Inspection of machining defects on mechanical parts using a computer vision system

Mechanical parts are currently mass produced by machines that perform high-speed machining. The mechanical parts will be assembled in a machine with other components, and therefore there will be holes for screws, grooves etc… These parts must have a precise shape. If the machining and drilling are not done in the right place or with a small dimensional error in this case the assembly process cannot be completed. The manuel inspection is cumbersome and the efficiency is low. To overcome this problem, the non-contact defect detection process is the subject of this research. We proposed a real-time measurement technique using edge detection, dilation and erosion algorithms to remove noise and enhance the collected part image to make the information clearer, then compare it with the perfect model. This technique uses the OpenCV library and a Raspberry Pi 4 board. The experimental results show that the defect on the part can be effectively inspected, located and recognized.