Research On Processing And Analysis Techniques Of Printed Circuit Board Three Dimensional CT Image

With the development of electronic technology, Printed Circuit Board(PCB) has become an important component of sophisticated electronic equipment in modern industry. PCB defects generated in sophisticated electronic equipment’s producing and using processes will severely affect the equipment’s function and performance. Hence, PCB detection methods have received more and more focuses and become one of the research hotspots. Cone-beam Computed Tomography(CBCT) based PCB three-dimensional non-destructive testing method can obtain the inner structural information of PCB in three-dimensional space with micron-sized resolution, thus provide a three-dimensional non-destructive way for inner flaw detection, and largely expand detection range and improve detection efficiency, and has become a new means for PCB non-destructive testing. As PCB producing technique has become more complicated and the demands for PCB testing have been diversified, requirements for CBCT based automatic PCB testing are getting higher. Therefore, research on CBCT based PCB three-dimensional image processing and automatic testing of circuit connectivity has a great theoretical significance and application value.Due to the influences of all kinds of noises from the imaging process of CBCT, the distribution of gray are uneven and there are bad metal artifacts in PCB three-dimensional images. Therefore how to test the circuit connectivity of PCB in three-dimensional space is a difficult problem in three-dimensional CT image processing and the systematic study of this problem has not been reported until now. This thesis has a depth analysis on the features of the imaging system’s noises and the data characteristic of PCB image targets. By introducing the theories of image enhancement, image segmentation and target detection in three-dimensional image engineering field, three-dimensional image processing methods and automatic non-destructive testing of circuit connectivity are studied for PCB. The main researches and innovations are as follows.1. In the aspect of three-dimensional PCB image enhancement, for the features of gray level distribution and the characteristic of noises in three-dimensional PCB images, three new image contrast enhancement algorithms are proposed for three different application purposes. Firstly, according to the distribution features of the histograms of PCB images, a double histogram equalization algorithm based on the distribution feature of the maximum histogram of PCB image is proposed, which can improve the image’s contrast and enhance the image details while making sure that the whole feature of the image’s gray level is basically unchanged. Secondly, by introducing the difference of the gray statistic characteristics between the background(basal lamina) and the object(metal), a double weighted image enhancement algorithm based on gray levels and their distances is presented to enhance the contrast between background and the object. Finally, the histogram equalization to low luminance images may lose the information of high brightness and invalidate the contrast enhancement. To solve this problem, a variable weights accumulation based histogram enhancement algorithm is given, which can give prominence to the metal with high brightness in low luminance images.2. In the aspect of PCB three-dimensional image segmentation, in order to adapt the size and data distribution feature of the PCB three-dimensional image, a spectral segmentation algorithm based on weighted area mean histogram is proposed, which constructs a new kind of similarity matrix. The proposed algorithm firstly creates three-dimensional sub-regions by region growing method, and then constructs the similarity matrix on the basis of histogram of sub-region mean, and weights the similarity matrix with the probability of sub-region mean. This new similarity matrix building method reduces the matrix size and increases the divisibility of different objects. Experimental results show that the algorithm can obtain better segmentation performance.3. In the aspect of PCB three-dimensional image target detection, to address the geometric distortions and metal artifacts in PCB, precise and fast wire detection method is proposed based on PCB’s gray level distribution, structures, and geometric features of wires and priori knowledge of circuit connectivity. Firstly a correction algorithm based on the location of central plane of circuit layer is proposed to correct the curve wires caused by geometric distortions, and then a volume morphology wire detection algorithm based on wire direction decomposition, which makes use of the characteristic of linear shape of the wire in three-dimensional space, is given to detect the wire. In the processing of wire detection, the prior knowledge of circuit connectivity is utilized to judge and recover automatically the wire connectivity worsened by metal artifacts. Experimental results show that the method can acquire better wire detection performance for actual PCB three-dimensional CT image.