Using Kansei Engineering and Neural Networks in Yarns Design

• Main Authors: Chia-Lin Lin, 林佳霖
• Other Authors: Meng-Dar Shieh
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/21456199717747595125

碩士 === 國立成功大學 === 工業設計學系碩博士班 === 93 === This research established a set take the Kansei Engineering as the foundation flow, and combined neural network to addresses the issue of how humans perceive to the yarn and response on the designers’ new works, so as to the acceleration design flow assisted the designer able to carry on design by more effective also an objective way. The textile industry was already the very mature industry in Taiwan's, because the advance in technology, the technology were developed gaining ground which, the consumer realized, has accelerated the product life cycle, formerly admired the artificial platoon regulation the way because the effectiveness for a period of time and the accuracy the insufficient demand, the product innovation has been gradually regarded by the enterprise the creation value. The 50 pieces yarn samples provided with Mean Time(明大) enterprise co., ltd.. The subjects were invited to measure their subjective impression of 50 different yarns using the Semantic Differential Method (SD). Composes the program to extract the Color Coherent Vector（CCV） as the color features of yarn images, LBP (local binary pattern), SCOV, SAC, VAR to analyze gray-scale of image as texture features, and to code the information of yarns’ weaves and the material. Take the color features, the texture features, the weave and the material’s code as the input layer, and the output layer is the value of 10 impression words. The Back-Propagation Networks（BPN）was trained to approximate the relationship between the kansei features and the features of yarns. The two ways are correlated with each other through a Neural Network mechanism, which is used to correlate the two feature spaces such that the retrieval system can enhance the designer to design the yarns efficiently and conveniently.