基于深度图像的奶牛表型特征获取系统设计与试验

Design and Experiment of Phenotypic Characteristics Acquisition System for Dairy Cow Based on Depth Image

奶牛表型特征是评价奶牛成长状况的一项重要参数,为减少奶牛的应激性且能便捷地获取奶牛几何表型尺寸,运用图像处理技术提取奶牛表型特征参数,设计了一款针对深度图像和点云数据的奶牛几何表型特征获取系统。对奶牛深度图像采用背景减去法、阈值分割、滤波和空洞填充等方法获取奶牛目标区域,对目标采用边缘检测、角点检测和凸包运算等检测特征点,最后对应点云数据获得奶牛表型特征尺寸。系统现场试验结果表明,系统获取的体重准确性在98%以上,体尺准确性在96%以上,系统工作稳定、测量精度高,为实现数字化养殖打下了基础,具有很好的应用前景。

Phenotypic characteristics of dairy cow is an important parameter to evaluate growth status of cows,in order to reduce excitability of cows and obtain geometric phenotypic size of cows easily,a phenotypic characteristics acquisition system based on depth image and point cloud data was designed by using image processing technology to extract phenotypic characteristic parameters of cows.Background subtraction,threshold segmentation,filtering and cavity filling were used to obtain target area of cows for depth image,and edge detection,corner detection,convex hull operation and other detection feature points were used for target.Finally,phenotypic characteristic size of cows was obtained from corresponding point cloud data.Field test of system was carried out and results showed that accuracy of body weight obtained was more than 98%,accuracy of body size was more than 96%,system had a high precision of measuring stability,which could lay a foundation for realization of digital breeding and has a good application prospect.