磨粒图象背景光照不均匀分布图象预处理  

The Pretreatment of Segmentation on Distributing Uneven Background Brightness of Wear Particle Image

作　　者：吕植勇[1] 严新平[1] 

机构地区：[1]武汉理工大学智能交通工程中心,武汉430063

出　　处：《润滑与密封》2005年第5期34-37,共4页Lubrication Engineering

基　　金：国家自然科学基金项目(50275111);教育部重点研究项目(重点02052教).

摘　　要：针对磨粒图片背景亮度呈现分布不均的问题,采用两种计算机图象处理方式进行了预处理,得到比较完整的二值图。第一种方式是:在没有铁磨粒情况下采集背景亮度变化,通过数字处理得到背景亮度变化的差值,然后将亮度补偿到铁谱图上。试验结果表明,该方法在边缘信息损失很小的情况下,得到磨粒二值图。第二种是:在不能得到背景光的条件下,通过一阶和二阶差分处理及其变形方法,来消除背景梯度变化,从而消除背景光的低频噪声。从试验效果来看,一阶和二阶差分不仅能进行边缘检测,而且能通过卷积运算消除背景光分布不均,磨粒边缘有点扩大,有些少量信息损失掉了,这种损失可以用图象形态学中的腐蚀的方法消除。两种方法相比,采用背景补偿的方法效果更好一些,在条件许可的情况下尽量采用背景光补偿方法。Two kinds of computer image processing methods were used for the analysis of uneven background brightness distributing of ferrography images. For the first method, the uneven background brightness distributing was compensated by collecting background brightness of ferrography image. This method has little loss of marginal information. For the second method ,the uneven distributing of ferrography images was treated through the first and the second orders operation to eliminate the change of background gradient when the background brightness of ferrography images could not been gotten. This method can get the good background of ferrography images and segment wear particle from background. The image processing test shows that the first and the second orders operation can segment image and eliminate the uneven distribute of images background, but compensated method by collecting background brightness is better than orders first and second order operation methods.

关 键 词：磨粒图象 噪声 预处理 图象分割 

分 类 号：TH165.3[机械工程—机械制造及自动化]