Abstract: The human ear has a lot of structural features. For the insufficiency of the single feature description affecting the ear recognition rate, an ear recognition algorithm based on the Haar local feature is proposed. The algorithm uses elliptical LBP operator and HOG operator which are in line with the external shape of the human ear to extract the texture features and edge features of the image, and fuse the two features. The advantages of using Haar feature calculations are introduced into LBP and HOG feature extraction. The elliptic LBP operator and HOG operator are constructed by designing four groups of Haar coding modes. The experiment and analysis of the improved algorithm are carried out. The experimental results show the effectiveness and practicability of the algorithm.

Key words: ear recognition, Haar feature, local binary pattern, histogram of oriented gradient

摘要： 人耳具有丰富的结构特征，针对单一特征描述影响人耳识别率的不足，提出一种融合Haar型局部特征的人耳识别算法。算法采用符合人耳外部形状的椭圆形LBP算子与HOG算子，分别提取图像的纹理特征和边缘特征，将两种特征进行融合。利用Haar特征运算快捷的优势，引入到LBP和HOG特征提取中。通过分别设计的4组Haar编码模式，构建椭圆形LBP算子与HOG算子。对改进算法进行实验与分析，实验结果表明了算法的有效性和实用性。

关键词: 人耳识别, Haar特征, 局部二值模式, 梯度方向直方图