Since burrs are small and often occur inside complex structures, it is very difficult to automate their detection. How to prevent burrs from being passed on to the next process has been a major manufacturing challenge. In the past, an inspector used an industrial endoscope to visually inspect all hydraulic valves for burrs. In this case, more than 100 people per plant were involved in this inspection work.
In this case, we applied a 3D restoration algorithm based on image matching to images obtained from an industrial endoscope camera. However, our existing algorithm was initially insufficient in computing speed, making real-time measurement difficult. To solve this problem, we reviewed the algorithm from the mathematical model and developed an ultra-fast algorithm that was 100 times faster than the conventional algorithm. We also implemented an automatic burr detection algorithm and burr inspection & measurement software using threshold value.