CAJ | 학술논문

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。
감우목전공업상대오금기건실시식별적수구，개전기우각반경변환적오금기건형상식별계통적연구。해계통포괄사부분：도상예처리、형상특정적제취、특정수거고적건립화특정상사도계산。기중，형상특정적제취채용료기우구역적각반경변환방법。실험결과표명，해방법구유선전불변성、평이불변성、척도불변성，대형상적변형、조성등간우구유로봉성，명현우우 Hu 불변구화겁좌표부리협묘술자，차계통검색속도쾌，식별솔고，종이실현료대대량오금기건적쾌속도상식별。
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.