Inspection And Recognition Of Components On Circuit Boards

Circuit board is an important part of modern electronic equipment, its quality affects the performance of products. Online inspection and error correction can improve the quality of products and reduce waste. The process of circuit board inspection is done manually all the time. Since people's ability is limited and the number of components on one board becomes higher, new approaches that can examine components automatically are needed.This paper proposes a new method for automatic component inspection by image processing technology. A circuit board inspection system is built, which is based on HSI color model. The concerned problems such as location, image matching and defect determination are solved. Based on the above method, the image characteristics of components, such as chip capacitors and IC chips etc., are researched. According to the research,' another inspection method is proposed, which is based on component characteristics. This method depends less on location accuracy, and can detect images of different resolution. The computation quantity is reduced, as the area without components is omitted during the inspection process.Recognition of nominal values on components is also an important task. In the process of character recognition, feature selection is a basic step. As translation, scale and rotation invariant features, the Orthogonal Fourier-Mellin Moments (OFMMs) are robust and have good ability of image classification and reconstruction. They have been widely used in target recognition and image classification. Among the numerous moments, how to select the most effective moments as features to classify images is a valuable subject. Based on the characteristic of OFMMs and Principal Component Analysis(PCA), this paper researches the selection of moments in image recognition and classification. In the experiments, the characters gathered from circuit boards are classified with this method. The results show that this method can improve the correct rate of image recognition effectively and has better performance than traditional methods.