CAJ | 학술논문

提出了一种基于小波结构矩的具有平移、旋转、缩放不变性的新型图像识别方法.小波结构矩是在小波矩的基础上通过改变图像函数的结构即几何矩的密度得到的.该算法结合了小波和结构矩的优点,不但实现了对图像特征的精细把握还增大了相似图像之间的距离.采用三次样条函数作为母小波,有效提取了图像的全局特征和局部特征.实验证明,小波结构矩比改进的Hu矩和结构矩具有更高的识别率,目前该算法已经成功运用到全自动金丝球焊机图像识别系统.
제출료일충기우소파결구구적구유평이、선전、축방불변성적신형도상식별방법.소파결구구시재소파구적기출상통과개변도상함수적결구즉궤하구적밀도득도적.해산법결합료소파화결구구적우점,불단실현료대도상특정적정세파악환증대료상사도상지간적거리.채용삼차양조함수작위모소파,유효제취료도상적전국특정화국부특정.실험증명,소파결구구비개진적Hu구화결구구구유경고적식별솔,목전해산법이경성공운용도전자동금사구한궤도상식별계통.
Wavelet-structure moments for image recognition were presented to deal with the images which were translated, scaled and rotated. This new theory was based on structure moments and wavelet transformation, after changing the structure of a image function, namely the density in geometric moments, the difference between images was enlarged, using mother wavelet of the cubic B-spine, both global feature and local feature can be extracted efficiently. Experiments demonstrate that wavelet-structure moments perform better than improved Hu moments and structure moments at recognition rate and correctness. And the method was applied into practical automatic golden ball wire bonder systems.