Machine Vision Based Inspection Technology Research Of Surface Mounted Devices

Automatic Optical Inspection (AOI) of Surface Mounted Devices (SMD) on printed circuit board (PCB) is a significantly important research aspect in the quality inspection domain of the electronic industry assembly line. It aims at recognizing distinguishing electronic components and inspecting the default of them. Using the optical and computer vision technology, AOI system can finish all the non-contact inspection items on the high density circuit board, which adoptes the miniaturization component, accurately at a high speed. This can conquer shortcoming bringed by human inspection, such as low efficiency and instability. Finally the production efficiency and passing rate of electronic products are enhanced dramatically.Basing on investigation of industry status and development of PCB components automatic optical inspection, this study proposes the main frame of the inspection system. Utilizing the analysis pointing the main modules, this study advances some principles about designing and establishing experiment platform. A simple hardware experiment system is constructed to verify the algorithm rationality, combining with the VC++ software platform.Utilizing the characteristics of color image in the Munsell color system, this paper integrates image hue, saturation and intensity information to locate valid information area on electronic components rapidly. Due to the character feature on the surface of resistance, this thesis uses gridding technic and statistical appearance modeling approach to achieve optical character verification of resistance by machine learning.Thanks to capacitance can be differentiated by the color information, using hue value, which is insensitive to illuminance, can realize stable classification and recogni zat ion.This paper proposes an easy and practical method to locate the Mark point quickly. Then implement rapid component partition basing it. By means of the grayscale inertia moment theory, this system calculate chip rotation angle through the contour information of SOP (Small Outline Package) chips. Utilizing the position coordinate of polar circle which is on the surface of QFP (Quad Flat Package) chips, the algorithm can finish angle measurement to this kinds of chips. Making use of the processing result, the system can achieve deflexion correction and character partition.Finally, this study adopts vector summation of circular projection operator and Zernike operator to realize rotation invariant inspection to the rotated chip mark. This approach solves invalidation problem which is normal in traditional template matching due to image roration, increases robustness of template matching.