Application of CNN-Based Method for Automatic Detection and Classification of the IC Packages

Abstract

Automatic detection and classification of integrated circuits' packages is one of the methods supporting the traditional production of electronic parts based on through-hole technology, typical for Printed Circuit Boards (PCBs), utilizing the advantages of modern machine vision solutions. As a result of the growing availability of cameras and 3D printers, as well as the popularity of IoT systems, prototyping of some electronic circuits with the use of simpler robotic systems may be supported by an automatic analysis of electronic components based on machine vision. Considering recent advances in the applications of Convolutional Neural Networks (CNNs) in computer vision, their applicability for this task has been analyzed and experimentally verified in this paper, also in comparison to previously proposed approach based on handcrafted features.