Efficient active contours with optical physical parameters for shoreline extraction: a general method and novel framework

Accurate shoreline extraction is important for global climate change, land-sea interactions, and ocean dynamics. However, many current shoreline extraction methods have limitations in accuracy and efficiency when applied to different data, resulting from either empirical model parameters or limited image types. Considering the advantages of active contour in applying variational ideas to solve energy generalizations and the universal adaptability of optical physical parameters, an active contours with optical physical parameters (ACOPP) method that rapidly converges for shoreline extraction is proposed in this study. This model uses active contours to calculate the pixel-level energy generalization function of the image and combines the energy of the optical physical parameters to accelerate the process of contour curve approximation to the target edge by energy confrontation through iterative fitting, and improve the accuracy and efficiency of the target curves. Different images with multiple time phases in different regions were selected as experimental data, including GF2, SPOT-5, Sentinel-2, and Landsat-8, to demonstrate the accuracy, efficiency, and universality, while shoreline complexity coefficient was proposed for the study area division, which was used for shoreline extraction by ACOPP. Finally, the extraction results were analysed and validated, and the conclusions are as follows: ACOPP has a faster convergence speed in the extraction process than traditional methods, and the extraction results have a higher spatial resolution. The extraction results were highly accurate, and the mean absolute error (MAE) compared with the validation sets was < 2 pixels. The model is applicable to remote sensing images in which the normalized difference water index (NDWI) can be calculated, that is, optical remote sensing images containing green and near-infrared (or mid-infrared) bands. The experimental results prove the feasibility of ACOPP and a novel framework that combines ACOPP to extract target feature edge information for remote sensing images is proposed, the feasible and extensible of the frame have been verified.