Research On Artifacts Correction For Printed Circuit Board

Electronic devices are becoming more and more important in modern society. Variety electronic elements are connected by Printed Ciruit Board (PCB), consequently it is necessary for us to test whether the PCB works normally or not. The schematic of PCB can be reverted through PCB Reverse Engineering technology, which can be used to detect the defects which exist in PCB. Industrial Cone-Beam Computed Tomography (CBCT) provides a new way for PCB Non-Destructive Testing (NDT) with characteristics of fast imaging and high space resolution.However, the reconstructed image based on industrial CBCT contains a lot of artifacts, especially the geometry artifact and metal artifact, affected by beam-hardening, photon-scattering, etc. And because of the specific application of industrial CBCT, the generality of current artifact correction methods is poor. Researching artifact correction methods for PCB reconstructed image is essential for completing the system of artifact correction methods and upgrading image quality.The main research content of this paper is depressing geometry artifact and metal artifact which exist in PCB image, the major achievements cover the following aspects:1. Aiming at the problem that geometry paprameters which were estimated in previous geometry artifact calibration algorithms are too many can't be got fast, an algorithm of geometry calibration using a two-ball template is presented in this paper. This method needs a two-ball template, only four geometry parameters in our defined coordinate system are used to establish a least square model through computing the point of intersection between the main X-ray beam and flat detector plane, which can be solved by Gauss-Newton algorithm. The results of experiment show that the geometry parameters can be estimated fast, moreover, the geometry artifact in PCB reconstruced image is eliminated apparently.2. As for the limitation of traditional metal artifact calibration algorithms, according to the features of metal artifacts in PCB reconstructed image, a new way for metal artifact correcting based on orthogonal scanning is proposed in this paper. This method firstly scans PCB twice and images are reconstructed respectively. Then two-dimensional slices belonging to the same layer of PCB are segmented. At last, image processing, image matching and image fusion techniques are adopted to decrease metal artifacts. The experiments show that much useful information is recovered, and the traditional calibration algorithm is ineffective in comparision.3. The metal artifact calibration method in this paper needs twice image reconstruction which is time-consuming. Parallel computing technique based on Compute Unfied Device Architecture (CUDA) implemented on Graphics Processing Unit (GPU) is used to accelerate the process of image reconstruction, and a new way of optimization strategy is brought forward. The optimization strategy contains maxmizing the occupancy of GPU, optimizing utilization of memory accessing and mathematics instruction. The results of experiment indicate that the program which runs on GPU using these optimization techniques is faster than chained program which runs on CPU by a factor of two hundred and thirty five, and faster than that doesn't use these optimization techniques by a factor of three point three.