基于边缘的图像分割在牛体尺测量中的应用

Application of Edge-based Image Segmentation in Cow Body Measurement

图像分割是图像处理、分析和理解的基础,目前它已经成为机器视觉研究领域最活跃的课题之一。边缘检测能勾画出目标物体,蕴含丰富的信息,是图像分割、识别和分析中抽取图像特征的重要方法。本文通过常用算子如Sobel、Roberts、Prewitt、Gauss-Laplace和Canny算法对牛体图像分割的效果进行实验对比,证明Canny算子总体上优于其他算子。针对Canny算子分割可能产生的断裂和不完整,运用OR运算结合模糊和边缘信息去除断边,利用数字形态学重新填充图像中的空洞,增强边缘轮廓显示的效果。在此基础上进行的牛体测量结果与实际结果相比误差较小,测量精度较高,其通用性较好,能够满足日常对于牛体体高和体长的测量要求。

Image segmentation is the basis of image processing,analysis and understanding,and it has become one of the most active topics in the field of machine vision.Edge detection can outline the target object and contain rich information.It is an important method to extract image features in image segmentation,recognition and analysis.In this paper,the common operators such as Sobel、Roberts、Prewitt、Gauss-Laplace and Canny algorithms are compared to compare the effects of bovine image segmentation,and it is proved that Canny operator is superior to other operators in general.In view of the possible fracture and incompleteness of Canny operator segmentation,OR operation combined with blur and edge information is used to remove broken edges,and digital morphology is used to refill holes in the image to enhance the effect of edge outline display.On this basis,compared with the actual results,the measurement results of cattle body have smaller error,higher measurement accuracy and better generality,which can meet the daily measurement requirements of body height and body length.