In view of the problem that most existing network situation prediction methods are unable to mine the deep information in the data and need to manually extract and construct features, a deep feature network situation prediction method named DFS-Seq2Seq (Deep Feature Synthesis-Sequence to Sequence) was proposed. First, the data produced by network streams, logs and system events were cleaned, and the deep feature synthesis algorithm was used to automatically synthesize the deep relation features. Then the synthesized features were extracted by the AutoEncoder (AE). Finally, the data was estimated by using the Seq2Seq (Sequence to Sequence) model constructed by Long Short-Term Memory (LSTM). Through a well-designed experiment, the proposed method was verified on the public dataset Kent2016. Experimental results show that when the depth is 2, compared with four classification models including Support Vector Machine (SVM), Bayes, Random Forest (RF) and LSTM, the proposed method has the recall rate increased by 7.4%, 11.5%, 6.5% and 3.0%, respectively. It is verified that DFS-Seq2Seq can effectively identify dangerous events in network authentication and effectively predict network situation in practice.

Most of the existing portrait segmentation algorithms ignore the hardware limitation of mobile devices and blindly pursue the effect, so that they cannot meet the segmentation speed requirement of mobile terminals. Therefore, a portrait segmentation network which could run efficiently on mobile devices was proposed. Firstly, the network was constructed based on the lightweight U-shaped architecture of encoder-decoder. Secondly, in order to make up for the fact that the Fully Convolutional Network (FCN) was limited by a small sensing domain, so that it was not able to fully capture the long-distance information, an Expectation Maximization Attention Unit (EMAU) was introduced after the encoder and before the decoder. Thirdly, for improving the accuracy of portrait boundary contour, a multi-layer boundary auxiliary loss was added at the training stage. Finally, the model was quantized and compressed. The proposed network was compared with other networks such as PortraitFCN+, ENet and BiSeNet on Veer dataset. Experimental results show that, the proposed network can improve the image reasoning speed and segmentation effect, as well as process the RGB images with the resolution of 224×224 at the accuracy of 95.57%.

Joint trajectory tracking control and flexible vibration suppression techniques for a Free-Floating Flexible Space Robot (FFFSR) were discussed under parameter uncertainty and limited torque. A composite controller containing a slow control subsystem for joint trajectory tracking and a fast control subsystem for flexible vibration description were proposed using singular perturbation method. A model-free Fuzzy Radial Basis Function Neural Network (FRBFNN) adaptive tracking control strategy was applied in the slow subsystem. FRBFNN was adopted to support the estimation of velocity signals performed by the observer, the approximation of the unknown nonlinear functions of the observer as well as the controller. The fast subsystem adopted an Extended State Observer (ESO) to estimate coordinate derivatives of flexible modal and uncertain disturbance, which could hardly be measured, and used Linear Quadratic Regulator (LQR) method to suppress the flexible vibration. Numerical simulation results show that the composite controller can achieve stable joint trajectory tracking in 2.5 s, and the flexible vibration amplitude is restricted in ±1×10 ^-3 m, when the control torque is limited within ±20 N·m and ±10 N·m.

In order to improve the performance of Differential Evolution (DE) algorithm and extend its adaptability for solving continuous optimization problems, an enhanced DE algorithm was proposed by using efficient local search under the Memetic framework. Specifically, based on the Davidon-Fletcher-Powell (DFP) method, an improved local search method named NDFP was put forward, which could speed up finding locally optimal solutions based on excellent individuals explored by the DE algorithm. Furthermore, a strategy on when and how to run the NDFP local search was also given, so as to strike a good balance between global search (i.e., DE) and local search (i.e., NDFP). The proposed strategy was also enhanced the adaptability of NDFP local search in the range of DE algorithm. To verify the efficiency of the proposed algorithm, extensive simulation experiments were conducted on up to 53 test functions from CEC2005 and CEC2013 Benchmarks. The experimental results show that, compared with DE/current-to-best/1, SaDE and EPSDE algorithms, the proposed algorithm can achieve better performance in terms of both precision and stability.

Subject to the imaging principle, manufacturing technology and other factors, the spatial resolution of spaceborne hyperspectral remote sensing imagery is relatively low. Therefore, the thesis proposed the image fusion of hyperspectral imagery and high spatial resolution imagery, and designed the best fusion algorithm to enhance spatial resolution of hyperspectral remote sensing imagery. According to the characteristics of Earth Observing-1 (EO-1) Hyperion hyperspectral imagery and Advanced Land Imager (ALI) panchromatic imagery, 4 kinds of fusion algorithms were selected to carry out a comparative study of the image fusion effect for the city and mountain regions from 9 kinds of remote sensing image fusion algorithms, namely Gram-Schmidt spectral sharpening fusion method, transform fusion method of Smoothing Filter-based Intensity Modulation (SFIM), Weighted Average Method (WAM) fusion method and Wavelet Transformation (WT) fusion method. And it carried out the comprehensive evaluation and analysis of the image fusion effect from 3 aspects of qualitative, quantitative and classification precision, which aims to determine the best fusion method for EO-1 hyperspectral imagery and panchromatic imagery. The experimental results show that: 1) from the image fusion effect, Gram-Schmidt spectral sharpening fusion method is the best in 4 kinds of fusion methods used; 2) from the image classification effect, the classification results based on the fusion image is better than the classification results based on the source image. The theoretical analysis and experimental results show that Gram-Schmidt spectral sharpening fusion method is an ideal fusion algorithm for hyperspectral imagery and high spatial resolution imagery, and it can provide powerful support to improve the clarity of hyperspectral remote sensing imagery, the reliability and the accuracy of the image object recognition and classification.

Concerning the problem of low robustness of general watermarking algorithms in resisting JPEG compression and geometric transform attacks, a zero-watermarking algorithm based on Cellular Automata (CA) and Singular Value Decomposition (SVD) was proposed. Firstly, an image was transformed by 2-dimensional cellular automata transform and the low-frequency subband approximation image were isolated, then the CA parameters was saved as key. After that, the approximation image was sub-blocked, and the blocks were decomposed by SVD, then the zero-watermark was constructed by CA rule in SVD matrix. In image authentication, the image could be certificated by comparing the similarity of two watermarks with the threshold value. The experimental result shows that this algorithm has good invisibility and perfect robustness in resisting JPEG compression and geometric transform attacks.

The structure of 2-adic Finite State Machine (2-adic FSM) was studied. To build the machine, multiple-input Galois Feedback with Carry Shift Register (FCSR) vanes were used as building blocks instead of the one-input vanes which were used in the old method. This leads to a new implementation method of 2-adic FSM. With this method, a general 2-adic FSM was transformed into an equivalent 2-adic FSM with integer matrices. Moreover, if there exist some entries whose denominators are not coprime in the same row of the input or the transition matrix, the length of the transformed 2-adic FSM is shorter than the one in the old method, thus reducing the number of registers.

By discussing the shortages of access control policy, and analyzing the complementarity and completeness between access control and state control, the necessity of state control was proposed. A formal description about state control policy was defined, and the policy's description rules based on XML were regulated. At the same time, according to different control goal and control object, some application patterns for state control policy were provided. In addition, the complexity of state control policy was discussed, and some solutions were provided.

Hyperspectral remote sensing imagery contains abundant spectral information because of its numerous bands, but it also causes the curse of dimensionality. How to resolve this conflict and improve the classification accuracy of hyperspectral remote sensing imagery is the major concern. Therefore, the thesis proposed ICA-DTC model that combined Independent Component Analysis (ICA) with Decision Tree Classifier (DTC) to research the hyperspectral imagery classification based on EO-1 Hyperion. First, ICA was applied to carry on the feature extraction on hyperspectral remote sensing imagery. Based on this, the characteristic components and other geography auxiliary elements were selected as test variables, the appropriate threshold was selected to set discriminating rule, and the DTC model was established to classify hyperspectral remote sensing imagery. Then the results obtained by this method were compared with that obtained by traditional maximum likelihood classification. The experimental results show that ICA can extract nonlinear characteristics from surface features well and ICA-DCT model can effectively improve the classification accuracy of surface features under complex terrain. In terms of the total classification accuracy, the former is up to 89.34%, 18.8% higher than the latter. In terms of the classification accuracy of a single surface feature, the former is only slightly lower than the latter on water, while 11 other surface features are higher than the latter.

In wireless mobile ad hoc environment, the characteristics of multi-hopping and dynamic topology make route information stale quickly, and the sender cannot activate route request in time when route fails. All these make the throughput of TCP over DSR degrades when mobile nodes move fast. RP-DSR was proposed based on DSR, in which the receiver participated in route discovery. The simulation results show that RP-DSR protocol can get more satisfying performance than DSR in the status of fast and dynamic topology.