Applied Haar Cascade and Convolution Neural Network for Detecting Defects in The PCB Pathway

摘要

The examination of the PCB pathway in Indonesia currently still uses human labor to look for defects, so each PCB must be checked one by one. This requires considerable time to check the amount of produced PCB. With the times, there are various digital image processing techniques that can be developed for quality control processes, including checking the quality of a product quickly and accurately. In this research, the Haar Cascade and Convolution Neural Network method were applied to an affordable mini PC. The result shows this low-cost mini PC has optimal performance for the PCB checking process. From a total of 1344 training data, the system is able to correctly detect the condition of the PCB as many as 1330 data or more than 99% while for integration testing on a mini PC, the system is able to produce accuracy up to 90%.