Automatic Hybrid Genetic Algorithm Based Printed Circuit Board Inspection

The paper presents a novel integrated system in which a number of image processing algorithm are embedded within a genetic algorithm (GA) based framework in order to provide an adaptation and better quality analysis with less computational complexity while maintaining flexibility to a broad range of defects. A specially tailored hybrid GA (HGA) is used to estimate geometric transformation of arbitrarily placed printed circuit boards (PCBs) on a conveyor belt without any prior information such as CAD data. A library of image processing functions is accessed by the HGA within an intelligent framework. These functions include operations such as fixed multi-thresholding, Sobel edge-detection, image subtraction and noise filters. The proposed framework allows novel composition of tasks such as edge-detection and thresholding in order to increase defect detection accuracy with low computational time. Our simulations on real PCB images demonstrate that the HGA is robust enough to detect any missing components and cut solder joint with any size and shape with significant reduction in computational time compared to conventional approaches