摘要
针对现有人耳特征提取方法主要采用几何形状法和代数法提取,存在偏差较大的问题,提出了一种新的人耳图像特征提取方法,并将其应用到矿工身份识别中。该方法利用三尺度canny算子提取人耳边缘图像,运用凸包算法提取人耳边缘特征点,采用轮廓搜索算法提取人耳外轮廓,在极平面上用外耳轮廓上的点到极点的距离与人耳长轴的比值构成人耳特征向量,解决了几何形状法提取人耳特征偏差大的问题。将用该方法提取的人耳图像特征用于矿工身份识别,正确识别率达96%。
In view of problem of big error of existing features extraction method for ear image adopting geometrical shape extraction and algebra extraction,a novel feature extraction method for human ear image was proposed to apply to miner identification.The method uses three-scale canny operator to extract image edges of the helix,and adopts convex hull algorithm to extract key points of ear edge image,uses outer contour search algorithms to extract outer contour of ear image;constructs the ear image feature vector with the length ratio of point to the pole distance of outer ear contour and human ear in epipolar plane,which solves the problem of big error of ear geometry feature.The identification accuracy is 96% with the method to extract human ear image feature for miner identification.
出处
《工矿自动化》
北大核心
2015年第11期30-34,共5页
Journal Of Mine Automation
基金
国家自然科学基金重点资助项目(51134024)
关键词
人耳识别
特征提取
CANNY算子
凸包算法
支持向量机
ear recognition
features extraction
canny operator
convex hull algorithm
support vector machine
