| Multi-type water bodies in the coast are important element in coastal ecosystem and play an important role in coastal socioeconomic development and human life.However,the rapid pace of coastal economic and urbanization has posed a number of threats to the water bodies in coast,including heavy metal pollution,eutrophication,red and green tides,which have a negative impact on the ecological environment and human health.Therefore,the monitoring of multi-type water bodies is crucial to sustainable development of ecological environment in coastal areas.At present,in optical remote sensing,Landsat-8 images have been widely used for mapping coastal water bodies at large scales due to its wide coverage and easy access.However,only based on a single Landsat-8 image,there is still a lack of study on the extraction of multi-type coastal water bodies.This thesis focuses on two aspects on the extraction of coastal multi-water bodies.On the one hand,the thesis studies a multiple features extraction algorithm for four common water bodies(aquaculture ponds,saltpans,rivers and lakes)in sequence.On the other hand,the thesis studies a sequential multiple features extraction algorithm for five water bodies(rice fields,aquaculture ponds,saltpans,rivers and lakes)in a bit complicated environment(such as rice fields with obvious water body characteristics and water photovoltaics).The difficulty of the former lies in the similarity of shape and spectral features among water bodies,which makes it difficult to extract water bodies.The difficulty of the latter not only has the similarity of shape and spectral features,but also has the interferences that affect the extraction performance of water bodies.The main work of this thesis is as follows:(1)A sequential extraction algorithm of coastal multi-water bodies based on multi-features is proposed.For the four common water bodies,in order to improve the extraction performance of the whole water body,the water body index,Hessian matrix and Laplace operator are used to extract the whole water body.In order to extract aquaculture ponds,based on the similarity of local spatial angle and distance of aquaculture ponds,a discriminant criterion based on regional projection is constructed to extract aquaculture ponds.In order to extract saltpans,based on the difference between the spectra of the crystallization and evaporation ponds of the saltpans,and other water bodies,and the spatial distribution characteristics of the saltpans,a criterion based on spectral and spatial search is constructed to extract the saltpans.Finally,by using the difference of the shape characteristics of rivers and lakes,the rivers and lakes are extracted by the proposed roundness,so as to realize the sequential extraction of four kinds of water bodies.In order to verify the effectiveness of the proposed algorithm,Liaoning and Shandong regions are taken as the study areas.By comparing with the existing algorithms,the proposed algorithm has about 5 % improvement in Kappa coefficient and Intersection Over Union(IOU).The experimental results also show that the proposed algorithm can effectively solve the problems caused by the similarity of shape and spectral characteristics among these water bodies.(2)A sequential multi-feature based extraction algorithm of coastal multi-water bodies in a bit complicated coast is proposed.For the five water bodies,in order to reduce the negative impact of urban building roofs and bare land on sequential extraction,the building index and bare land index are used to generate the mask.In order to extract rice fields with obvious water body characteristics,the neighborhood feature rules based on geometric features and vegetation index are constructed to complete the extraction.Then,the aquaculture ponds,saltpans,rivers and lakes can be extracted by(1).Finally,the photovoltaic index is constructed to extract the water photovoltaic,so as to reduce the influence of water photovoltaic on the extraction performance of each water body,so as to complete the sequential extraction of the five types of water bodies.In order to verify the effectiveness of the proposed algorithm,Jiangsu and Guangdong regions are taken as the study areas.By comparing with the existing algorithms,the proposed algorithm has more than 8 % improvement in Kappa coefficient and Intersection Over Union.The experimental results show that the proposed algorithm can effectively solve the similarity problem of shape and spectral features among water bodies in a bit complicated coastal environment,and can reduce the negative influence on the extraction performance of multi-water bodies. |
