基于Haar特征的Adaboost算法的猪脸检测  

Pig Face Detection by Adaboost Algorithm Based on Haar Feature

作　　者：李景 吴玉秀 张捍东 LI Jing;WU Yuxiu;ZHANG Handong(School of Electrical and Information Engineering,Anhui University of Technology,Maanshan 243000)

机构地区：[1]安徽工业大学电气与信息工程学院,安徽马鞍山243000

出　　处：《常州工学院学报》2023年第3期14-19,共6页Journal of Changzhou Institute of Technology

摘　　要：为了能够更好地帮助猪场养殖户对猪场进行日常信息管理和全流程追溯,利用基于Haar特征的Adaboost算法对猪脸进行检测。检测方法如下:首先对采集的猪脸图像进行灰度处理,去除噪声;其次利用Adaboost算法搭建模型,训练出可以区分猪脸的弱分类器,这些弱分类器通过优化后再级联到一起构成强分类器;再次采用Haar特征来计算样本中对应的特征值信息;最后对图像进行多次缩放保证进行猪脸检测时能够捕捉到足够的信息。实验结果表明,与传统的主成分分析法相比,基于Haar特征的Adaboost算法生成的猪脸分类器能够更好地进行猪脸检测,在运行速度和检测准确率上都有较大提升。In order to help pig farmers better manage daily information and trace the whole process of pig farms,the Adaboost algorithm based on Haar feature was used to detect pig face.The detection methods are as follows:first,the collected pig face images were grayed out to remove noise;the Adaboost algorithm was used to build a model to train weak classifiers distinguishing pig faces,and these weak classifiers were optimized and then cascaded together to form a strong classifier;then Haar features were used to calculate the corresponding eigenvalue information in the samples;finally,the images were scaled several times to ensure that sufficient information could be captured when pig face detection was performed.Experimental results show that,compared with traditional principal component analysis method,the pig face classifier ge\|nerated by the Adaboost algorithm based on Haar features can effectively detect pig faces,and has a greater improvement in running speed and detection accuracy.

关 键 词：猪脸检测 ADABOOST HAAR特征 分类器 级联 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]