Image Filtering Of Engine Fuel Spray And Software Implementation

A large number of spray images were obtained by direct photography,and it was necessary to manually select some of the better spray images from the large number of images into the particle measurement system for measurement of the spray characteristics.In order to reduce the subjective error and the workload of artificial selection,this paper contrasts and observes a large number of spray images,and then uses the computer program to automatically filter the spray image by setting different screening criteria.In order to improve the screening efficiency,the screening process of the spray image is divided into two steps.The preliminary screening mainly used the spray image gray scale peak,the image local brightness difference value,the image sharpness value as the standard screening.In order to better select the screening index value,all the spray images are counted before the initial screening.Finally,the spray particles(particle number and particle uniformity)are used as the screening index,so the spray image needs to be pre-processing and threshold processing to facilitate the identification and statistics of sprayed particles.Among them,in the uniformity of the particle,the coefficient of variation(the ratio of the standard deviation to the mean)is used as a measure,and the coefficient of variation is not dimensionally more objective when comparing different spray images.Based on MATLAB / GUI development of spray image screening interface,the main functions are image reading and display,screening index value statistics,single spray image value query,spray image preliminary screening,image preprocessing and threshold processing,spray image final screening.This software is used to filter all the spray images taken in a kind of working condition.The results of software screening and artificial screening are compared,and the results are more consistent,which indicates that the screening program has certain applicability.Finally,the filtering effect of different threshold processing is compared,and the result of iteration threshold is closest to the Malvern data.