基于深度卷积神经网络的蛋鸡体温监测系统  

Laying Hen Body Temperature Monitoring System based on Deep Convolutional Neural Network

作　　者：杨雨彤 句金 任守华[1] YANG Yutong;JU Jin;REN Shouhua(Heilongjiang Bayi Agricultural Reclamation University,Daqing Heilongjiang 163000,China)

机构地区：[1]黑龙江八一农垦大学,黑龙江大庆163000

出　　处：《现代畜牧科技》2023年第10期51-55,共5页Modern Animal Husbandry Science & Technology

摘　　要：蛋鸡养殖是我国畜牧业的重要组成部分,蛋鸡的体温是影响其生长、产蛋和健康的重要指标。以蛋鸡体温监测为研究对象,提出一种基于深度卷积神经网络的蛋鸡体温监测系统,实现对蛋鸡的自动检测、定位、温度计算和异常报警。利用热红外相机采集蛋鸡的热红外图像,分别采用YOLOV3和YOLOV42种算法进行目标检测,对比分析2种算法在蛋鸡目标检测任务上的性能,结果表明,YOLOV4算法具有更高的准确率(95.60%)、召回率(94.80%)和mAP(96.10%)。设计了一种基于聚类分析的蛋鸡体温异常检测方法,通过对像素值进行聚类,得到蛋鸡的异常温度阈值,对检测到的蛋鸡目标进行温度计算和异常判断,实现对异常蛋鸡体温的自动报警。Chicken farming is an important component of animal husbandry in our country,and the body temperature of chickens is a crucial indicator that affects their growth,egg production,and health.In this study,we focus on the monitoring of chicken body temperature and propose a chicken body temperature monitoring system based on deep convolutional neural networks.This system achieves automatic detection,localization,temperature calculation,and anomaly alerting for chickens.Initially,thermal infrared images of the chickens are captured using a thermal infrared camera.Then,two algorithms,YOLOV3 and YOLOV4,are employed for object detection in chicken target identification.A comparative analysis of the performance of these two algorithms in chicken target detection tasks is conducted,and the results demonstrate that the YOLOV4 algorithm exhibits higher accuracy(95.60%),recall rate(94.80%),and mean Average Precision(mAP)(96.10%)than YOLOV3.Additionally,we design a chicken body temperature anomaly detection method based on cluster analysis.By clustering pixel values,we obtain the abnormal temperature threshold for chickens.The calculated temperatures and abnormality judgments are applied to detected chicken targets,enabling automatic alert generation for abnormal chickens.

关 键 词：蛋鸡 热红外图像 体温监测 深度卷积神经网络 聚类分析 

分 类 号：S831[农业科学—畜牧学]