多粒子运动追踪算法污染度检测技术研究  

Research on Contamination Detection Technology Based on Multi Particle Motion Tracking Algorithm

作　　者：王涵 左洪福[1] 刘珍珍 耿聪 WANG Han;ZUO Hongfu;LIU Zhenzhen;GENG Cong(Civil Aviation College,Nanjing University of Aeronautics and Astronautics,Jiangsu Nanjing 211106,China;Aviation Industry(Xinxiang)Metrology and Test Science Technology Co.,Ltd.,He’nan Xinxiang 453000,China)

机构地区：[1]南京航空航天大学民航学院,江苏南京211106 [2]航空工业(新乡)计测科技有限公司,河南新乡453000

出　　处：《机械设计与制造》2025年第1期1-4,共4页Machinery Design & Manufacture

基　　金：国家自然科学基金项目(U1933202)。

摘　　要：油液污染度监测机械设备的常规检测手段之一,基于磨粒图像分析原理的污染度检测计数是其中广泛应用的一种技术。为得到高精度和高重复率的污染度检测结果,避免对连续图像样本中的相同粒子重复计数,提出一种基于多粒子追踪算法的颗粒计数方法。首先应用高斯混合模型背景差分进行运动颗粒提取。然后运用卡尔曼滤波进行目标预测和运用匈牙利算法进行目标匹配,得到相应的颗粒计数值。最后对污染浓度进行校正,得到污染度计量结果。使用含标准颗粒物实验样本油进行测试,最终结果均满足JJG 1061-2010标准,其中粒径>4μm、>6μm、>14μm颗粒测量值的计数误差和计数重复度均满足标准要求,证明了所提出的检测算法的有效性。Oil contamination monitoring is one of the conventional detection methods of mechanical equipment.Contamination detection and counting based on wear particle image analysis principle is a widely used technology.In order to obtain the pollu⁃tion detection results with high precision and high repetition rate,a wear particle counting method based on multi particle track⁃ing algorithm is proposed in this paper.Firstly,Gaussian mixture model background difference is used to extract moving par⁃ticles.Then,Kalman filter is used for target prediction and Hungarian algorithm is used for target matching to obtain the corre⁃sponding particle count value.Finally,the pollution concentration is corrected to obtain the measurement results of pollution de⁃gree.The test is carried out with the test sample oil containing standard particles,and the final results meet the JJG 1061-2010 standard.The counting error and counting repeatability of the particle size>4μm,>6μm and>14μm particle measurements all meet the standard requirements,proving the effectiveness of the detection algorithm proposed in this paper.

关 键 词：颗粒计数 多粒子运动追踪 污染度检测 

分 类 号：TH16[机械工程—机械制造及自动化] V519.2[航空宇航科学与技术—航空宇航推进理论与工程]