An improved distance regularization level set evolution method for coastline change detection

Dynamic monitoring of coastline is of great significance to the ecological environment protection in coastal areas. An improved distance regularization level set evolution method for coastline change detection is proposed in this paper. To reduce the sensitivity of the level set method to the initial contour, the normalized difference water index extracted from the remote sensing image is clustered by K-means to generate the preliminary binary segmentation result, afterwards the initial level set function is obtained by morphological operators. In light of the parameter initialization of level set method, the local entropy of remote sensing image in each channel is weighted for initializing parameter through nonlinear mapping, which avoids the disadvantage of manual parameter adjustment. Meanwhile, an exponential edge stop function based on posterior probability weight is introduced, which makes use of the posterior probability of the target in the background and the characteristics of the exponential edge stop function to accelerate the curve evolution in the background. Finally, the change detection map is acquired by comparing and analyzing the coastline extraction results of bi-phase images. Experiments on the island data set indicate that the effectiveness of the proposed method in coastline extraction and coastline change detection compared with state-of-the-art methods.