The problem of determining the dimensions occupied by a spatially distributed target (SDT) in a two-dimensional raster image of the observed area of the Earth’s (water) surface, formed by a spaceborne synthetic aperture radar (SAR), is relevant in various space monitoring tasks. In this work, this problem is addressed specifically in the context of determining the dimensions of SDT of surface ships against reflections from the sea surface. Both phenomenological and real models based on SSDD database are used as SAR images. Decision-making regarding the dimensions of surface ships is performed using classical (parametric and non-parametric) algorithms as well as machine learning algorithms leveraging artificial neural networks. The results of the comparative analysis of these algorithms are presented.