Automated PCB identification and defect-detection system (APIDS)

Shams Ur Rehman, Ka Fei Thang, Nai Shyan Lai

Abstract


Ever growing PCB industry requires automation during manufacturing process to produce defect free products. Machine Vision is widely used as popular means of inspection to find defects in PCBs. However, it is still largely dependent on user input to select algorithm set for the PCB under inspection prior to the beginning of the process. Continuous increase in computation power of computers and image quality of image acquisition devices demands new methods for further automation. This paper proposes a new method to achieve further automation by identifying the type of PCB under inspection prior to begin defect inspection process. Identification of PCB is achieved by using local feature detectors SURF and ORB and using the orientation data acquired to transform the PCB image to the reference image for inspection of defects. A close-loop system is produced as a prototype to reflect the practicality of the idea. A Graphical User Interface was developed using MATLAB to present the proposed system. Test data contained 29 PCBs. Each PCB was tested 5 times for camera acquired images and 3 times for database images. The identification accuracy is 98.66% for database images and 100% for images acquired from the camera. The time taken to detect the model of PCB is recorded and is significantly lower for ORB based identification than SURF based. The system is also a close loop system which detects defects in PCB units. The detection of defects has highest accuracy of 92.3% for best controlled environment scenario. With controlled environment, the system could detect defects in PCB pertaining to smallest of components such as SMDs.

Keywords


machine vision; PCB defects; PCB identification; SURF; ORB

Full Text:

PDF


DOI: http://doi.org/10.11591/ijece.v9i1.pp297-306
Total views : 670 times


Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

ISSN 2088-8708, e-ISSN 2722-2578