基于图像分割的马体尺测量系统设计-《计算机技术与发展》

文章信息/Info

Author(s):: ZHANG Jing-jing; CHENG Yun-tao; School of Computer and Information Engineering,Xinjiang Agricultural University,Urumqi 830052,China

Keywords:: YOLACT; COCO data set; Matlab simulation; body measurement; linear regression

摘要:: 基于马体与背景的图像分割技术,以马体测量系统为设计目标,展开其体高、体长、胸围、管围指标的机器测量与系统设计研究。在图像处理部分,采用单阶段的实例分割模型YOLACT,在MS-COCO 数据集上实现马体与背景的分割; 在数据测量部分,借助 Matlab 平台中提供的多种方法及库函数,依次完成了马体尺测量系统的界面设计、图像导入、坐标定位、距离测量、回归分析、测量值筛选、数据显示及保存等内容。通过引入胸径、管径的二维参数,建立马体胸围与管围的三维预测方程;采用图像的局部放大功能,加强马体特征点定位的准确性;在验证马体尺测量值的有效性时,采用 rcoplot方法进行数据的残差分析;借助 Matlab-GUI 工具,进行体尺测量系统的可视化设计,为马体尺的快速获取及智能化测量提供典型范例。

Abstract:: Based on the image segmenting technology of horse body and background,taking the horse body measurement system as the design objective,the machine measurement and system design of its body height,body length,chest circumference and pipe circumference are studied. In the image processing,YOLACT,a single-stage instance segmentation model,is used to segment horse body and background by using MS-COCO data set. In the data measurement,the interface design, image import,coordinate positioning,distance measurement,regression analysis,selection of measurement values and data display of horse body measurement system are completed successively with the help of various methods and library functions provided by the platform of Matlab. By introducing the two-dimensional parameters of breast diameter and pipe diameter,the three-dimensional prediction equation of horse’s chest and pipe diameter is established;the accuracy of calibrating horse’s feature points is enhanced by using thelocal enlargement function ofimage;theresidual analysis of data is carried out by rcoplot method when verifying the validity of horse’s measurement value;and the visualization design of horse’s body measurement system is carried out by means of the tool of Matlab-GUI. It provides a typical example for fast acquisition and intelligent measurement of horse’s body size.