摘要
鉴于目前工业上对五金器件实时识别的需求,开展基于角半径变换的五金器件形状识别系统的研究。该系统包括四部分:图像预处理、形状特征的提取、特征数据库的建立和特征相似度计算。其中,形状特征的提取采用了基于区域的角半径变换方法。实验结果表明,该方法具有旋转不变性、平移不变性、尺度不变性,对形状的变形、噪声等干扰具有鲁棒性,明显优于Hu不变矩和极坐标傅里叶描述子,且系统检索速度快,识别率高,从而实现了对大量五金器件的快速图像识别。
Given the real-time recognition demand of hardware articles currently in the industry,we carried out the research on angular radial transform-based hardware articles shape recognition system.The system comprises of four parts：image preprocessing,shape feature extraction,feature database establishment and characteristics similarity calculation.Amongst them,for shape feature extraction the system uses the area-based angular radial transform method.Experimental results show that the method has good rotation,scale and translation invariance,it is robust to deformations and noises,and is superior to Hu invariant moment and polar Fourier descriptor,moreover,the system retrieves fast and has high recognition rate,so realises the fast image recognition on the large number of hardware articles.
出处
《计算机应用与软件》
CSCD
2015年第10期181-183,195,共4页
Computer Applications and Software
基金
国家自然科学基金项目(61201392)
关键词
五金器件
角半径变换
形状特征
不变性
图像识别
Hardware articles
Angular radial transform
Shape feature
Invariance Image recognition
