Quantitative evaluation of color system of small and medium-sized cities based on color recognition of street view images: A case study of Feng County, China

The research on color planning for small and medium-sized cities in China started relatively late, and has not formed a scientific planning, design and control method, which is yet to be analyzed in depth. Currently, there is a relative lack of technical means for quantitative research on urban color, which is mainly reflected in the inefficiency of data collection methods, and the lack of effective extraction and analysis techniques for the color information of the massive street photos. In this paper, the spatial database technology through the geographic information system (GIS), the semantic segmentation technology through a full convolutional network realized by PyTorch, the color recognition technology through Matlab and the color expression technology through Munsell system are associated and integrated. They are integrated into the process of storing, analyzing and expressing color information, forming a quantitative evaluation method for urban color planning in small and medium-sized cities. Finally, the color database of Feng County is constructed as an example. Its color positioning is evaluated and judged to determine the appropriate color range of N6–N9, 2.5YR7/1-10Y9/3, and to propose the reference color spectrum and the forbidden color spectrum, to provide a reference for the small and medium-sized cities’ color planning and make urban color planning more inclined to quantitative expression.