In the era of big data, utilizing online platforms allows for an accurate understanding of the public’s perception and preferences towards coastal landscape features in different regions. This provides valuable references and support for coastal city image construction and spatial planning. Taking 12 coastal cities in China as the research subjects, the research employed Python to crawl coastal landscape images from five major online platforms, including Baidu, Bing, Tuchong, Sina Weibo, and Ctrip. With the aid of the pre-trained FCN model from the ADE20K dataset, we conducted semantic segmentation on the images. By clustering and quantitatively analyzing the segmentation results of 150 landscape elements, we explored the imagery types, spatial variations, and visual features of coastal landscape imagery in public perception. The results indicate: (1) Natural landscape imagery types are perceived more prominently than urban spaces, recreational areas and cultural landscape imagery types in coastal cities of China. (2) Significant spatial variations in the perception of coastal landscape imagery exist among different cities, with higher perception uniqueness observed in the South China Sea and Yellow Sea regions, and lower in the Bohai Sea and East China Sea regions. Overall, the central coastal cities exhibit higher diversity, while the coastal cities on the northern and southern ends show lower diversity. (3) The visual characteristics of coastal cities in China include moderate overall blue dominance, low green dominance, spatial enclosure, and facility provisioning. This research contributes to the preservation of distinctive landscape features in various coastal areas and the exploration of coastal landscape diversity. It also provides insights and references for studying public perception of coastal city appearances and preferences based on big data imagery analysis.