兽类学报

• 论文 • 上一篇    下一篇

利用角形几何特征和计算机视觉技术确定岩羊年龄的新方法

苏梦雨1,2 张同作3 连新明3 苏建平3 都玉蓉1,2   

  1. (1 青海师范大学生命科学学院,西宁810008) ;
    (2 青海省青藏高原药用动植物资源重点实验室,西宁810008);
    (3 中国科学院西北高原生物研究所,青海省动物生态基因组学重点实验室,西宁810001)
  • 出版日期:2019-05-30 发布日期:2019-05-10
  • 通讯作者: 都玉蓉 E-mail: xndyr@163.com

A new method for determining the age of blue sheep using horn geometric characteristics and computer vision techniques

SU Mengyu, ZHANG Tongzuo, LIAN Xinming, SU Jianping, DU Yurong   

  1. (1 School of Life and Geography Science, Qinghai Normal University, Xining 810008, China)
    (2 Key Laboratory of Medicinal Plant and Animal Resources the Qinghai-Tibetan Plateau in Qinghai Province, Xining 810008, China)
    (3 Qinghai Provincial Key Laboratory of Animal Ecological Genomics, Northwest Institute of Plateau Biology, Chinese Academy of Sciences, Xining 810001, China)
  • Online:2019-05-30 Published:2019-05-10

摘要: 确定野生动物活体年龄是动物种群生态学研究的一项重要基础工作,然而,迄今为止,没有一种传统方法能高效精准地完成这项任务。岩羊角形态复杂且终身生长,其角形随年龄增长和观察角度变化极大。野外条件下,高效获取岩羊个体年龄信息的唯一方法就是拍摄其角的清晰照片。本研究旨在提出一种利用角形几何特征和计算机视觉技术确定野生岩羊活体年龄的新方法,其原理和操作如下:首先建立覆盖全部年龄段(1.5-15.5岁)的已知年龄羊角实体标本的3D模型(左、右角各22个);再在R程序控制下使其在3D空间中旋转,每旋转1°输出一张角形图片;最后计算每张角形输出图片与待定年龄角照片的形状相似性,找出相似性最高的角形图片对应的3D模型,该3D模型的年龄即为待定年龄羊角的实际年龄。30张已知年龄羊角照片的实验验证结果表明,上述年龄鉴定方法能高效地获得精准结果(准确率达100%)。

关键词: 岩羊, 年龄鉴定, 3D数字模型, R编程, 计算机视觉

Abstract: Studies on the population ecology, behaviour and social structure of ungulates often require accurate measurements of the age of individuals. However, to date there is no simple noncontact method of determining the ages of individuals. The horns of ungulates are structurally complex, and grow and change shape as animals age. If 3D models of the horns of individuals of known ages are built, the ages of other individuals can be estimated from photographs of their horns by comparison with the reference models. This paper describes a new age-determining method for blue sheep (Pseudois nayaur) using horn geometric characteristics and computer vision techniques. The procedure consists of three stages: 1) creating 3D reference images based on the horns of individuals of known age (22 individuals, two horns per individual; age range 1.5 to 15.5 years); 2) creating new 3D images of the horns of target individuals, which are then rotated on a computer using an in-house R script to obtain images at stepwise intervals of 1°; and finally, 3) by calculating the similarity between the new and the reference horn images. A final image with the highest similarity to the reference images is generated, and used to determine the actual age of the target individual. Of the 30 photographs used to evaluate our new method the age of the target individual in every picture was correctly determined. Our new method has potential use for studies of all wild/feral sheep and probably also other ungulate species.

Key words: Wild blue sheep (Pseudois nayaur), Age determining, 3D model, R programing, Computer vision