Research On Image Quality Optimization In Geometric Dimensions Measurement By Industrial Computed Tomography

Dimensional metrology is a key step in reverse engineering. Dimensional metrology based on Computed Tomography(CT) is non-destructive and accurate, attracting more and more interests. However, suffered from the multi-energy characteristic of X-ray source, limited dynamic range of the detector and non-point feature of focal spot, CT image quality will be polluted, thereby affecting the measurement accuracy. As a result, research on CT image quality optimization is quite significant in improving the accuracy of the dimensional metrology by industrial CT.In order to improve the accuracy of CT dimensional metrology, this paper focuses on three main factors affecting the image quality, including the beam hardening artifact, dynamic range of detector and point spread effect and the main achievements are as follows:1, Aming at weaken the beam hardening artifacts, this paper proposes a beam hardening correction method based on the exponential model. The parameters of beam hardening exponent model are treated as unknown variables. A gray entropy based cost function with constraints is built for the optimization model by introducing both the CT image uniformity feature and the prior information to calibration parameters as constraints. The beam hardening parameters are solved by the simplex method and applied to correcting the original projection data. Both numerical and experimental results show that the proposed method can effectively suppress the beam hardening artifact and improve CT image quality, the dimensional accuracy is significantly improved using the image after optimization.2, Aming at the low image quality caused by limited dynamic range of the detector, a dynamic time wrapping(DTW) based multi-current projection fusion method is proposed. This method firstly uses gray value-effective edge gradient sequences to describe the difference at the edge, and then the sequences are matched by DTW method, auto determining the best threshold for fusion. The sub region of multi-energy projection is extracted by the threshold, realizing the expanding of detector dynamic range. Both numerical and experimental results show that this method can overcome the problem of image missing due to the over-exposure of the detector, improving the signal to noise ratio of CT image at the meantime, based on which the accuracy of dimensional metrology is significantly optimized.3, Aming at the problem of image degradation due to the point spread effect, an image restoration method based on point spread function(PSF) measurement and alternating direction total variation(TV) is proposed. This method firstly builds the PSF model on the assumption that the PSF of CT is a Gaussian function and separable, and the golden section method is utilized to solve the parameters. Then the optimized model is built using TV in blurred image as the target function, and the alternating direction method is used to restore the image. The edge of the restored image is well maintained. Both numerical and experimental results show that the proposed method can reduce the degradation, and higher measurement accuracy is obtained based on the CT image after restoration.