mixed pixels from W-V-I-S model was analyzed in detail.So this approach contributed to improving the detection accuracy of mixed W-V-I-S pixels and endmember spectra determination.2.The basic principle of Fully-Constrained-Least-Squares(FCLS)method was introduced in detail,and the two effective constraints of the method were also described in detail: the sum of the endmembers' abundances should be equal to one,and that the angle of endmember abundance should be limited to between 0 and 1.Therefore,this method efficiently calculated abundance of water,vegetation,impervious surfaces and soil,and improved the unmixing accuracy from mixed W-V-I-S pixels.3.According to the calculated abundance of water,vegetation,impervious surfaces and soil based on FCLS,the basic principle of the spatial attraction model was first introduced in detail.And the specific spatial location of the seawater,vegetation,impervious surfaces and soil in the mixed W-V-I-S pixel was analyzed,so the coastline of the research regions of the South China Sea,East China Sea and the Bohai Sea,China was accurately extracted.4.This paper comprehensively analyzed the coastline extraction by automatic sub-pixel unmixing technology.In the South China Sea,East China Sea and Bohai Sea,China experimental regions,the quantitative accuracy results indicated that the proposed method in this paper can accurately extract coastline from the mixed pixels of complex coastal zones and achieve more precise results than based on pixel-level and sub-pixel-level.This method was more suitable for the coastline extraction of complex coastal zones,which can be enable to offer a good foundation for extracting coastline of complex coastal zones in the future.