文献类型：期刊文献

英文题名：Image colour application rules of K'o-ssu based on machine learning algorithm

作者：Li, Junxuan[1];Fu, Rongrong[1];Wang, Yuewei[1];Fan, Ruiyang[1];Cui, Hangrui[1]

机构：[1]East China Univ Sci & Technol, Shanghai 200237, Peoples R China

年份：2025

卷号：280

外文期刊名：EXPERT SYSTEMS WITH APPLICATIONS

收录：;EI(收录号：20251518215245);WOS:【SCI-EXPANDED(收录号:WOS:001469718100001)】；

语种：英文

外文关键词：K'o-ssu; Colour extraction; Machine learning; Colour features; Grey Relation Analysis; Image colour application rules

摘要：Colour in brand packaging design has an important role in enhancing brand recognition and cultural association, and is also an effective path to achieve brand differentiation. At present, Chinese brand colour recognition, colour matching and cultural association mainly rely on the designer's subjective perception, and have not yet formed the rules of colour application in line with Chinese-style colour aesthetics. In order to obtain the reference target colour resources with brand characteristics and in line with Chinese colour aesthetics and cultural associations, this study proposes a scientific method of colour extraction and clustering of K'o-ssu (silk tapestry with cut designs) through machine learning, forming the K'o-ssu colour dataset,discovering the colour application rules of K'o-ssu. At the same time, we updated the packaging design process by matching the colour matching rules in the dataset with the brand's perceptual vocabulary through Grey Relation Analysis (GRA), so that the discovered colour rules are applied to the colour design of the product packaging. Firstly, YOLOv5s target detection model and Grabcut algorithm are selected for pattern detection, segmentation and extraction. And then SLIC super pixel algorithm and BIRCH clustering algorithm are used to realize the colour gradual clustering of segmented samples, so as to extract the K'o-ssu feature colour dataset. The FP-Growth algorithm is used to carry out colour correlation research, summarize the colour matching relationship of each sample set, and excavate the colour matching law. Secondly, this study uses two round GRA combined with the KANO model to analyze colour combination for brand cognition, which adjusts the weights according to the type of user demand. Finally, the first-ranked colour combinations and shades are applied to the colour design of brand product packaging. In order to verify the scientific validity of the colour application rules proposed in this study, the effectiveness of the colour application design scheme was evaluated using eye-tracking experiments and satisfaction questionnaires, and the results proved that the design scheme obtained in this study meets the users' aesthetics and consumption needs.