Air-tightness detection system based on optical flow method

WU Chun-long， PAN Hai-peng， XIA Yong-ming
（Academy of Mechanical Engineering & Automation， Zhejiang Sci-Tech University， Hangzhou 310018， China）

Abstract： Aiming at solving potential safety hazard and low efficiency of using the common bubble method to detect minitype pressure vessel，the bubble image processing method was studied. An image capture system was built. The Horn-Schunck method based on the optical flow concept was researched to deal with the bubble image in Matlab. A well-bedded bubble image was gotten by using customized threshold of image segmentation region. An semi-automatic air-tightness test system based on machine vision was devised. The number and the size of bubbles which was used to account the leakage rate and leakage part was caught automatically by this system. The results indicate that，the minimum bubble diameter detected by this system is 4 mm. The most appropriate diameter is 8 mm. From then on the surface of bubble will get more irregular and the detection precision will drop as the diameter increase. So this system is successful in counting the number and the size of bubbles in certain conditions. It can realize the gas leakage detection.
Key words： optical flow method； air-tightness detection； machine vision； high speed camera system