Image clustering algorithm and psychological perception in historical building colour rating research: A case study of Guangzhou, China

Abstract

This study aims to construct a new method for collecting colours from Guangzhou historical buildings using the Natural Color Sytem (NCS) colour system for on-site comparison. It seeks to detect and classify the primary and secondary colours of the buildings through image attributes, and to analyse the group co-occurrence and clustering algorithm characteristics of the colours of the historical buildings, as well as the colour hierarchy based on the image clustering algorithm. The study further evaluates the colour ratings of the historical buildings through questionnaire validation, explores the relationship between colour attributes (hue, lightness, and chroma), and classifies the colour evaluation criteria into five tiers. It was found that the symbiotic clustering diagrams of the colours of the historic buildings in Guangzhou exhibited distinct colour rating relationships, with the clustering colour hierarchy showing a high correlation with the results of the colour ratings assessed by questionnaire validation. A close relationship between the computer-generated output and the perceptual judgement is demonstrated, further proving the reliability and accuracy of the computer model. The results of the study provide comparable data and practical tools for colour planning and management of urban historic building facades.