奶牛面部识别图像倾斜校正算法的研究  

Research of Image Incline Correction Algorithm in Cow Face Recognition

作　　者：臧学莲[1] 王建平[1] 喻轩[1] 倪志江[2] 

机构地区：[1]中国农业大学信息与电气工程学院 [2]中国农业大学工学院,北京100083

出　　处：《农机化研究》2008年第9期150-152,共3页Journal of Agricultural Mechanization Research

基　　金：国家星火计划项目(2005EA105003)

摘　　要：实现基于计算机视觉的奶牛面部自动识别是奶牛场数字化的关键步骤之一。为此,提出了基于奶牛面部结构特征的图像倾斜校正算法:使用彩色图像预处理和图像二值化分离图像中奶牛面部信息与背景信息,再结合奶牛面部几何结构特征,提出了使用旋转—投影法估计图像倾斜的角度,通过改进积分投影函数、降低图像分辨率的方法提高算法的运算速度,最后根据奶牛面部的对称性提高角度计算精度,并根据该角度校正奶牛面部图像。经实验验证,该算法运算速度快,校正效果好。Automatic cow face identification based on computer vision is one of the important steps to realize digital dairy farm management. The paper introduced an algorithm for correcting slantwise image which based on the feature of cow＇ s facial structure： color image preprocessing and image binarization were used to separate cow＇ s facial information and background information of images;combined with the cow＇ s facial geometric structure features, the rotating - projection algorithm is proposed to estimate the incline angle, enhanced the computing speed by improving integral projection function and reduce image resolution, finally, improved the accuracy of the angle based on cow facial symmetry, and corrected cow facial image in accordance with the angle. Experiments showed that the algorithm is good at computing speed and correcting effect.

关 键 词：奶牛面部 倾斜校正 积分投影 结构特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]