Remote Sensing Images Coastline Detection Methods Based On Active Contour Models

Coastline detection is critical for autonomous navigation, geographical exploration, coastline resource management, coastal environmental protection and sustainable coastal development. Remote sensing images are particularly suited to be used to detect the coastline because features can be easily captured on Earth’s surface in large area. Due to the subjectivity and substantial effort involved in manual delineation, a great deal of research effort on automatic method has been devoted to coastline detection., but post-processing and over-segment often occurred. Recently, active contour models have been extensively applied to coastline detection field by its smooth and closed contours. At present, most existing algorithms are focused on the edge-based models. However, edge-based models are dependent on contour initialization and more sensitive to image noise. In this paper, to provide solutions to these challenges, three novel active contour models are proposed.The main contributions of this thesis are as follows:1) A novel region-based local region mean decending active contour model is proposed. First, based on the model of remote sensing images with artifacts and weak areas, we propose a novel artifacts removal method and utilize two restrictions to obtain a novel energy functional. Second, by minimizing this energy, the variable expressions are deduced. Third, an algorithm description is provided. And then, the method has been validated on the real remote sensing images. Experiments show that our method is faster and more accurate than some existing models.2) A novel region-based local region mean ascending active contour model is proposed. First, base on the model of remote sensing images with artifacts and speckle noise, we proposed a novel artifacts removal method to compensate strong inhomogeneity intensity in sea and utilized two restrictions to obtain a novel energy functional. Second, by minimizing this energy, the variable expressions are deduced. Third, a proposition is demonstrated to prove convergence of local region mean. Then, an algorithm description is given. Finally, the method has been validated on the real remote sensing images. Experiments show that our method is faster and more accurate than some existing models.3) A novel edge-based active contour model is proposed. First, based on the remote sensing images with low contrast and Local Binary Fitting model, a new signed pressure force function is proposed and deduced. Second, an algorithm description is shown. Finally, the proposed model has been validated on the remote sensing images. Experiments show that our method is more accurate than IVC model.