摘要
为了实现可靠的机械零件识别,设计了一套基于ART2神经网络的零件形状识别系统,采用一种新的图像特征,即对预处理后的图像进行正交傅里叶-梅林矩变换.该特征对图像大小、方位、强度变化具有良好的不变性.同时采用ART2神经网络识别方法,并进行了算法改进.实验结果表明,该系统能够进行有效识别.最后,对正交傅里叶-梅林矩的计算进行分析,并提出进一步研究的方向.
A shape recognition system based on ART2 neural network is developed in this paper to realize the reliable recognition of workpiece. A new type of features, orthogonal Fourier-Mellin moments, is extracted after image preprocessing. This type of features is advantage of being invariant to dimension, orientation and intensity variance. The algorithm of ART2 nerual networks is modified and employed to recognize the pattern. Experimental results verify the effectiveness of this system. Further research according to analysis of the OFFM computation is .put forward as well.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
2009年第3期117-120,共4页
Journal of Harbin Institute of Technology
基金
国家自然科学基金资助项目(60475028)
关键词
模式识别
特征提取
正交傅里叶-梅林矩
归一化
ART2
pattern recognition
feature extraction
orthogonal fourier-mellin moments
normalization
ART2