| Summary: | 碩士 === 國立成功大學 === 工業設計學系碩博士班 === 91 === The objective of this research is to develop an intelligent fabric retrieval system using computer-based Kansei algorithms to assist fashion designers in designing costume textiles, as well as to help consumers find their preferred textile samples quickly and efficiently. There are two major research aspects in this research: The first one addresses the issue of how humans perceive and measure similarity within the domain of the Kansei of images. To understand and describe this mechanism, a subjective experiment is performed. It used the SD method to find humans’ impression of textile, and structured the factor space of pyscological domain through the Principle component analysis. By conferring the result of Hierarchical Cluster Analysis and the Multidimensional scaling (MDS) from the ex-periments, the suitable features of images that influence human Kansei feeling are dis-cussed. The next step is to define the feature space of images according color composi-tion of fabric images. The two feature vectors, Kansei feature vector and an image feature vector, define the feature spaces. The fabric retrieval system, based on human psychologi-cal Kansei and perception, is able to assist designers in creating new ideas through two ways interactively. The first is the high-level Kansei searching algorithm, where predefined impression words are systematically adjusted. The other is the low-level perception query, which includes perception feature modification and image similarity indexing. Moreover, 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 be “intelligent” for the fashion designers and consumers.
|
|---|
