Design And Research Of An Automatic Optical Inspection System For PCB

With the development of Surface Mounting Technology, the density of Surface Mounting Components is higher and higher, and the size of components is getting smaller and smaller. In order to adapt to this trend, Automatic Optical Inspection technology is applicated in the quality inspection of Print Circuit Board more and more widely, and the AOI is becoming a core device in SMT production line.The functional modules of Automatic Optical Inspection system are researched and an AOI system for PCB is designed in this paper. The PCB detection algorithm solutions are proposed focusing on the image processing core module of AOI system.A method based on least-squares and Sub-pixel edge detection is proposed for the Sub-pixel calibration of the most commonly used circular PCB Mark. In order to to enhance the adaptability of the algorithm, the non-circular interference targets are excluded by geometric characteristics. In addition, Douglas-Peucker algorithm is introduced for the calibration of polygon PCB Mark.The performance of time-consuming template matching based on the Normalized Cross Correlation is improved with the mathematical characteristics, "Box filtering" and image pyramid technique for the locating of Surface Mounting Components of PCB, without loss of accuracy but much faster computing speed. For the locating position with sub-pixel precision, according to the condition of the correlation peak, the outline of the longitudinal section along the correlation peak is obtained by cubic spline interpolation for the optimum position, which is also the position of the Surface Mounting Components with Sub-pixel precision.In order to achieve the inspection of the solder of IC chip pins and the measurement of the deflection angle of Surface Mounting Components, several new blob detection algorithms are proposed with the function of main axis angle computing. For the character recognition of resistor, four character recognition methods are introduced in this paper, and the correction of rotation, glitches and width is achieved. The color features in HSV color space and the distance of color are used for capacitance detection. In addition, the mosic algorithm is used for the thumbnail of PCB, and a learning- mechanism is proposed in this paper to make the system have ability to learn.