基于显微图像识别的在线润滑油中磨粒分析方法  

On-line Analysis for Particles in Lubricating Oil Based on Micro-image Recognition Method

作　　者：郝延龙 何红坤 常青[1] 严志军[1] 潘新祥[1] 朱新河[1] 程东[1] 马春生[1] 

机构地区：[1]大连海事大学轮机工程学院,辽宁大连116026

出　　处：《润滑与密封》2016年第5期59-64,共6页Lubrication Engineering

基　　金：国家自然科学基金项目(E091004)

摘　　要：针对现有显微图像油液磨粒监测技术的芯片结构的不足,改进微通道与外通道结合处结构,减小了微通道被阻塞的可能性,提高了检验精度。为了提高机器润滑油中金属磨粒在线图像识别的准确性,分析图像运动模糊退化模型,并以自相关函数估算模糊尺度,采用基于维纳滤波的图像恢复算法获得清晰的磨粒图像;应用图像差值与最大类间方差法(Otsu)对图像进行分割并对分割后的磨粒图像计数获得磨粒等效直径和分布。研究结果表明,该图像分割方法提高了油液磨粒的在线实时检测的准确率,其精度达95%以上。Aimed at the shortage of microfluidic chip structure,the channel junction's structure was improved,the likelihood was decreased that the micro channel was blocked,and the inspection accuracy was improved. To improve the accuracy of the lubricating oil pollution level analysis and the metal particle image identification,all motion blur image degradation model was analyzed; the autocorrelation function was used to estimate the fuzzy measurement; the second Wiener filter was applied to restore blurred image; the wear particle image segmentation method based on lubricating oil background image and Otsu was applied; the last the numbers of the particles were counted and the equivalent diameter was calculated.Experiments show that the online real-time detection system for oil detection has very good effect. Its precision is above95%.

关 键 词：在线检测 磨粒 模糊尺度 图像分割 

分 类 号：TH117.1[机械工程—机械设计及理论]