Research On The Technology Of Industry X-ray Image Processing

Gas insulated switchgear (GIS) is widely used in high voltage transmission and distribution system which plays a vital role in the power system, once it is failure, there will be a serious loss. With the continuous development of the non-destructive detecting, X-ray detecting, compared with the traditional method of sensor physical and chemical parameters, is an effective external diagnostic method for GIS to ensure it reliable operation without disassembly. The digital X-ray imaging system, includes the high-frequency X-ray machine and flat-panel, detect the GIS by radiographic, and the detecting image is transfer to a computer to give an objective basis. Therefore, some key issues in image intelligent for portable digital X-ray imaging system were deeply research in the dissertation, detailed as follows:(1) High-resolution perception realized in multi-dimensional reconstructed space by using non-relevant observation data of low resolution of X-ray image. It is not only an important scientific problem of compressed sensing optimization, but also a major problem of X-ray high-resolution displaying in the limit of sensor hardware. Aim to study on its sparsity and the equidistant limited conditions of sprase transform (RIP) for the X-ray images by the model of weighted1-norm compressed perception. A discrete wavelet transform transfer the X-ray image to sparse domain. The matrix0, with maximum incoherent observation construction, is built. The original image reconstructed in substrate by orthogonal tracking algorithm, experiments shows the effectively in time and effect.(2) For the error system of the X-ray detector hardware, significant feature in image showed unclearly. The dual-platform of multi-scale and adaptive histogram is proposed to enhance significant feature information. First, the global grayscale restrictions function of limit contrast equalization for the initial sub-block. It decomposed in the three scales factor projection by transform function coefficients, and which adjusts automatically in partial sub graph contrast. This method to adjust the contrast from the overall and local grayscale, on the one hand the limiting function is selected from the overall situation, on the other hand the significant feature details retained on partial.(3) The unknown quantum noise influence on X-ray image, in order to weaken the excluding non-significant feature, the bivariate non-local fuzzy average filtering is put forward which introduce the gradient operator to reduce the correlation between adjacent pixels on the overall sample, then fuzzy operation to set the filter window size and the number of sub-blocks. The adaptive non-local filter adopt in every sub-block to eliminate the image non-significant features the impact effectively.(4) To solve the portable X-ray detector is unable to complete the imaging device at a time of large-scale GIS. It is multiple X-ray image group that captured by the mobile X-ray detector. The potential splicing area identified by the maximum cross-correlation mutual information based on particle swarm optimization. Moreover, the best match feature points in is matched by the LLE-SIFT based on nonlinear manifolds methods. Then the overlapped different is decreased by portion of the gradient-weighted image fusion method to obtain a large viewing image.(5) Research on typical defects for GIS equipment, the moment invariant special detection based on the fast neighborhood consistency matching model background recovery is proposed. The fast neighbor matching the background model improved from the fast matching model (FMM), and the monolayer point mode revised to the point and regional dual repair. It improves the convergence rate of the repair function. In order to improve the accuracy, X-ray image digital subtraction method based on second moment invariant feature, contains the GIS contour, to detect the defect type and location. It also eliminates the GIS equipment security risks.