Lubrication Oil Contamination Detection Tecnology Based On Image Recognition

Nowadays, mechanical equipments play an important role in production, the workingstate of machinery is inextricably linked to lubricant. In order to ensure that themechanical equipments can be maintained a steady working state, the polluted lubricant isneeded to be replaced, then the oil contamination detection technology comes into being.Detection of particulate contamination in oil is the most important means of oilcontamination detection technology. In this paper, with the help of image technologywhich is in rapid development today, visual detection method for lubricating oilcontamination is studied on the basis of literature at home and abroad, including severalissues, such as design of digital microscope system, image segmentation and accuratepositioning of tiny moving particle, the size measurement and distribution of particles inoil, etc.According to oil contamination level standards, in order to evaluate the state of oilpolution, the sizes and distribution of particles in oil are needed to be measured. In thispaper, the overall structure of visual detection system of particles in oil is designed, anin-depth discussion on depth of field of digital Microscopy Imaging system is done andthe sampling channel is designed on the basis of force analysis of particles in microsampling channel, combined with subsequent image segmentation algorithm.In the particles extracting, it gets incomplete information as the contrast rationbetween objects and background, the paper presents an algorithm combining edgedetection with gray-scale difference, this algorithm obtains the adaptive threshold withclustering method and fill the holes in the objects with morphological method. Thisalgorithm to achieve particles of rapid and relatively complete segmentation.Particles in the sample channel have the phenomenon of adhesion and overlap. Theevalution result of oil contamination will be inaccurate if the particles don’t be segmentedaccurately. Adhesion and overlappong particles are segmented by distance transform,‘peak minusing’, elimination of confusion extreme points, distance reconstruction. Thesegmented results show that the algorithm can not only solve the problem of the adhesion and overlapping particles segmentation, but also improve the over-segmentation intraditional watershed algorithm.The area and equivalent diameter of particles are recorded by Matlab software,. Theresults are classified by the oil contamination standard of NAS1638. After compared withthe test results and average results from 10 sample data, the data obtained by thisalgorithm are stability and accuracy and appropriate for oil contamination particlesdetection.