文档介绍：南京航空航天大学
硕士学位论文
基于图像分析的机器油液污染在线监测系统研究与开发
姓名:孙福珍
申请学位级别:硕士
专业:交通信息工程及控制
指导教师:左洪福
20080101
南京航空航天大学硕士学位论文
摘要
本文主要研究油液污染在线监测系统中油液污染度测量和磨损颗粒分类的
问题,以及为了对某个发动机轴承试验机磨损颗粒在线监测而开发的油液颗粒污
染分析系统。本文主要开展以下几个方面的研究工作:
。针对微流动油液运动图像的特点,本文提出了基于卡
尔曼运动模型模糊图像恢复以及基于背景分割技术的运动图像目标提取方法。
。首先,根据油液污染度评定方法和等级划分,本文讨论
了不同污染度等级判断标准的尺寸特征参数的提取与计算方法;其次,根据微管
道油液速度场分析,提出运动油液的体积计算方法。最后,从颗粒大小和颗粒个
数两个方面对系统可能的误差来源进行分析。
。在进行磨粒显微形态学特征分析基础上进行
分类特征参数优化选择,从磨粒分类的准确性和快速性两个角度出发,提出了支
持向量机与最近邻相结合快速识别算法。
。首先,在实验室现有的条件下,改进了
系统的硬件配置,开发了油液颗粒污染度分析软件系统;其次,分析了系统若干
关键问题;最后,对油液污染度计算和磨损颗粒识别方法进行对比实验,验证本
文提出方法的有效性。

关键词: 在线监测图像处理污染度测量磨粒分类误差分析

i
基于图像分析的机器油液污染在线监测系统研究与开发
Abstract
In this paper, the problems of oil pollution examination and wear debris
identification are studied in oil pollution on-line monitoring system, furthermore, an
analysis system of oil particles pollution is carried out for monitoring the wear debris
of an engine bearing test. Based on the engineering background, the main research
work is introduced as follows:
1. Research on the methods of image processing. According to the characteristics
of micro-flow oil, a new method for image quality descending analysis to the motion
blur images based on the Kalman motion model is presented and a new method for
target extraction is proposed based on the background segmentation technology in the
paper.
2. Oil pollution examination. Firstly, according to the evaluation method and
grading partition, different dimension parameters are calculated based on different
judging standards of pollution grading; Then, a new method for calculating
micro-flow oil volume is presented according to oil speed field analysis in
micro-piping; Finally, the possible errors of system are discussed from two aspects of
particle size and number.
3. Research on the wear debris identification. The classification paramet