江杰，岳伟，曹孟珍
（内蒙古科技大学 信息工程学院，内蒙古 包头014010）
关键词：关键帧提取;辅助标识法;测点提取;体尺测量
中图分类号:TP391.413 文献标志码:A
摘 要：羊体体尺直接反映羊生长发育情况，利用机器视觉实现羊体体尺无接触式测量。首先，采用辅助标识法羊体提取背部轮廓曲线；采用D-P算法和海伦-秦九绍公式寻找曲线曲率最大的点作为臀部测点；其次，分析轮廓曲线特征信息，提取关键帧。最后，在寻找到的关键帧的基础上，寻找肩胛点。结合空间分辨率计算出羊体体尺参数。实验结果表明，可以准确提取关键帧，体尺测量平均误差不超过3%。
Research on sheep body size measurementmethod
based on computer vision
JIANG Jie YUEWei CAO Meng-zhen
(School ofInformation Engineering, Inner Mongolia University of Science and Technology， Baotou Nei Mongol 014010, China)
Key words:key frame extraction； auxiliaryidentification method；measuring point extraction； body measurement
Abstract:The measurementof sheep body size directly reflects thedevelopingsituationof the sheep.Machinevision wasused to realize sheep body measurementswithoutcontacting the sheep body.Firstly, auxiliaryidentification method was adopted toextract sheep’s backcontour curve.Themaximum point of the curve curvature as a hipmeasuring point was obtained by usingD- Palgorithmand Helen- Qin Jiushao formula.Secondly,thecontour is analyzed to get itsfeature information and to extract key frames.Finally, theshoulder pointwas found onthe basis of the foundkey frames ,and thesheep body parameters werecalculated combinedwith spatial resolutionmethod. The experimental results show that thekey frames can beextractedaccurately and theerror forbody measurementis lessthan3% on average.