The fine-grained image has characteristics of large intra-class variance and small inter-class variance， which makes Fine-Grained Image Categorization （FGIC） much more difficult than traditional image classification tasks. The application scenarios， task difficulties， algorithm development history and related common datasets of FGIC were described， and an overview of related algorithms was mainly presented. Classification methods based on local detection usually use operations of connection， summation and pooling， and the model training was complex and had many limitations in practical applications. Classification methods based on linear features simulated two neural pathways of human vision for recognition and localization respectively， and the classification effect is relatively better. Classification methods based on attention mechanism simulated the mechanism of human observation of external things， scanning the panorama first， and then locking the key attention area and forming the attention focus， and the classification effect was further improved. For the shortcomings of the current research， the next research directions of FGIC were proposed.

In order to accurately and quickly measure the spatial straightness of train axle, a measurement system of train axle spatial straightness was constructed and the algorithms of spatial circle fitting, spatial straight line fitting and straightness measurement were studied. Firstly, the spatial circle fitting algorithm based on spatial plane and spatial sphere tangent was introduced according to the characteristics of the object under test. Then, the RANdom SAmple Consensus (RANSAC) algorithm was used to iterate out the best point set of the model. On the basis of the data obtained from the spatial circle fitting of the train axle section, the data of the train axle section spatial circle center was analyzed. And the wolf colony algorithm was used to fit the spatial straight line, that is, according to the circle center coordinates of the spatial circle of the train axle section at the position of the space section, the spatial straight line of the train axle was fitted. Finally, the wolf colony algorithm was used to measure the spatial straightness of train axle, and the measured data were compared with the data of laser tracker. Experimental results show that the accuracy of measuring the spatial straightness of train axle based on wolf colony algorithm is 0.01 mm, which can meet the requirements of high accuracy, high stability and repeatability in measurement of the spatial straightness of train axle.

It is difficult to realize on-line measurement for the large dimension range of China Railway High-speed (CRH) body, the complexity of testing items and the variety of vehicles. Firstly, a measurement scheme of key dimensions for a large-scale bullet train was proposed, where binocular Charge Coupled Device (CCD) stereo vision was used to set up the measuring sub stations of each key dimension, and the laser tracker and coordinate transformation algorithm were used to complete the global calibration of each CCD camera's measuring sub station. In each measuring sub station, the stereo spatial ball detection technology was used to measure local key dimensions. At the same time, a neural network temperature error compensation model based on wavelet analysis was constructed, and the precision of space distance compensation reached 0.05 mm. The comparison between the proposed method and three-coordinate measuring machine, shows that the proposed method is simple in operation, high in flexibility and high in precision, which can effectively solve the key dimension detection problem of CRH body.

Concerning the problem of a large amount of noise in Low-Dose Computed Tomography (LDCT) reconstructed images, a deep residual Convolutional Neural Network for Stationary Wavelet Transform (SWT-CNN) model was proposed to estimate Normal-Dose Computed Tomography (NDCT) image from LDCT image. In training phase, the high-frequency coefficients of LDCT images after Stationary Wavelet Transform (SWT) three-level decomposition were taken as inputs, the residual coefficients were obtained by subtracting the high-frequency coefficients of NDCT images from high-frequency coefficients of LDCT images were taken as labels, and the mapping relationship between inputs and labels could be learned by deep CNN. In testing phase, the high-frequency coefficients of NDCT image could be predicted from the high-frequency coefficients of LDCT image by using this mapping relationship. Finally, the predicted NDCT image could be reconstructed by Stationary Wavelet Inverse Transform (ISWT). With the size of 512 x 512, 50 pairs of normal-dose chest and abdominal scan sections of the same phantom and reconstructed images with noise added to the projection field were used as data sets, of which 45 pairs constituted a training set and the remaining 5 pairs constituted a test set. The SWT-CNN model was compared with the-state-of-the-art methods, such as Non-Local Means (NLM), K-Singular Value Decomposition (K-SVD) algorithm, Block-Matching and 3D filtering (BM3D), and Image domain CNN (Image-CNN). The experimental results show that, the Peak Signal-to-Noise Ratio (PSNR) and Structural SIMilarity (SSIM) of NDCT image predicted by SWT-CNN model are higher, and its Root Mean Square Error (RMSE) is smaller than that of other algorithms. The proposed model is feasible and effective in improving the quality of low-dose CT images.

Concerning that the Wiener filtering algorithm leads signal distortion with low Signal-to-Noise Ratio (SNR) when dealing with the noise of non-stationary speech signal, a new speech denoising algorithm named SSA-WF was proposed combining with Singular Spectrum Analysis (SSA) and Wiener Filtering (WF). To obtain the speech signal as smooth as possible, SSA was used to denoise the nonlinear and non-stationary speech signal to improve the SNR of the noisy speech. Then the processed signal was put into WF to further eliminate the high frequency noise that still existed in the speech signal. The simulation results from different intensity of background noise show that the proposed algorithm is superior to the traditional methods in SNR and Root-Mean-Square Error (RMSE). The results also demonstrate that the new algorithm can not only remove the background noise efficiently, but also reserve the details of the original signal, it is suitable for the denoising of nonlinear and non-stationary speech signal.

Since the compressive sampling of Block-Based Compressed Sensing (BCS) in spatial domain lacks of considering the global features of an image, image fusion based on conventional BCS sampling suffers from reduced quality and blocking artifacts during reconstruction. Firstly, the input images were sparsely represented by Contourlet Transform (CT), then the Contourlet Transform Block-Based Compressed Sensing (CTBCS) sampling was implemented in the CT domain. Secondly, the compressive samplings were fused by the rule of linear weighting. Finally, the fused image was reconstructed by Iterative Thresholding Projection (ITP) algorithm with consideration of blocking artifacts. The fusion method based on CTBCS was proposed for remote-sensing images, and the implementation algorithm was also presented in detail. In the simulation experiments, BCS and CTBCS were used for compressive sampling, then ITP algorithm was used for image reconstruction. The simulation results show that, compared with BCS, CTBCS sampling which considered the global characteristics has higher convergence speed, less computational complexity and higher reconstructing accuracy, the corresponding Peak Signal-to-Noise Ratio (PSNR) of recovery image is also higher. The real data tests indicate that the compressive fusion based on CTBCS achieves better result than that based on BCS. With very small amount of samples, the CTBCS-based compressive fusion can achieve a comparable result with fusion by the conventional CT method. Therefore, the proposed fusion method effectively implements the compressive fusion for the remote-sensing images with large amounts of data.

Aiming at the problem that the traditional nonlinear robust filtering will be severely degraded when the distribution of measurement noise deviates from the assumed Gaussian distribution, a new robust nonlinear Kalman filter based on M-estimation and detection method was proposed. The proposed robust filtering algorithm set a threshold using Chi-square test to delete mutation outliers, and modified the measurement update using M-estimation. Several conventional nonlinear filtering methods were evaluated under different measurement noises in terms of accuracy and stability. Under non-Gaussian noise and strong interference, the proposed algorithm outperforms the traditional robust algorithm with higher estimation accuracy by 25.5% and lower estimation covariance by 18.3%. The experimental results show that the proposed filtering algorithm can suppress the influence of non-Gaussian noise and strong interference, and increase the estimation accuracy and stability.

As for the large differences between the visual and infrared images in gray value caused by different imaging mechanism, inconsistent contour, the low matching probability of traditional matching methods based on gray or feature, the gray information of visual and infrared images was introduced after researching a variety of Hausdorfff Distance (HD) algorithms. Image matching method based on the neighbor grayscale information Hausdorfff distance was proposed. Based on the calculation of the similarity of edge feature points, the calculation of image normalized grayscale variance was added into this method, which effectively solved the low probability problem caused by different edge of visual/infrared image in Hausdorff distance matching algorithms. The simulation results of visual and infrared images matching show that under various conditions, compared with the conventional Hausdorff distance method, the proposed algorithm effectively improves matching effect under different light conditions and anti-jamming of noise.

The separation method of the graph was proposed for  improving destroy-resistance and throughput at the assumption of known user distribution. Link Weight Factor (LWF) that eclectically thinks over geography complexities, destroyresistance and portfolio is defined. Putting forward the intersection circle algorithm that seeks the feasible seat of nodes and virtual cell algorithm that seeks the service center.The Communication Station design closely combines with the geography information systems, therefore the result is feasible and available.

The paper analyzed some methods of real-time data sampling in Windows 2000/XP platform and proposed an optimal solution using the mechanisms for real-time application in Windows 2000/XP. For disserting the design idea, A vibration signal sampling system was proposed as an example application and experimental data was given which indicates that this solution can fulfill the requirements of real-time application. The practical result has proved to be successful.