监督模糊聚类神经网络及其应用  

Supervised fuzzy clustering neural network for garment seams

作　　者：潘永惠[1] 包芳[1] 

机构地区：[1]江阴职业技术学院,江苏江阴214405

出　　处：《计算机工程与设计》2009年第15期3612-3614,3623,共4页Computer Engineering and Design

摘　　要：为在面料成衣之前客观评价其缝纫性能,提出了一种基于监督模糊聚类客观评价方法。通过引入输出空间对FCM聚类算法进行改进,同时反映输入空间的聚类特征和输出空间的逼近特性。用FAST系统测量服装面料的力学性能指标,运用核主成分法对所测指标进行分析,提取5个核主成分作为神经网络的输入。实验结果表明,系统可以根据中厚型棉织物的不同结构及物理性能快速准确地给出该织物成衣后的缝纫性能评价指标。To evaluate the sewing ability of the fabric objectively before being garment, a prediction system based on supervised fuzzy clustering neural network （SFCNN） for garment seams is proposed. Improving the objective function of FCM clustering algorithm by importing output space to get the clustering feature of input space and the approximation feature of output space at the same time, and using FAST system to test low stress mechanical properties of fabric, KPCA method is applied to analyze the measured mechanical properties, then five kernel principal components is extracted as the input of SFCNN. Experimental results demonstrate that the proposed approach could efficiently be used as an objective seam pucker evaluation system with high convergence and accuracy, which is robust for various structures and mechanical properties of thick cotton fabric.

关 键 词：中厚型棉织物 缝纫平整度 监督模糊聚类 输出空间 FAST系统 核主成分分析 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP274.3[自动化与计算机技术—控制科学与工程]