The Effect Mechanism Of Optical Properties Of Printed Circuit Board On Its Automatic Visual Inspection System

The automatic visual inspection(AVI) system of PCB(Printed Circuit Board, PCB) is the key inspection system in the PCB production and has been widely used in developed country. It integrates high uniformity illumination system, high-speed line scan cameras acquisition systems, high-precision servo drive system and the corresponding image processing algorithms and software technologies into one system. Research and development in the field of automatic visual inspection system are active in Israel, the United States, Japan and other developed countries. Some results have been commercialized. But due to the lack of fundamental research in the effect mechanism of PCB to its inspection performance, there is a gap between domestic and developed countries in the research level of this system.In consideration of the importance of automatic visual inspection in both industry and academia, This dissertation presents the research of the mechanism of PCB optical properties affects the inspection performance of its automatic visual inspection system. Furthermore, it also presents the proven method to improve the inspection performance. These works improve domestic research level in the field of automatic visual inspection and enable the competitive products of our own. In this dissertation, the major research works and innovations are as follows:(1) A high uniformity illumination method based on the optical properties of PCBAfter the study of several commonly used metal surface reflection model. This dissertation selects the Phong model as the reflection model for PCB. Based on this model, a new high uniformity illumination method for PCB visual inspection had been proposed. It can eliminate the image shadow fringes and reachs high defect contrast. Experiments show that the lighting configuration can improve the defects detection rate in about 15%, from 79% to 94.7%. The flexibility of this method used in different materials is proven through simulation.(2) Illumination assessment and optimization method based on image analysis in RGB color spaceThe image requirements in the system had been analyzed in detail, then an illumination assessment and optimization method had been proposed. This method based on image analysis in RGB color space. It mainly contains a merit function, which consists of three components: major color distance, color fluctuations and contrast of prior knowledge of defects. Experiments show that this merit function can quantify the effect of illumination design; furthermore, it can also used in illumination optimization. The experiment also proven the contrast of minor defects is enhanced. This method is suitable for visual defects inspection system.(3) A physics-based defects model and inspection algorithm for visual defectsIn order to solve the problem of false and missed alarm, this dissertation proposed a new physics-based defects model and inspection algorithm for automatic visual inspection, denoted as the localized defects image model(LDIM). It is using a local defect merit function to quantify the cost that a pixel is defective. This function comprises two components: color deviation and color fluctuation. Parameters related to statistical data of the background region of images are also taken into consideration. Test results demonstrate that the model matches the definition of defects, as defined by international industrial standards IPC-A-610 D and IPC-A-600 G. Furthermore, the proposed approach enhances small defects to improve detection rates. Evaluation using a defects images database returned a 100% defect inspection rate with 0% false detection, at its best performance. Proving that this method could be practically applied in manufacture to quantify inspection standards and minimize false alarms resulting from human error.(4) A reproducibility optimization method of imaging system using linear variable metallic neutral density ?lterSince the stability of inspection performance is only depends on the reproducibility of image acquisition system. Therefore, in order to reduce the variation of inspection performance of the system, an evaluation and optimization method for line scan machine vision system has been proposed in this dissertation. This method use a linear variable metallic neutral density filter as the object, and a two steps curve fitting is use to get the reproducibility parameters. Test results demonstrate that the method is applicable to quantity the reproducibility of imaging system. Furthermore, the method can also be used for optimization. The final optimization result in this work is about 0.5 pixels, which is 5 times better than its initial performance.(5) The effect mechanism of optical properties of PCB to its inspection performance, with a new design ruleBased on previous research works, from the perspective of defects information from illumination to camera, an effect mechanism of optical properties of PCB on the inspection performance of its AVI system is summarized. A new design rule of automatic defects inspection is proposed. This rule based on the image parameters distribution of defective and non-defective region. This rule aims to model the design goal of automatic defects inspection system, namely, minimize the overlap of these two distributions. Using this rule, researcher can build and optimize the system to its optimum performance. The definition of minimum detectable defect is also discussed.