摘要
以伊犁马为研究对象,通过马场图像采集、实地测量,完成了马体尺测量中关键技术的研究。基于YOLACT实例分割技术,在MS-COCO数据集完成马体与背景的快速、高性能分割;采用边缘检测Canny算子对分割后的图像进行轮廓提取;在获取的马体轮廓上,对比动物特征点的Harris角点检测算法,提出动态网格的测点标定方法,完成马体尺特征点的数据标定,同时部分解决了马体站姿与摄像头不平行带来的体长修正问题;比较Regress及Polynomial的多元线性回归方式,量化、完成马体尺数据中胸围、管围的数据拟合及三维预测,并以像素为640*480两匹伊犁马体图像为例,获得了体尺测量结果。结果表明,基于深度学习和图像测量技术,可有效进行伊犁马体尺的自动测量并将其误差控制在较小范围之间,就大体型动物的体尺测量技术而言,该研究具备范例参考意义。
Taking the Yili horse as the research object,the key technologies in the measurement of horse body size are studied by image acquisition and field survey.Based on the example segmentation technology of YOLACT,the fast and high performance segmentation of horse body and background is completed in MS-COCO data set.The Canny operator of edge detection is used to extract the contour of the segmented image.Compared with Harris corner detection algorithm of animal feature points,the method of measuring points calibration of dynamic grid is proposed to complete the data calibration of horse scale feature points.At the same time,the length correction problem caused by the non parallel of the horse body and the camera is partly solved.By comparing the multiple linear regression methods of Regress and Polynomial,the data fitting and three-dimensional prediction of chest circumference and tube circumference in horse body size data are quantified and completed.Then taking the image of two Yili horses with the pixel of 640*480 as an example,the body size measurement results are obtained.It is shown that the automatic measurement of Yili horse’s body size based on depth learning and image measurement technology can be effectively carried out and its error can be controlled in a small range.As far as the measurement technology of large animal’s body size is concerned,this study has a reference value for example.
作者
张婧婧
ZHANG Jing-jing(School of Computer and Information Engineering,Xinjiang Agricultural University,Urumqi 830052,China)
出处
《计算机技术与发展》
2020年第11期180-184,189,共6页
Computer Technology and Development
基金
新疆维吾尔自治区高校科研计划项目(XJEDU2020Y020)。
