Study On The Technology Of Oil Pollution Class Based On Image Processing

Oil pollution levels can quantitatively reflect the degree of oil pollution and it is an important theoretical basis of oil pollution control. According to statistics, 70%-85% failure of the hydraulic system is caused by oil pollution. Therefore, the detection of oil pollution level has great significance in improving the reliability and extending the service life of hydraulic system. At present the general testing equipment have some shortcomings, such as expensive, inconvenient to operate, and low accuracy. In order to independently design a cheap, efficient, accurate and convenient particle analyzer to achieve the detection of oil pollution, the whole system is designed based on the technology of microscopic image formation and image processing. It includes two parts: the image acquisition and the image analysis. Our work focuses on how to use computer image processing technology to identify oil pollution particles in the images and gain related parameters, and then calculate the level of oil pollution. For this purpose, the following research is made:Firstly, introduce how to use learning vector quantization (LVQ) neural network in particle edge detection. According to the characteristics of oil image, an algorithm based on the average background to get the target image is proposed. Meanwhile, owing to the difference between noise and edge point, three vectors are proposed which have the feature of edge point and can restrain noise. The three vectors make up one feature vector to represent the sample image. The training samples are composed of the feature vectors and the target image. After training, the neural network can detect the edge of particles. The simulation results show that the network after training can identify the pollution particles image accurately.Secondly, do some research about the BP neural network application in edge detection. Use the same training samples and build a three-layer BP neural network. Simulation results show that BP neural network achieved better results than LVQ neural network.Finally, Use label operation, get the number and the size of each particle and calculate the level of oil pollution. In order to get more accurate level, do some research for distinguishing the droplet, bubble and particle.