Application Research On The Detection Of Engine Valve Getting Into Die With Machine Vision Technology

Valve is an essential part of mobile engine, automizing its prodution line is of greatimportance to the development of domestic mobile industry. This paper based on the valveelectric upsetting and forming equipments in a factory in Guangdong province, did someresearch on the detection problem of valve getting into die. In order to solve this problem, thispaper using Machine Vision technology, raised two programs:static detection program basedon Image Procession technology and dynamic detection program based on Object Trackingand Positioning technology.This paper designed the hardware structure of these programs, and finished their typeselection after analysing the function requirement of these programs. After that, this paperpresented the methods on how to realize these two programs. Regarding to the static detectionprogram, in order to adapt the complex detection environment and fulfill diverse detectiointests, this paper raised a software architecture combining an image procession interfaceprogram and vision detection commands in robot control system after analysing the softwarefunction requirement of this static program, and then presented the way on how to realize it.Regarding to the dynamic detection program, the most important point lied in the research ofits key algorithms. In order to track and position the valve when it was carried into the die,this paper raised a tracking algorithm based on Mean Shift and a positioning algorithm basedon BP neural network. At last, this paper carried on some experiments on these two programs.The results showed that the static program can detect the state of valve accurately when it wascarried into the die, and also the dynamic program was effective. After comparing theadvantages and the weaknesses of these two programs, this paper drew a conclusion that thedynamic detection program had more using value and was more suitable to the users than thestatic detection program.This paper did initial research on the application of Machine Vision technology on valvedetection problem, it not only provided some valuable experience to subsequent research, butalso provided some research ideas for the application of Machine Vision technology on othersdetection problems.