摘要
探讨基于Halcon中多层感知机织物色差检测研究。首先将数据库中部分样本特征数据输入神经网络,然后采用均方误差最小化训练优化神经网络,构建样本特征与色差类别的多层感知机映射函数;最后将面积更大、上色情况更复杂的织物输入多层感知机,通过形态学算法分割出织物中不同类别的色差区域,并统计三种颜色织物的色差检测平均准确率,最低可达到89.7%。认为:织物图像中的不同色差区域均能准确判别并分割出来,能满足工业色差检测的准确率与灵敏度要求。
The study of fabric color difference detection based on Halcon multilayer perceptron was dis- cussed. Firstly, part of the sample characteristics data from database were input into neural network, then neural network was trained and optimized by adopting minimum mean squared error. Multilayer perceptron mapping function for sample characteristics and the classification of color difference was established. In the end, the fabric with larger area and more complicate coloring situation was input into multilayer perceptron. By morphological al- gorithm, the area with different classifications of color difference in the fabric was divided. The average accuracy rate of color difference detection for the fabrics with three kinds of colors were counted. The lowest accuracy was 89.7 %. It is considered that the different color difference areas in the fabric image can be accurately distinguished and divided. It can meet the requirement of accuracy and sensitivity for industrial color difference detection.
作者
孟秀萍
苏工兵
吴奇明
周会勇
MENG Xiuping, SU Gongbing, WU Qiming ,ZHou Huiyong(Wuhan Textile University, Hubei Wuhan, 43007)
出处
《棉纺织技术》
CAS
北大核心
2018年第5期60-65,共6页
Cotton Textile Technology
基金
国家自然科学基金(51375351)
关键词
图像处理
织物色差
多层感知机
色差检测模型
神经网络
Image Processing, Fabric Color Difference, Multilayer Perceptron, Color Difference DetectionModel, Neural Network
