Abstract: In order to accurately and efficiently extract coastlines from synthetic aperture radar images,this paper proposes a coastline extraction algorithm based on neural network-optimized Canny operator.The algorithm combines neural networks and statistical regression to adaptively determine the optimal values of the three Canny parameters that are Gaussian filter standard deviation,high threshold,and low threshold,in order to improve the Canny operatorly.Firstly,a neural network model is trained on the training set to obtain the optimal CaPP values for each SAR image.Then,statistical regression and optimization criteria are used to establish the optimal linear combination of CaPP and the SAR image's mean and standard deviation.Finally,the algorithm is experimentally verified using a test set.The experimental results show that the proposed algorithm can adaptively obtain the optimal CaPP values,with the SSIM mean of the coastline extraction results in the test set being 0.912,and the overall accuracy and Kappa coefficient means are 98.55%and 0.966 3,respectively.This demonstrates that the proposed algorithm can accurately extract the coastlines of SAR images by adaptively obtaining the optimal CaPP values.

Key words: coastline extraction, SAR images, Canny algorithm, neural network