Key Technology Of Image Inspection And Processing In PCB Automatic Optical Inspection

The quality of printed circuit board (PCB) is required higher than ever with the rapid development of electronics technology. It could not produce high-quality PCB without automatic optical inspection (AOI). And image inspection and processing is the important basis of PCB AOI. Many interests have been focused on this field in some developed countries or regions such as Taiwan, Japan and some western countries, and some achievements have been widely used in some fields. So the study of image inspection and processing in PCB AOI is very important.Considering the potential applications of image inspection and processing in PCB AOI and the larger gap of the researches between the mainland of China and those countries or regions mentioned above, some key technologies of the inspection and processing of bare PCB image have been studied in order to improve the level of study in the field, and they have been applicated in the study of the inspection hole machine and the appearance inspection machine. The main contents and originality are listed as follows:1. On the basis of nature suture algorithm of fade in and fade out, an algorithm of image automation mosaic based on hardware has been proposed. In an experiment, the whole and massive bare PCB AOI image has been captured by the algorithm proposed in this dissertation, and the velocity of bare PCB has reached to 5 m/min. A variance filter based on mean denoising and image enhancement (VMDIE) has been presented for the sake of denoising and enhancing in PCB AOI image. Contrast experiments have verified that the results of the filter were better than that of the other filters mentioned in this dissertation. Two kinds of algorithms of PCB AOI image enhancement have been proposed. One was image enhancement of combining removal of gray redundance (RGR) with gray transform (GT), the contrast of image after being enhanced by the algorithm was 1.6073 times the value of original image, and the other was that of merging RGR and butterworth high pass filter based on Fourier transform (BHPBFT), the results of enhancement were better than that of other comparative algorithms mentioned in this dissertation.2. An improved method of piont Hough transform (PHT) in evaluating the parameters of circle based on Hough transform and PHT has been proposed. The comparative experiments have been executed in the circle inspection of PCB AOI image by using the three methods mentioned aboved. They showed the inspection precision of the improved method was higher than that of Hough transform and PHT, and the velocity was 50.1979 times and 1.6765 times the value of Hough transform and PHT respectively. In order to detect the centre and radius of concentric circles effectively in PCB AOI circle hole image, the methods of point sets selection of edge points in PHT and the image scan direction of searching the point sets have been improved by using the geometrical characteristic of the concentric circles based on Hough transform. In the comparative inspection experiments of PCB AOI orientation circle hole image, the inspection method proposed in this dissertation had some obvious advantages, and the velocity was 3.3305 times the value of X. Cao's method and 3.3431 times that of S. Margarida's method. In accordance with the respective advantages of randomized Hough transform (RHT) and circle fitting based on least square method (CFBLSM), a method of circle inspection that combines both advantages has been presented. The real edges of defective circle hole of PCB AOI microscope image were obtained as far as possible when we applied the method to detect defective circle holes which include fuzziness and noise.3. Nonuniform illuminative PCB AOI image was enhanced with homostasis filter, then an improved Otsu method has been proposed by combining the advantage of Otsu method. Contrast experiment has been executed by using different methods in PCB AOI image with nonuniform illumination ecdysis phenomena, and the results of the improved method were better than that of other methods mentioned in this dissertation.4. Coordinate transform of image space between distorted image and the corresponding corrected image has been executed with two kinds of methods. After adopting least square method (LSM), polynomial fitting and gray restoration of bilinearity interpolation, corresponding coordinates of control points gemination have been found between the distorted image and the corrected image by adaptive control relationship. The distorted PCB AOI images have been corrected effectively, and the errors of mean square were about 0.12. 5. In order to detect the information of lines in PCB AOI image with noise and fuzziness more accurately, a method of line inspection combining RHT with LSM has been proposed. The inspection experiments have been executed by detecting the lines of PCB AOI images, and the results of the method proposed were better than that of RHT by the comparison of experiments.