Defect Detection in Printed Circuit Boards Using Leaky-LeNet 5

Abstract

Each electronic device includes printed circuit boards (PCBs), where defect detection is an important process to enhance the quality of PCB production. To accomplish error-free PCBs, the researchers and experts converted traditional manual inspections into automated systems. The manual inspection results are ineffective, where the non-defective PCBs are classified as defective PCBs. A subsequent study added a technique called LeNetwork-5 (LeNet-5) and speeded up robust feature extraction (SURF) techniques to identify defects. The existing method was unreliable, so further research was conducted in this area using Leaky techniques. Two objectives are achieved using SURF technique: (1) registration of the PCB to be checked against a reference PCB and (2) identification of feature points on the PCB helps to find missing components. Additionally, the Leaky LeNet-5 classifier is employed to classify the defects in PCBs. The proposed method achieved accuracy of 99.83%, sensitivity of 96.26%, specificity of 99.15%, and F1-Score of 98.45%.