产品材料质感偏好意象进化认知算法与系统  

Preference learning for evolutionary cognition algorithm and system of product material texture

作　　者：汪颖[1,2] 张三元[1] 张克俊[1] 孙守迁[1] 

机构地区：[1]浙江大学计算机科学与技术学院,浙江杭州310027 [2]浙江理工大学艺术与设计学院,浙江杭州310018

出　　处：《计算机集成制造系统》2014年第4期762-770,共9页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(61272304;61070075;61303137);浙江省自然科学基金资助项目(Y13F020143)~~

摘　　要：为了使产品设计师能够通过材质设计增加产品差异性,提高消费者的满意度,提出一种基于基因表达式编程的材料质感偏好意象进化认知算法。综合考虑产品材质的视觉、触觉物理量与视觉心理量对产品偏好的影响,对124个地板样本的材色、光泽度和粗糙度物理量进行了测量。60名受试者对地板样本分别进行了纹理和偏好心理量的主观评分,采用主观量表获取实验数据。经过无量纲化处理和主成分分析后,归纳出四个主成分为自变量、偏好意象为因变量,将解析材料质感要素与消费者偏好意象之间的认知关系转化为复杂函数关系建模问题。结果表明,在求解产品材料质感偏好意象认知关系的问题上,该算法比神经网络和支持向量机算法具有更好的预测效果和更强的鲁棒性。基于该偏好进化认知算法开发了地板喜好度测评系统,并在实际应用中验证了算法的有效性。To enhance the product identification by material design and to improve the customer＇s satisfaction, an evo- lutionary cognition algorithm of material texture preference based on gene expression programming was proposed. By considering the influence of product texture＇s visual sense, tactile physical quantity and visual psychological capa- bility on product preference comprehensively, 124 floor tiles were measured in terms of color, glossiness and rough- ness. 60 subjects scored these 124 floor tiles in terms of texture and preference subjectively. After dimensionless treatment and principal component analysis, four components were extracted as independent variables while the product preference as dependent variable, and the cognitive relationship between product material texture aspects and user preference was transformed into a modeling problem of complex functions. The results showed that this gene expression programming had obvious advantages over neutral network and support vector machines in terms of prediction accuracy and robustness on solving product material texture cognition problem. A floor preference assess- ment system was prototyped based on this proposed algorithm and the effectiveness was proved in practical applica- tion.

关 键 词：进化认知 主成分分析 演化算法 材料质感 偏好学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]