一种基于迭代直径圆检测的大气颗粒物图像分割方法

An image segmentation method of atmospheric particulates based on iterative diameter circle detection

大气颗粒物图像的分割是研究大气颗粒物形态特征的重要前提,针对球形大气颗粒物的扫描式电子显微镜(scanning electron microscope,SEM)图像,提出了一种基于迭代直径圆检测的图像分割方法,首先获取大气颗粒图像的边界,并利用迭代直径抽样策略检测颗粒物图像的边界,去除边缘噪声;然后利用细化方法计算可能圆的圆心和半径,检测可能圆获得候选圆;最后利用检测到的候选圆分割球形大气颗粒图像。选取4幅球形大气颗粒的SEM图像进行分割实验,结果表明所提方法的平均分割准确率可达到94%以上。

Segmentation of atmospheric particle images is an important prerequisite for studying geometrical morphology.An image segmentation method based on circular detection using iterative diameter is proposed for scanning electron microscope(SEM)images of spherical atmospheric particle.In the method,boundary of atmospheric particle image is first obtained and boundary of atmospheric particle images is detected using iterative diameter sampling strategy.Then,center and radius of potential circle is obtained by refined method and potential circles are detected to get candidate circles.Finally,segmented image of spherical atmospheric particle is obtained.In order to verify validity of this algorithm,four SEM images of atmospheric particulate matter were used for segmentation experiments.The experimental results show that average segmentation accuracy of this method is up to 94%.