The results of existing publicly available automatic ice lead identification algorithms have been verified using expert identification of ice cover leads in the Laptev and East Siberian seas based on optical satellite imagery. It has been found that none of these algorithms can be used to obtain data that allows for calculating lead characteristics such as orientation, length, and spatial density. For automatic ice lead identification, it has been proposed to develop an algorithm using a convolutional neural network trained on data from the AARI electronic archive of sea ice leads.
Dyment et al. (Mon,) studied this question.