To address the problem of low recognition accuracy caused by vanishing gradient and overfitting in the cattle eye image feature extraction process, an improved DenseNet based cattle eye image feature extraction method was proposed. Firstly, the Scaled exponential Linear Unit (SeLU) activation function was used to prevent the vanishing gradient of the network. Secondly, the feature blocks of cattle eye images were randomly discarded by DropBlock, so as to prevent overfitting and strengthen the generalization ability of the network. Finally, the improved dense layers were superimposed to form an improved Dense convolutional Network (DenseNet). Feature information extraction recognition experiments were conducted on the self-built cattle eyes image dataset. Experimental results show that the recognition accuracy, precision and recall of the improved DenseNet are 97.47%, 98.11% and 97.90% respectively, and compared to the network without improvement, the above recognition accuracy rate, precision rate, recall rate are improved by 2.52 percentage points, 3.32 percentage points, 2.94 percentage points respectively. It can be seen that the improved network has higher precision and robustness.

With the rapid development of Internet of Things (IoT), more and more IoT node devices are deployed, but the accompanying security problem cannot be ignored. Node devices at the network layer of IoT mainly communicate through wireless sensor networks. Compared with the Internet, they are more open and more vulnerable to network attacks such as denial of service. Aiming at the network layer security problem faced by wireless sensor networks, a network intrusion detection system based on sequence model was proposed to detect and alarm the network layer intrusion, which achieved higher recognition rate and lower false positive rate. Besides, aiming at the security problem of the node host device faced by wireless sensor network node devices, with the consideration of the node overhead, a host intrusion detection system based on simple sequence model was proposed. The experimental results show that, the two intrusion detection systems for the network layer and the host layer of wireless sensor network both have the accuracy more than 99%, and the false detection rate about 1%, which meet the industrial requirements. These two proposed systems can comprehensively and effectively protect the wireless sensor network security.

Recently, deep learning has been a hot research topic in the field of image super-resolution due to the excellent performance of deep convolutional neural networks. Many large-scale models with very deep structures have been proposed. However, in practical applications, the hardware of ordinary personal computers or smart terminals are obviously not suitable for large-scale deep neural network models. A light-weight Network with Automatic Residual Scaling (ARSN) for single image super-resolution was proposed, which has fewer layers and parameters compared with many other deep learning based methods. In addition, the specified residual blocks and skip connections in this network were utilized for residual scaling, global and local residual learning. The results on test datasets show that this model achieves state-of-the-art performance on both reconstruction quality and running speed. The proposed network achieves good results in terms of performance, speed and hardware consumption， and has high practical value.

Sparse coding similarity measure has good robustness to gray-scale offset field in non-rigid medical image registration, but it is only suitable for single-modal medical image registration. A non-rigid multi-modal medical image registration method based on multi-channel sparse coding was proposed to solve this problem. In this method, the multi-modal registration was regarded as a multi-channel registration, with each modal running in a separate channel. At first, the two registered images were synthesized and regularized separately, and then they were divided into channels and image blocks. The K-means-based Singular Value Decomposition ( K-SVD) algorithm was used to train the image blocks in each channel to get the analytical dictionary and sparse coefficients, and each channel was weightedy summated. The multilayer P-spline free transform model was used to simulate the non-rigid geometric deformation, and the gradient descent method was used to optimize the objective function. The experimental results show that compared with multi-modal similarity measure such as local mutual information, Multi-Channel Local Variance and Residual Complexity (MCLVRC), Multi-Channel Sparse-Induced Similarity Measure (MCSISM) and Multi-Channel Rank Induced Similarity Measure (MCRISM), the root mean square error of the proposed method is decreased by 30.86%, 22.24%, 26.84% and 16.49% respectively. The proposed method can not only effectively overcome the influence of gray-scale offset field on registration in multi-modal medical image registration, but also improve the accuracy and robustness of registration.

Focusing on the issue that the video watermarking information is vulnerable to geometric attacks and the balance between robustness and adaptability of the watermarking algorithm, a dual watermarking scheme based on human visual characteristics and Scale Invariant Feature Transform (SIFT) was proposed. Firstly, the human visual threshold in the video sequence was taken as the maximum embedding strength of the watermark; secondly, the video frame was processed by Discrete Wavelet Transform (DWT). An adaptive watermarking algorithm based on video motion information was proposed for medium-high frequency subband coefficients; based on statistical properties of wavelet coefficients, an anti-geometric attack video watermarking scheme was proposed for low-frequency ones. Finally, SIFT was acted as the trigger to judge whether the video frame was subjected to geometric attacks. The video frames were corrected by using the SIFT scale and orientation invariance when it was under geometric attack, and the watermark signal of the video frame was extracted after correction. For video frame under non-geometric attack, the medium-high frequency extraction scheme was used directly. In comparison with the real-time robust video watermarking algorithm, called VW-HDWT (Video Watermarking based on Histogram in DWT domain) algorithm, the Peak Signal-to-Noise Ratio (PSNR) value was improved by 7.5%. Compared with the watermarking algorithm based on feature area, the capacity of watermark embedding could be increased by about 10 times. The experimental results show that the proposed scheme is robust to common geometric attacks in the condition of fine watermarking transparency.

Many existing scheduling approaches in cloud data centers try to consolidate Virtual Machines (VMs) by VM live migration technique to minimize the number of Physical Machines (PMs) and hence minimize the energy consumption, however, it introduces high migration overhead; furthermore, the cost factor that leads to high payment cost for cloud users is usually not taken into account. Aiming at energy reduction for cloud providers and payment saving for cloud users, as well as guaranteeing the deadline of user tasks, a heuristic task scheduling algorithm called Energy and Deadline-Aware with Non-Migration Scheduling (EDA-NMS) was proposed. The execution of the tasks that have loose deadlines was postponed to avoid waking up new PMs and migration overhead, thus reducing the energy consumption. The results of extensive experiments show that compared with Proactive and Reactive Scheduling (PRS) algorithm, by selecting a smart VM combination scheme, EDA-NMS can reduce the static energy consumption and ensure the lowest payment with meeting the deadline requirement for key user tasks.

The existing Low-rank representation methods for face recognition fuse of local and global feature information of facial images inadequately. In order to solve the problem, a new face recognition method called Characterized Dictionary-based Low-Rank Representation (LRR-CD) was proposed. Firstly, every face image was represented as a set of characterized patches, then the low-rank reconstruction characteristic coefficients based on training samples as well as the corresponding intra-class characteristic variance were minimized. To obtain the efficient and high discriminative reconstruction coefficient matrix of face image patches, a new mathematical formula was presented. This formula could be used to completely preserve both global and local features of original hyper-dimensional face images, especially the local intra-class variance features, by the way of minimizing the low-rank constraint problem of corresponding patches in training samples and correlated intra-class variance dictionary. What's more, owing to the adequate mining of patch features, the proposed method obtained good robustness to the general noise such as facial occlusion and luminance variance. Several experiments were carried out on the face databases such as AR, CMU-PIE and Extended Yale B. The experimental results fully illustrate that the LRR-CD outperforms the compared algorithms of Sparse Representation Classification (SRC), Collaborative Representation Classification (CRC), LRR with Normalized CUT (LRR-NCUT) and LRR with Recursive Least Square (LRR-RLS), with the higher recognition rate of 2.58-17.24 percentage points. The proposed method can be effectively used for the global and local information fusion of facial features and obtains a good recognition rate.

The traditional information propagation model is more suitable for homogeneous network, and cannot be effectively applied to the non-homogeneous scale-free Social Network (SN). To solve this problem, an information propagation model based on local information was proposed. Topological characteristic difference between users and different effect on information propagation of user influence were considered in the model, and the probability of infection was calculated according to the neighbor nodes' infection and authority. Thus it could simulate the information propagation on real social network. By taking simulation experiments on Sina microblog networks, it shows that the proposed model can reflect the propagation scope and rapidity better than the traditional Susceptible-Infective-Recovered (SIR) model. By adjusting the parameters of the proposed model, it can verify the impact of control measures to the propagation results.

Currently, the query of transmission lines monitoring system in smart grid is mostly aiming at the global query of Wireless Sensor Network (WSN), which cannot satisfy the flexible and efficient query requirements based on any area. The layout and query characteristics of network were analyzed in detail, and a query algorithm based on mesh structure in large-scale smart grid named MSQuery was proposed. The algorithm aggregated the data of query nodes within different grids to one or more logical query trees, and an optimized path of collecting query result was built by the merging strategy of the logical query tree. Experiments were conducted among MSQuery, RSA which used routing structure for querying and SkySensor which used cluster structure for querying. The simulation results show that MSQuery can quickly return the query results in query window, reduce the communication cost, and save the energy of sensor nodes.

In this paper, an improved manifold distance based dissimilarity measure was designed to identify clusters in complex distribution and unknown reality data sets. This dissimilarity measure can mine the space distribution information of the data sets with no class labels by utilizing the global consistency between all data points. A K-medoids algorithm based on the improved manifold distance was proposed using the dissimilarity measure. The experimental results on eight artificial data sets with different structure and the USPS handwritten digit data sets indicate that the new algorithm outperforms or performs similarly to the other two K-medoids algorithms based on the existing manifold distance and Euclid distance and has the ability to identify clusters with simple or complex, convex or non-convex distribution.

Bayesian network is one of the most important theoretical models for the representation and reasoning of uncertainty. At present, its structure learning has become a focus of study. In this paper, a Bayesian network structure learning algorithm was developed, which was based on topological order and quantum genetic algorithm. With the richness of the quantum information and the parallelism of quantum computation, this paper designed generator strategy of topological order based on a quantum chromosome to improve not only the efficiency of search, but also the quality of Bayesian network structure. And then by using self-adaptive quantum mutation strategy with upper-lower limit, the diversity of the population was increased, so that the search performance of the new algorithm was improved. Compared to some existing algorithms, the experimental results show that the new algorithm not only searches higher quality Bayesian structure, but also has a quicker convergence rate.

The existing watermarking algorithm based on image moments has the disadvantages of small capacity, large complexity and its robustness should be improved in further study. A new zero-watermarking against geometric attacks was proposed. Using the image normalization and the features of Tchebichef moments, the rotation normalized Tchebichef moments of original image was calculated in the unit circle, and the upper-left corner of Tchebichef moments was scanned into numerical matrix. Afterwards, binary secret key was generated according to numerical matrix and watermark image, and saved to zero-watermarking information database. In detection, the same process was executed to draw out numerical matrix from the unauthenticated image, and watermark image was extracted by using the secret key and numerical matrix. The experimental results show that this algorithm is robust against rotation attacks of random angles, scaling attacks and common signal processing operations, even combined attacks.

To prevent the privacy disclosure caused by linking attack and reduce information loss resulting from anonymous protection, a (λα,k) multi-level anonymity model was proposed. According to the requirement of privacy preservation, sensitive attribute values could be divided into three levels: high, medium, and low. The risk of privacy disclosure was flexibly controlled by privacy protection degree parameter λ. On the basis of this, clustering-based approach for multi-level anonymization was proposed. The approach used a new hierarchical clustering algorithm and adopted more flexible strategies of data generalization for numerical attributes and classified attributes in a quasi-identifier. The experimental results show that the approach can meet the requirement of multi-level anonymous protection of sensitive attribute, and effectively reduce information loss.

In view of that the traditional decision tree algorithm that cannot solve the mass data mining and the multi-value bias problem of ID3 algorithm, the paper designed and realized a parallel decision tree classification algorithm based on the MapReduce framework. This algorithm adopted attribute similarity as the choice standard to avoid the multi-value bias problem of ID3 algorithm, and used the MapReduce model to solve the mass data mining problems. According to the experiments on the Hadoop cluster set up by ordinary PCs, the decision tree algorithm based on MapReduce can deal with massive data classification. What's more, the algorithm has good expansibility while ensuring the classification accuracy and can get close to linear speedup rate.

In order to deal with the frequently encountered non-stationary random signals in signal processing, they can be divided into sub-stationary random signals, and autocorrelation function can be used to reflect the essential characteristics of sub-stationary signals. The computation of piecewise stationary stochastic process autocorrelation function was discussed. In order to reduce the amount of calculation and errors of the existing function models, a new model to approximate autocorrelation function of piecewise stationary stochastic process was proposed in this paper. The computer simulation shows that the model can effectively approximate autocorrelation function. The computing speed is faster, and the errors are much fewer and smoother. Applying the model to the restoration of blurred digital images, a very good restoration effect can be got.

Concerning the insufficiency of integrated level of virtual apparatus experiment teaching system between colleges, taking advantage of Service-Oriented Architecture (SOA) and Web services technology, and Web-based virtual apparatus technology, the interscholastic virtual apparatus experiment teaching sharing platform was designed and implemented in the.NET architecture, which found a solution to achieve the target for resource share of virtual apparatus experiment teaching system.

An anonymous authentication protocol was proposed to improve the anonymity and to control anonymity abuse in wireless communications. The protocol implemented the anonymity of Home Location Register (HLR) using Visitor Location Registers (VLRs) which had been employed to forward the authentication message by Mobile Station (MS). Through checking the authentication token, this protocol could control anonymity abuse. The probability statistics were adopted to analyze the security and anonymity of this protocol. The analysis results show that this protocol not only improves the anonymity of HLR but also defends anonymity abuse effectively.