Research progress on methods of automatic coastline extraction based on remote sensing images

Abstract

The coastline refers to the boundary line between land and sea. The coastline detection is an important part of studying the effects of land-sea interaction and human activities on coastal zones. It is also significant for the effective development, integrated management and sustainable exploitation of coastal resources and protection of coastal ecosystems. Therefore, it is critical to achieve the extraction of coastline quickly and accurately. Remote sensing technology has gradually become a way to detect the location of the coastline and monitor its dynamic changes with its wide range, high temporal resolution, high spatial resolution, multispectral, low cost and other prominent advantages, overcoming the shortcomings of traditional coastline detection methods of long time period, high intensity of labor and so on. This paper reviews the research progress on methods of automatic coastline extraction based on remote sensing images proposed in recent years. First, the definition and classification of the coastline are given. The two stages of extraction of the instantaneous waterline and the real coastline and the specific extraction process are clarified. Then, these methods are divided into eight categories: threshold segmentation, edge detection operator, active contour model, data mining, multiresolution analysis, object-oriented, polarization and other methods. And the basic ideas of the main methods of coastline extraction are elaborated. Finally, the advantages and disadvantages of all kinds of methods are analyzed and compared. The principle of tidal correction and the methods of accuracy evaluation are expounded, and the research work of the next step is prospected. The detailed feature description of main methods are as follows. Threshold segmentation method is simple and easy to implement. While the selection of appropriate threshold is a certain degree of difficulty, and the extraction accuracy needs to be improved. The edge detection operator method has a good effect on the extraction of edge. Simultaneously, this method is susceptible to noise and prone to detect pseudo edges, even the continuity of extracted coastline is poor. The extraction result of the active contour model is accurate. However, the model has high complexity and large computational burden. Data mining method utilizes intelligent means to achieve automatic extraction of coastline. But a variety of methods need to be combined to make the extraction accuracy higher. Although the multiresolution analysis method can obtain rich edge information, the mostly applied wavelets have a limitation of dealing with directional information. Object-oriented approach can achieve image segmentation of a higher level and reduce the influence of texture and other characteristics inside the cell. It cannot make full use of the implicit information of the image when the amount of data is large. And the polarization method is essentially a threshold segmentation method. The only difference is the selection method of threshold. The polarization method achieves threshold selection by usage of different polarization methods of SAR images. In the end, in view of the deficiency of currently existing methods, the following feasible research approaches and prominent research prospects of automatic coastline extraction based on remote sensing images are put forward: constructing dataset, utilizing hyperspectral data, employing deep learning method, adopting swarm intelligence algorithm to optimize the parameters in the method of active contour model, using multiscale transform which contains directional information to improve the extraction accuracy of multiresolution analysis method, combining various kinds of methods to synthesize their advantages and achieving the automatic extraction of coastline with subpixel precision.