基于双目-红外多源数据的猪只体尺自动测量方法  

Method of automatically measuring body size of pigs based on stereo-infrared multi-source data

作　　者：周明彦 黎煊[1,2] 祝志慧 陈萌放[2,3] 徐迪红 ZHOU Mingyan;LI Xuan;ZHU Zhihui;CHEN Mengfang;XU Dihong(College of Engineering,Huazhong Agricultural University,Wuhan 430070,China;Ministry of Agriculture and Rural Affairs Key Laboratory of Smart Farming for Agricultural Animals,Wuhan 430070,China;College of Information,Huazhong Agricultural University,Wuhan 430070,China)

机构地区：[1]华中农业大学工学院,武汉430070 [2]农业农村部智慧养殖技术重点实验室,武汉430070 [3]华中农业大学信息学院,武汉430070

出　　处：《华中农业大学学报》2025年第2期9-16,共8页Journal of Huazhong Agricultural University

基　　金：武汉市生物育种重大专项(2022021302024853);生猪现代育种技术研发及新品种选育(HBZY2023B006-03);华中农业大学-中国农业科学院深圳农业基因组研究所合作基金项目(SZYJY2022031)。

摘　　要：针对猪只体尺测量过程中普通彩色图像易受到环境光线影响、深度图像易产生空洞等问题,提出一种基于双目-红外图像的非接触式猪只体尺测方法。该方法使用红外图像提取猪只前景图像,基于椭圆拟合法和投影差分法分割猪只头尾部,采用凸包分析法和对称性检验法筛选猪只理想姿态。同时,利用双目图像获得猪只深度信息,使用单应性矩阵联系红外与双目图像的位置信息,并在此基础上设计了猪只体尺测量方法。结果显示:基于红外图像提取出的猪只轮廓清晰且平滑;理想姿态筛选算法精确度为94.0%;体长、体宽、臀宽、体高和臀高测量的平均相对误差分别为1.78%、3.26%、3.20%、1.92%和2.27%,平均相对误差的平均值为2.49%,平均绝对误差为1.42 cm。研究表明,本文提出的算法能够满足猪只体尺测量的精度要求,为猪只体尺的连续、自动测量提供了可行的解决方案。A non-contact method of measuring the body size of pigs based on stereo-infrared images was established to address the issues of common color images easily affected by environmental lighting and depth images prone to producing holes in the process of measuring the body size of pigs.Infrared images were used to extract foreground of pigs.Ellipse fitting and projection difference methods were used to segment the head and tail of pig.Convex hull analysis and symmetry testing methods were used to select the ideal posture of the pig.Stereo images were used to obtain depth information of the pig,and a homography matrix was used to fuse the advantages of both the infrared and stereo images.A method of measuring the body size of pigs was designed based on these.The results showed that the contours of the pig extracted from the infrared images were clear and smooth.The algorithm of selecting ideal posture had an accuracy of 94.0%.The average relative error of the measurements for body length,body width,hip width,body height,and hip height was 1.78%,3.26%,3.20%,1.92%,and 2.27%,respectively,with an overall average relative error of 2.49% and an average absolute error of 1.42 cm.It is indicated that the algorithm proposed meets the requirements of accuracy for measuring the body size of pigs.It will provide a feasible solution for continuously and automatically measuring the body size of pigs.

关 键 词：猪 体尺测量 单应性矩阵 数据融合 姿态检测 

分 类 号：S818[农业科学—畜牧学]