摘要
由于人耳识别具有不受表情、化妆等影响的独特优势,研究了人耳识别特征技术,对不变矩算法进行了改进,获得6个改进的高阶不变矩特征向量。分别用两种算法对人耳图像进行识别处理,比较两种识别结果。改进后不变矩算法的高阶不变矩具有平移、旋转不变和尺度缩放不变的特性。采用BP人工神经网络对60只人耳图像进行识别验证,正确识别率达到91.8%。
This paper introduces the process of ear identification. An improvement in moment invariants algorithm allows six feature vectors to be distilled by the method of high-order moment invariants. The use of these two kinds of algorithms in ear identification finds that the high-order moment invariants in the improved algorithm shows the feature of invariance in translation, rotation and scale change. The application of a BP artificial neutral network to identify 60 ear images shows an identification accuracy of 91.8%.
出处
《黑龙江科技学院学报》
CAS
2008年第1期51-53,57,共4页
Journal of Heilongjiang Institute of Science and Technology
关键词
人耳识别
改进不变矩算法
图像处理
ear recognition
improved algorithm of moment invariant
image processing
