Study On Extraction Method Of The Shoreline Using Remotely Sensed Imagery

The shoreline is the site where the land and the ocean meet each other.It is one of the important signs of the coastal zone.Its analytical techniques and applications are of great significance in marine scientific research,production activities and national defense construction.With the extensive application of satellite data,the extraction of shoreline based on remote sensing image has gradually become the mainstream method of interpreting shoreline.In this paper,the remote sensing image is used as the data source to improve the extraction accuracy of the shoreline.Based on the actual requirements and application conditions,the paper studies and analyzes the shoreline extraction methods of the existing two categories(according to whether the mixed pixel problem is considered),and points out their own shortcomings and disadvantages.The improved algorithms are proposed.Specifically,there are two main aspects of the work:1.Regardless of the effect of mixed pixels on the extraction accuracy of the shoreline,the shoreline extraction algorithm based on the KPCA spectral feature constraint(SEKSFC)is proposed to solve the problems that the spectral information of the multi-spectral remote sensing images used by the geodesic active contour(GAC)model is not sufficient and the shoreline extraction accuracy is poor in the terrain complex and changeable circumstances.Based on the multi-spectral remote sensing image as the data source,the spectral feature term constructed by KPCA combines the image data item based on the GAC model to establish the shoreline extraction model,which effectively improves the extraction accuracy of the shoreline.2.Considering the influence of mixed pixel problem on the extraction accuracy of shoreline,sub-pixel mapping is used to determine the specific spatial position of water and land in mixed pixels.The main two aspects are studied and discussed:(1)Locally adaptive sub-pixel shoreline mapping based on SEKSFC algorithm(LA_SPSMS)is proposed to solve the problem that the accuracy of sub-pixel mapping may be affected because the global endmembers are usually chosen in some sub-pixel mapping methods while ignoring the same category in remote sensing images may exist a large spectral difference.The model uses the SEKSFC algorithm to extract the initial shoreline,and then uses the mathematical morphology to determine the mixed pixels which need to be processed.Finally,locally adaptive sub-pixel shoreline mapping of fusion spectral information is used to locate the boundaries of water and land in mixed pixels.The model reduces the influence of the spectral decomposition error caused by using the global endmembers and improves the accuracy of the sub-pixel mapping of shoreline.(2)Aiming at the problem that the sub-pixel mapping algorithms are diverse and which algorithm is better can't be determined in many cases,considering the auxiliary data can provide more constraints,a sub-pixel shoreline mapping model based on panchromatic images and multiple algorithms is proposed.This model chooses three typical algorithms including LA_SPSMS,sub-pixel mapping based on Markov random field and sub-pixel mapping based on Hopfield neural network,and adds the constraint conditions of high-resolution panchromatic images to each algorithm.By analyzing the sub-pixel mapping results obtained by each algorithm,the class of each subpixel is determined by weighting electoral law,which improves the accuracy of the shoreline.