基于改进Otsu算法的生猪热红外图像耳根特征区域检测  

Pig Ear Root Detection Based on Adapted Otsu

作　　者：周丽萍[1] 陈志[2] 陈达[2] 苑严伟[1] 李亚硕 郑建华[1] 

机构地区：[1]中国农业机械化科学研究院,北京100083 [2]中国机械工业集团有限公司,北京100080

出　　处：《农业机械学报》2016年第4期228-232,14,共6页Transactions of the Chinese Society for Agricultural Machinery

基　　金：'十二五'国家科技支撑计划项目(2014BAD08B07);北京市科技计划项目(D141100003814003)

摘　　要：为研究规模化生猪养殖场中非接触式体温检测方法,以自行设计的热红外图像采集器采集生猪图像,选取HSV颜色空间对图像进行变换,生成S层图像,应用形态学特征闭运算对二值化图像去噪。用改进后的Otsu算法,分别对仔猪、育肥猪和妊娠猪图像耳根特征区域进行检测。结果表明,该方法可以100%正确检测具有完整耳根部特征的仔猪、育肥猪、妊娠猪图像;对于耳根部特征不完整的仔猪图像23%可正确进行检测,育肥猪图像25%可正确进行检测,妊娠猪图像33%可正确进行检测;无法检测不具有耳根部特征的图像。In order to find the method of auto detecting body temperature in pig ＇s cultivation,the homemade thermal infrared imager was used to acquire the infrared thermograms of pigs. The imager was loaded on the arm of the auto-inspection trolley with the camera lens toward pigs,so it can clearly collect the images of pig ear roots. When the thermograms were collected,they were firstly transformed by HSV colorspace into the S-layer images. Furthermore,the morphological closing operation was used for binary image denoising. Finally,adapted Otsu was applied to ear roots detection of piglets,finishing pigs and pregnant pigs separately. The objective function was selected,because it was simple to find a grayscale to maximize η with the minimum computation at the same time. The results showed that the method can be perfectly used to analyze images with complete ear roots for piglets,finishing pigs and pregnant pigs.When the method was applied to those images with incomplete ear roots,the detection rates were 23% for piglets,25% for finishing pigs and 33% for pregnant pigs. The method cannot be used to analyze images without any ear roots of pigs. The method was applied to the auto-inspection system which can help to find the high temperature pigs in the cultivation. When the abnormal pigs were found,the auto inspection trolley would alert,and then stopped inspecting and began to spray.

关 键 词：生猪体温 温度特征 生猪耳根 OTSU算法 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] S127[自动化与计算机技术—计算机科学与技术]