Bercovich A, Edan Y, Alchanatis V, Moallem U, Parmet Y, Honig H, Maltz E, Antler A, Halachmi I
Department of Industrial Engineering and Management, Faculty of Engineering Sciences, Ben-Gurion University of the Negev, Beer-Sheva 84105, Israel.
J Dairy Sci. 2013;96(12):8047-59. doi: 10.3168/jds.2013-6568. Epub 2013 Oct 4.
Body condition evaluation is a common tool to assess energy reserves of dairy cows and to estimate their fatness or thinness. This study presents a computer-vision tool that automatically estimates cow's body condition score. Top-view images of 151 cows were collected on an Israeli research dairy farm using a digital still camera located at the entrance to the milking parlor. The cow's tailhead area and its contour were segmented and extracted automatically. Two types of features of the tailhead contour were extracted: (1) the angles and distances between 5 anatomical points; and (2) the cow signature, which is a 1-dimensional vector of the Euclidean distances from each point in the normalized tailhead contour to the shape center. Two methods were applied to describe the cow's signature and to reduce its dimension: (1) partial least squares regression, and (2) Fourier descriptors of the cow signature. Three prediction models were compared with manual scores of an expert. Results indicate that (1) it is possible to automatically extract and predict body condition from color images without any manual interference; and (2) Fourier descriptors of the cow's signature result in improved performance (R(2)=0.77).
体况评估是评估奶牛能量储备以及估计其肥胖或消瘦程度的常用工具。本研究提出了一种计算机视觉工具，可自动估计奶牛的体况评分。在以色列一家研究型奶牛场，使用位于挤奶厅入口处的数码单反相机收集了151头奶牛的顶视图图像。奶牛的尾根区域及其轮廓被自动分割和提取。提取了尾根轮廓的两种特征：（1）5个解剖点之间的角度和距离；（2）奶牛特征，即归一化尾根轮廓中每个点到形状中心的欧几里得距离的一维向量。应用了两种方法来描述奶牛特征并降低其维度：（1）偏最小二乘回归；（2）奶牛特征的傅里叶描述符。将三种预测模型与专家的人工评分进行了比较。结果表明：（1）无需任何人工干预即可从彩色图像中自动提取和预测体况；（2）奶牛特征的傅里叶描述符可提高性能（R(2)=0.77）。
