摘要
针对利用红外热成像进行生猪体温自动提取困难的问题,在设施猪场生猪体温红外巡检装置的基础上,提出将生猪耳部区域作为其体温的代表区域,探索一种基于红外热像图的生猪耳温自动提取算法(IT-PETE)。该算法通过高效而准确地识别生猪耳部区域并提取耳部区域的温度最大值和平均值,实现生猪体温非接触式自动监测。IT-PETE算法首先用拉普拉斯算子对生猪热红外图像进行预处理,然后基于YOLO v4和形态学对热红外图像中的生猪耳部进行提取,并结合耳部分割图像和温度矩阵自动获取耳部区域温度的最大值和平均值。采用5折交叉验证方法训练生猪耳部区域检测模型,训练集和验证集图像共2000幅,测试集400幅。试验表明,YOLO v4耳部区域检测准确率为97.6%,比Faster R-CNN和SSD分别提高了2.0个百分点和7.8个百分点,单帧图像的平均检测时间为12 ms。同时对20头猪的人工统计耳温数据与算法提取体温进行相关性分析,得到两者在耳部区域温度最大值和平均值的决定系数分别为0.9849和0.9119,表明IT-PETE算法对体温数据的提取具有可靠性和可行性。因此,IT-PETE算法在一定程度上可为生猪体温自动化监测和预警系统提供技术支撑。
In order to solve the problem of pig temperature automatic extraction by infrared thermography,based on the infrared inspection device of pig body temperature in facility farm,an automatic pig ear temperature extraction algorithm IT-PETE,based on infrared thermography,was proposed to identify the ear region of pigs efficiently and accurately and extract the maximum and average values of the ear region to achieve non-contact automatic monitoring of pig body temperature.Specifically,Laplace operator was used to preprocess the thermal infrared image of pigs,then extract the ear of pigs from the thermal infrared image based on YOLO v4 and morphology,and automatically the maximum and average temperature of the ear region was obtained by combining the ear segmentation image and the temperature matrix.The detection model of pig ear region was trained by five fold cross validations,with 2000 infrared thermal images as training and validation sets,and 400 images as test sets.The YOLO v4 detection accuracy of ear region reached 97.6%,which was 2.0 percentage points and 7.8 percentage points higher than that of Faster RCNN and SSD,respectively.The average detection time of single frame image was 12 ms.Meanwhile,through the analysis on the correlation between the artificial statistics of the ear temperature data of 20 pigs and the body temperature extracted by the algorithm,it was obtained that the correlation between their maximum and average temperature in the ear area was 0.9849 and 0.9119 respectively,which showed the reliability and application of the IT-PETE algorithm to extract the body temperature data.Therefore,IT-PETE algorithm can provide technical support for automatic monitoring and early warning system of pig body temperature to a certain extent,with a good application prospect.
作者
肖德琴
林思聪
刘勤
黄一桂
曾瑞麟
陈丽
XIAO Deqin;LIN Sicong;LIU Qin;HUANG Yigui;ZENG Ruilin;CHEN Li(College of Mathematics and Informatics,South China Agricultural University,Guangzhou 510642,China;Wens Foodstuff Group Co.,Ltd.,Yunfu 527300,China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2021年第8期255-262,共8页
Transactions of the Chinese Society for Agricultural Machinery
基金
广东省重点研发计划项目(2019B090922002、2019B020215004、2019B020215002)。
关键词
生猪
耳温提取
YOLO
v4
热红外成像
live pigs
ear temperature extraction
YOLO v4
infrared thermography