Research And Development Of Measuring System In Pump Testboard Based On Image Recognition

The fuel injection pump testboard is used to correct the injection characteristics of thefuel injection pump of diesel engine. Nowadays, the measurement of fuel injection quantityis implemented by the method of artificial reading, which leads to inefficiency in acquringinjection quantity, so as to make more serious errors. In this paper, the fuel pump testboardfuel detection system, which based on image recognition, is developed. It can realize theautomation of acquiring fuel injection quantity, which could improve the efficiency of fuelmeasurement, and reduce the errors in the same time.The fuel detection system based on image recognition is studied and developed in thisdissertation, and it can capture fuel injection quantity accurately; the AM1310CCDcamera is selected as the image sensor for its high resolution and stable performance meetthe needs of this system; the image processing algorithms is simulated using MATLAB toselect the best image processing algorithms, and after comparing the effect and efficiencyof image filtering algorithm, median filtering method is selected for image preprocessing;it is found that there is more boundary information detected by Canny operator, Cannyoperator is used to extract containers’ boundary information, and the positions of thecontainers can be found in the image using the boundary information; OTUS imageclassified segmentation is used to detect the fuel level of the container, and then we get theheight of the fuel level in the container, so that we can calculate the quantity of the fuel inthe container; the fuel detection system based on image recognition is installed in the fuelpump testboard, and a PC software is developed by LabVIEW and embedded by imageprocessing dll. As a result, the measurement of fuel charge has been realized, as well as thecontrol of test board and the acquirement of related signals.This image recognition system has been tested for several times in laboratory, and theexperiments show that, with an error of±0.05ml which is smaller than artificial reading, the fuel charge measurement system gives an accuracy result against real value. Meanwhile,this measuring system automated the measurement of fuel charge and improved itseffectiveness. Thus the method based on image recognition is reliable and validate.