基于双目-红外多源数据的猪只体尺自动测量方法
作者:
作者单位:

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

作者简介:

周明彦,E-mail:1069580244@qq.com

通讯作者:

徐迪红,E-mail:xudihong@mail.hzau.edu.cn

中图分类号:

S818

基金项目:

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)


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

1.College of Engineering, Huazhong Agricultural University, Wuhan 430070, China;2.Ministry of Agriculture and Rural Affairs Key Laboratory of Smart Farming for Agricultural Animals, Wuhan 430070, China;3.College of Information, Huazhong Agricultural University, Wuhan 430070, China

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • | |
  • 文章评论
    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。

    Abstract:

    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.

    图1 猪只图像采集平台Fig.1 Platform of pig image acquisition
    图2 猪只图像采集工作流程Fig.2 Work flow chart of pig images acquisition
    图3 猪只红外和双目图像Fig.3 Infrared and stereo images of pigs
    图4 红外图的猪只前景目标提取Fig.4 Pig target extraction with infrared image
    图5 猪只头尾分割过程Fig.5 Pig head tail segmentation
    图6 猪只头部和尾部凸包分析图Fig.6 Analysis of head and tail convex hull
    图7 非理想姿态猪只头部凸包分析图Fig.7 Analysis of convex hull of pig head in non ideal posture
    图8 体尺测量示意图Fig.8 Schematic diagram of body scale measurement
    图9 双目相机标定与深度信息计算Fig.9 Stereo camera calibration and depth information computation
    图10 猪只前景提取结果Fig.10 Prospect extraction results of pigs
    图11 连续体尺测量变化图Fig.11 Body scale variation diagram
    图12 平均相对误差的箱线图Fig.12 Boxplot of the average relative error
    表 1 体尺测量的平均误差Table 1 Average relative error of body size measuremen
    参考文献
    相似文献
    引证文献
引用本文

周明彦,黎煊,祝志慧,陈萌放,徐迪红.基于双目-红外多源数据的猪只体尺自动测量方法[J].华中农业大学学报,2025,44(2):9-16

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-11-29
  • 在线发布日期: 2025-04-02
文章二维码