In order to solve the problem of low accuracy of some distortion types judgment by image single distortion type judgment algorithm, an image single distortion type judgment method based on two-channel Convolutional Neural Network (CNN) was proposed. Firstly, the fixed size image block was obtained by cropping the image, and the high-frequency information map was obtained by Haar wavelet transform of the image block. Then, the image block and the corresponding high-frequency information map were respectively input into the convolutional layers of different channels to extract the deep feature map, and the deep features were fused and input into the fully connected layer. Finally, the values of the last layer of the fully connected layer were input into the Softmax function classifier to obtain the probability distribution of the single distortion type of the image. Experimental results on LIVE database show that, the proposed method has the image single distortion type judgement accuracy up to 95.21%, and compared with five other image single distortion type judgment methods for comparison, the proposed method has the accuracies for judging JPEG2000 and fast fading distortions improved by at least 6.69 percentage points and 2.46 percentage points respectively. The proposed method can accurately identify the single distortion type in the image.

Aiming at the problem of inaccurate matching of weak texture and pure color region in stereo matching and long time consumption of image segmentation algorithms, a stereo matching algorithm fused with image segmentation was proposed. Firstly, the initial image was filtered by Gaussian and smoothed by Sobel to obtain the edge feature map of the image. Secondly, the red, green and blue channel values of the original image were dichotomized by using the Otsu method and then refused to obtain the segmentation template map. Finally, the obtained grayscale map, edge feature map and segmentation template map were applied in the parallax calculation and parallax optimization process in order to calculate the parallax map. The proposed algorithm has the accuracy improved by 14.23 percentage points on average with the time cost per 10 000 pixels increased by 7.16 ms in comparison with Sum of Absolute Differences (SAD) algorithm. The experimental results show that the proposed algorithm can obtain smoother matching results in pure color and weak texture regions and parallax discontinuity regions, and it can automatically calculate the threshold and segment the image faster.

Deep convolutional neural networks are generally large in scale and complex in computation, which limits their application in high real-time and resource-constrained environments. Therefore, it is necessary to optimize the compression and acceleration of the existing structures of convolutional neural networks. In order to solve this problem, a hybrid compression method combining pruning and stream merging was proposed. In the method， the model was decompressed through different angles, further reducing the memory consumption and time consumption caused by parameter redundancy and structural redundancy. Firstly, the redundant parameters in each layer were cut off from the inside of the model. Then the non-essential layers were merged with the important layers from the structure of the model. Finally, the accuracy of the model was restored by retraining. The experimental results on the MNIST dataset show that the proposed hybrid compression method compresses LeNet-5 to 1/20 and improves its running speed by 8 times without reducing the accuracy of the model.

Focused on the issue that traditional machine learning algorithms still need to manually represent features, a protein subcellular localization algorithm based on the deep network of Stacked Denoising AutoEncoder (SDAE) was proposed. Firstly, the improved Pseudo-Amino Acid Composition (PseAAC), Pseudo Position Specific Scoring Matrix (PsePSSM) and Conjoint Traid (CT) were used to extract the features of the protein sequence respectively, and the feature vectors obtained by these three methods were fused to obtain a new feature expression model of protein sequence. Secondly, the fused feature vector was input into the SDAE deep network to automatically learn more effective feature representation. Thirdly, the Softmax regression classifier was adopted to make the classification and prediction of subcells, and leave-one-out cross validation was performed on Viral proteins and Plant proteins datasets. Finally, the results of the proposed algorithm were compared with those of the existing algorithms such as mGOASVM (multi-label protein subcellular localization based on Gene Ontology and Support Vector Machine) and HybridGO-Loc (mining Hybrid features on Gene Ontology for predicting subcellular Localization of multi-location proteins). Experimental results show that the new algorithm achieves 98.24% accuracy on Viral proteins dataset, which is 9.35 Percentage Points higher than that of mGOASVM algorithm. And the new algorithm achieves 97.63% accuracy on Plant proteins dataset, which is 10.21 percentage points and 4.07 percentage points higher than those of mGOASVM algorithm and HybridGO-Loc algorithm respectively. To sum up, it can be shown that the proposed new algorithm can effectively improve the accuracy of the prediction of protein subcellular localization.

Aiming at the problems of limited energy of nodes, short life cycle and low throughput of Wireless Sensor Network (WSN), a WSN Clustering Routing algorithm based on Genetic Algorithm (GA) and Fuzzy C-Means (FCM) clustering (GAFCMCR) was proposed, which adopted the method of centralized clustering and distributed cluster head election. Network clustering was performed by the base station using a FCM clustering algorithm optimized by GA during network initialization. The cluster head of the first round was the node closest to the center of the cluster. From the second round, the election of the cluster head was carried out by the cluster head of the previous round. The residual energy of candidate node, the distance from the node to the base station, and the mean distance from the node to other nodes in the cluster were considered in the election process, and the weights of these three factors were real-time adjusted according to network status. In the data transfer phase, the polling mechanism was introduced into intra-cluster communication. The simulation results show that, compared with the LEACH (Low Energy Adaptive Clustering Hierarchy) algorithm and the K-means-based Uniform Clustering Routing (KUCR) algorithm, the life cycle of the network in GAFCMCR is prolonged by 105% and 20% respectively. GAFCMCR has good clustering effect, good energy balance and higher throughput.

Heterogeneous Multi-core Processors (HMPs) platform has become the mainstream solution for modern embedded system design, and online mapping or scheduling plays a vital role in making full use of the advantages of high performance and low power consumption. Aiming at the dynamic mapping problem of application tasks in HMPs, a mapping and scheduling approach based on machine learning prediction model was proposed. On the one hand, a machine learning model was constructed to predict and evaluate the performance of different mapping strategies rapidly and efficiently, so as to provide support for online scheduling. On the other hand, the machine learning model was integrated with genetic algorithm to find out the optimal resource allocation strategy efficiently. Finally, an Motion-Join Photographic Experts Group (M-JPEG) decoder was used to verify the effectiveness of the proposed approach. The experimental results show that, compared with the Round Robin Scheduler (RRS) and sampling scheduling approaches, the proposed online mapping/scheduling approach has reduced the average execution time by about 19% and 28% respectively.

Traditional vision-based or sensor-based falling detection systems possess certain inherent shortcomings such as hardware dependence and coverage limitation, hence Fallsense, a passive falling detection method based on wireless Channel State Information (CSI) was proposed. The method was based on low-cost, pervasive and commercial WiFi devices. Firstly, the wireless CSI data was collected and preprocessed. Then a model of motion-signal analysis was built, where a lightweight dynamic template matching algorithm was designed to detect relevant fragments of real falling events from the time-series channel data in real time. Experiments in a large number of actual environments show that Fallsense can achieve high accuracy and low false positive rate, with an accuracy of 95% and a false positive rate of 2.44%. Compared with the classic WiFall system, Fallsense reduces the time complexity from O( mN log N) to O( N) ( N is the sample number, m is the feature number), and increases the accuracy by 2.69%, decreases the false positive rate by 4.66%. The experimental results confirm that this passive falling detection method is fast and efficient.

In a new platform for big data stream processing called Heron, the round-robin scheduling algorithm is usually used for task scheduling by default, which does not consider the topology runtime state and the impact of different communication modes among task instances on Heron's performance. To solve this problem, a task scheduling strategy based on Data Stream Classification in Heron (DSC-Heron) was proposed, including data stream classification algorithm, data stream cluster allocation algorithm and data stream classification scheduling algorithm. Firstly, the instance allocation model of Heron was established to clarify the difference in communication overhead among different communication modes of the task instances. Secondly, the data stream was classified according to the real-time data stream size between task instances based on the data stream classification model of Heron. Finally, the packing plan of Heron was constructed by using the interrelated high-frequency data streams as the basic scheduling unit to complete the scheduling to minimize the communication cost by transforming inter-node data streams into intra-node ones as many as possible. After running SentenceWordCount, WordCount and FileWordCount topologies in a Heron cluster environment with 9 nodes, the results show that compared with the Heron default scheduling strategy, DSC-Heron has 8.35%, 7.07% and 6.83% improvements in system complete latency, inter-node communication overhead and system throughput respectively; in the load balancing aspect, the standard deviations of CPU usage and memory usage of the working nodes are decreased by 41.44% and 41.23% respectively. All experimental results show that DSC-Heron can effectively improve the performance of the topologies, and has the most significant optimization effect on FileWordCount topology which is close to the real application scenario.

Aiming at the problem of noise residual in Non-negative Matrix Factorization (NMF) speech enhancement algorithm in low Signal-to-Noise Ratio (SNR) unsteady environment, a Perceptual Masking-based reconstructed NMF (PM-RNMF) single-channel speech enhancement algorithm was proposed. Firstly, psychoacoustic masking features were applied to NMF speech enhancement algorithms. Secondly, different masking thresholds were used for different frequencies to establish an adaptive perceptual masking gain function, and the residual noise energy and speech distortion energy were constrained by the thresholds. Finally, Speech Presence Probability (SPP) was combined to realize perceptual gain correction, the NMF algorithm was reconstructed and a new objective function was established. The simulation results show that under three kinds of unsteady noise environments with different SNR, the average Perceptual Evaluation of Speech Quality (PESQ) of PM-RNMF algorithm is improved by 0.767, 0.474 and 0.162 respectively and the average Signal-to-Distortion Ratio (SDR) is increased by 2.785, 1.197 and 0.948 respectively compared with NMF, RNMF (Reconstructive NMF) and PM-DNN (Perceptual Masking-Deep Neural Network) algorithms. Experimental results show that PM-RNMF has better noise reduction effect in both low frequency and high frequency.

Deep neural network has powerful feature self-learning ability, which can obtain the granularity features of different levels by multi-layer stepwise feature extraction. However, when the target subject of an image has strong correlation with the background color, the feature extraction will be "lazy", the extracted features are difficult to be discriminated with low abstraction level. To solve this problem, the intrinsic law of feature extraction of deep neural network was studied by experiments. It was found that there was correlation between feature extraction preference and background color of the image. Eliminating this correlation was able to help deep neural network ignore background interference and extract the features of the target subject directly. Therefore, a data enhancement algorithm was proposed and experiments were carried out on the self-built dataset. The experimental results show that the proposed algorithm can reduce the interference of background color on the extraction of target features, reduce over-fitting and improve classification effect.

In the mobile robot path planning problem in static multi-obstacle environment, Particle Swarm Optimization (PSO) algorithm has the disadvantages of easy premature convergence and poor local optimization ability, resulting in low accuracy of robot path planning. To solve the problem, a Multi-Objective Grasshopper Optimization Algorithm (MOGOA) was proposed. The path length, smoothness and security were taken as path optimization targets according to the mobile robot path planning requirements, and the corresponding mathematical model of multi-objective optimization problem was established. In the process of population search, the curve adaptive strategy was introduced to speed up the convergence of the algorithm, and the Pareto optimal criterion was used to solve the coexistence problem of the above three targets. Experimental results show that the proposed algorithm finds shorter paths and shows better convergence while solving the above problems. Compared with the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm, the proposed algorithm has the path length reduced by about 2.01 percentage, and the number of iterations reduced by about 19.34 percentage.

To solve the problem that the influencing factors of crowdsourced testing are complex and diverse, and the test quality is difficult to assess, a method for analyzing the quality influencing factors based on Spearman correlation coefficient was proposed. Firstly, the potential quality influencing factors were obtained through the analysis of test platforms, tasks, and testers. Secondly, Spearman correlation coefficient was used to calculate the correlation degrees between potential factors and test quality and to screen out key factors. Finally, the multiple stepwise regression was used to establish a linear evaluation relationship between key factors and test quality. The experimental results show that compared with the traditional expert artificial evaluation method, the proposed method can maintain smaller fluctuations in evaluation error when facing a large number of test tasks. Therefore, the method can accurately screen out the key influencing factors of mobile application crowdsourced test quality.

In order to solve the image degradation in the non-uniform dust environment with multiple scattering lights, a slices reconstruction method for single image dedusting was proposed. Firstly, the slices along the depth orientation were produced based on McCartney model in dust environment. Secondly, the joint dust detection method was used to detect dust patches in the slices where non-dust areas were reserved but the dust zones were marked as the candidate detected areas of the next slice image. Then, an image was reconstructed by combining these non-dust areas of each slice and the dust zone of the last slice. Finally, a restored image was obtained by a fast guided filter which was applied to the reconstructed area. The experimental results prove that the proposed restoration method can effectively and quickly get rid of dust in the image, and lay the foundation of object detection and recognition work based on computer vision in dust environment.

When the length of the test speech is sufficient, the information and discrimination of single feature is sufficient to complete the speaker recognition task. However, when the length of the test speech was very short, the performance of speaker recognition is decreased significantly due to the small data size and insufficient discrimination. Aiming at the problem of insufficient speaker information under the short speech condition, a short utterance speaker recognition algorithm based on multi-featured i-vector was proposed. Firstly, different acoustic feature vectors were extracted and combined into a high-dimensional feature vector. Then Principal Component Analysis (PCA) was used to remove the correlation of the feature vectors, so that the features were orthogonalized. Finally, the most discriminating features were picked out by Linear Discriminant Analysis (LDA), which led to reduce the spatial dimension. Therefore, this multi-featured system can achieve a better speaker recognition performance. With the TIMIT corpus under the same short speech (2 s) condition, the experimental results showed that the Equal Error Rate (EER) of the multi-featured system decreased respectively by 72.16%, 69.47% and 73.62% compared with the single-featured systems including Mel-Frequency Cepstrum Coefficient (MFCC), Linear Prediction Cepstrum Coefficient (LPCC) and Perceptual Log Area Ratio (PLAR) based on i-vector. For the different lengths of the short speech, the proposed algorithm provided rough 50% improvement on EER and Detection Cost Function (DCF) compared with the single-featured system based on i-vector. Experimental results fully indicate that the multi-featured system can make full use of the speaker's characteristic information in the short utterance speaker recognition, and improves the speaker recognition performance.

To solve the problem of uncertain Quality of Service (QoS)-aware cloud service composition optimization, an Uncertain-Long Time Series (ULST) model and Tournament strategy based Genetic Algorithm (T-GA) was proposed. Firstly, based on different access rules of users to services in different periods, the long-term change of QoS was modeled as an uncertain-long time series, which can accurately describe the users' actual QoS access record to service over a period of time. Secondly, an improved genetic algorithm based on uncertain QoS model was proposed, which used tournament strategy instead of basic roulette wheel selection strategy. Finally, a lot of experiments were carried out on real data. The uncertain-long time series model can effectively solve the problem of uncertain QoS-aware cloud service composition; the proposed T-GA is superior to the Genetic Algorithm based on Elite selection strategy (E-GA) in optimization results and stability, and the execution speed is improved by almost one time, which is a feasible, high efficient and stable algorithm.

Clustering algorithm is used to preprocess personal privacy information in order to achieve differential privacy protection, which can reduce the reconstruction error caused by directly distributing histogram data, and the reconstruction error caused by different combining methods of histogram. Aiming at the problem of sensitivity to input data parameters in DP-DBSCAN (Differential Privacy-Density-Based Spatial Clustering of Applications with Noise) differential privacy algorithm, the OPTICS (Ordering Points To Identify Clustering Structure) algorithm based on density clustering was applied to differential privacy protection. And an improved differential privacy protection algorithm, called DP-OPTICS (Differential Privacy-Ordering Points To Identify Clustering Structure) was introduced, the sparse dataset was compressed, the same variance noise and different variance noise were used as two noise-adding ways by comparison, considering the probability of privacy information's being broken by the attacker, the upper bound of privacy parameter ε was determined, which effectively balanced the relationship between the privacy of sensitive information and the usability of data. The DP-OPTICS algorithm was compared with the differential privacy protection algorithm based on OPTICS clustering and DP-DBSCAN algorithm. The DP-OPTICS algorithm is between the other two in time consumption. However, in the case of having the same parameters, the stability of the DP-OPTICS algorithm is the best among them, so the improved OP-OPTICS differential privacy protection algorithm is generally feasible.

The conventional Deep Belief Network (DBN) often utilizes the method of randomly initializing the weights and bias of Restricted Boltzmann Machine(RBM) to initialize the network. Although it could overcome the problems of local optimality and long training time to some extent, it is still difficult to further achieve higher accuracy and better learning efficiency owing to the huge difference between reconstruction and original input resulting from random initialization. In view of the above-mentioned problem, a kind of DBN model based on Sparse Denoising AutoEncoder (SDAE) was proposed. The advantage of the advocated model was the feature extraction by SDAE. Firstly, SDAE was trained, and then, the obtained weights and bias were utilized to initialize DBN. Finally, DBN was trained. Experiments were performed on card game data set of Poker hand and handwriting data sets of MNIST and USPS to verify the performance of the proposed model. In Poker hand data set, compared with the conventional DBN, the error rate of the proposed model is lowered by 46.4%, the accuracy rate and the recall rate are improved by 15.56% and 14.12% respectively. The results exhibit that the proposed method is superior to other existing methods in recognition performance.

In order to solve the problem that Local Directional Pattern (LDP) adopts the fixed average block method in the face feature extraction process, which cannot reflect the characteristics of different images well, an improved LDP based on interest point location was proposed. The positions of interest points contained rich feature information, and the interest points could be obtained automatically according to particular image. Firstly, the locations of interest points were decided by Speed Up Robust Feature (SURF) algorithm and K-means clustering algorithm. Secondly, 4-direction LDP (4-LDP) coding was calculated by the feature extraction windows established with each interest point as the center. Finally, the Support Vector Machine (SVM) was used to identify the face. The proposed method was evaluated in Yale and FERET databases and compared with the original LDP, 4-LDP and PCA-LDP (feature extraction method combined Principal Component Analysis and LDP). The experimental results show that the proposed method can obviously improve the recognition rate and stability while ensuring the real-time performance of the system.

Aiming at the vocalization-based bird species classification in natural acoustic environments, an automatic bird vocalization identification method was proposed based on a new Mel-subband parameterized feature. The field recordings were first divided into consecutive frames and the distribution of log-energies of those frames were estimated using Gaussian Mixture Model (GMM) of two mixtures. The frames with respect to high likelihood were selected to compose initial candidate acoustic events. Afterwards, a Mel band-pass filter-bank was first employed on the spectrogram of each event. Then, the output of each subband, i.e. a time-series containing time-varying band-limited energy, was parameterized by an AutoRegressive (AR) model, which resulted in a parameterized feature set consisting of all model coefficients for each bird acoustic event. Finally, the Support Vector Machine (SVM) classifier was utilized to identify bird vocalization. The experimental results on real-field recordings containing vocalizations of eleven bird species demonstrate that the precision, recall and F1-measure of the proposed method are all not less than 89%, which indicates that the proposed method considerably outperforms the state-of-the-art texture-feature-based method and is more suitable for automatic data analysis in continuous monitoring of songbirds in natural environments.

Considering the road network constraint and the uncertainty of moving object location, a new reverse-kNN query on road network termed Probabilistic Bichromatic Reverse-kNN (PBRkNN) was proposed to find a set of uncertain points and make the probability which the kNN of each uncertain point contains the given query point be greater than a specified threshold. Firstly, a basic algorithm called Probabilistic Eager (PE) was proposed, which used Dijkstra algorithm for pruning. Then, the Pre-compute Probabilistic Eager (PPE) algorithm which pre-computes the kNN for each point was proposed to improve the query efficiency. In addition, for further improving the query efficiency, the Pre-compute Probabilistic Eager External (PPEE) algorithm which used grid index to accelerate range query was proposed. The experimental results on the road networks of Beijing and California show that the proposed pre-computation strategies can help to efficiently process probabilistic bichromatic reverse-kNN queries on road networks.

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.

Aiming at the problem that segmented gesture by Kinect depth information usually contains wrist data, which easily causes subsequent false gesture recognition, a gesture segmentation and positioning algorithm based on improved depth information was proposed. Firstly, the gesture binary image was detected based on depth information threshold limit in experimental space. Secondly, according to characteristics of common gestures, accurate gesture was segmented by gesture endpoint detection and variable threshold algorithm. In order to obtain stable segmentation results, morphological processing of segmented gesture was conducted. Lastly, the gesture positioning algorithm was proposed based on the method of combining gesture gravity center coordinates and maximum inscribed circle center coordinates. The experimental results show that the proposed gesture segmentation method has better accuracy and stability than the existing algorithm. The combined gesture positioning is more stable than gesture gravity center positioning and skeletal data positioning of Kinect Software Development Kit (SDK) and it has no singular points.

To defense application-layer Distributed Denial of Service (DDoS) built on the normal network layer, a defense model based on Web behavior trajectory in the Web application server was constructed. User's access behavior was abstracted into Web behavior trajectory, and according to the generation approach about attack request as well as behavior characteristics of user access to Web pages, four kinds of suspicion were defined, including access dependency suspicion, behavior rate suspicion, trajectory similarity suspicion, and trajectory deviation suspicion. The deviation values between normal sessions and attack sessions were calculated to detect the application-layer DDoS to a specific website. The defense model prohibited the user access from DDoS when detecting the attack request generated by the user. In the experiment, real data was acted as the training set. Then, through simulating different kinds of attack request, the defense model could identify the attack request and take the defense mechanism against the attack. The experimental results demonstrate that the model can detect and defense the application-layer DDoS to a specific website.

Concerning the problems of high computation complexity and poor real-time processing capability in Monte Carlo Localization based on Cubature particle filter (CMCL), a new Monte Carlo localization algorithm based on particle filter with Adaptive Multi-Proposal Distribution (AMPD-MCL) was proposed. The proposal distribution in this algorithm was improved by using Cubature Kalman filter and the extended Kalman filter, in which the most recent measurements were added to weaken particle set degeneracy phenomenon. According to the distribution of particles in state space, Kullback-Leibler Distance (KLD) sampling was utilized in re-sampling to adjust the number of particles required for the next iteration of the filter, which reduced the amount of computation. Simulation results proved the effectiveness of Particle Filter with Adaptive Multi-Proposal Distribution (AMPD-PF). Experiments carried out on the Robot Operating System (ROS) showed that the improved algorithm achieved the average localization accuracy at 19.891cm, the number of particles needed for localization was 60, and the localization time was 45.543s; compared with CMCL algorithm, the localization accuracy was increased by 71.03%, the localization time was shortened by 63.10%. The results demonstrate that AMPD-MCL algorithm reduces localization error, adjusts the number of particles in real-time, reduces computation cost, and enhances real-time processing capability.

In order to make the mobile device more convenient and faster decrypt the outsourcing data stored in the cloud, on the basis of Identity-Based Broadcast Encryption (IBBE) system and Identity-Based Encryption (IBE) system, using the technique of outsourcing the decryption proposed by Green et al. (GREEN M, HOHENBERGER S, WATERS B. Outsourcing the decryption of ABE ciphertexts. Proceedings of the 20th USENIX Conference on Security. Berkeley:USENIX Association, 2011:34), a Modified Asymmetric Cross-cryptosystem Proxy Re-Encryption (MACPRE) scheme across the encryption system was proposed. The proposed scheme is more suitable for mobile devices with limited computing power to securely share the data stored in the cloud. When the mobile user decrypts the re-encrypted data, the plaintext can be restored by performing one exponent operation and one bilinear pairing operation, which greatly improves the decryption efficiency of the mobile user and saves the power consumption of the mobile user. The security of this proposed scheme can be reduced to the security of the IBE and IBBE scheme. The theoretical analysis and experimental results show that, the proposed scheme can allow the mobile devices to decrypt data stored in the cloud by spending less time, and ease the problem of limited computing power of the mobile devices. The proposed scheme is more practical.

Focusing on the issue that the Oriented fast and Rotated Brief (ORB) algorithm does not have scale invariance, an improved algorithm based on Speed-Up Robust Feature (SURF) and ORB was proposed. First, the feature points were detected by Hessian matrix, which made the extracted feature points have scale invariance. Second, the feature descriptors were generated by the ORB. Then the K-nearest neighbor algorithm was used for rough matching. Finally, the ratio test, symmetry test, the Least Median Squares (LMedS) theorem was used for purification. When the scale changed, the proposed algorithm's matching precision was improved by 74.3 percentage points than the ORB and matching precision was improved by 4.8 percentage points than the SURF. When the rotation changed, the proposed algorithm's matching precision was improved by 6.6 percentage points than the ORB. The proposed algorithm's matching time was above the SURF, below the ORB. The experimental results show that the improved algorithm not only keeps the rotation invariance of ORB, but also has the scale invariance, and the matching accuracy is improved greatly without decreasing the speed.

Boolean function plays an important role in the design and analysis of symmetric cryptography. Firstly, by studying the balanced property of subfunctions of the disjoint spectra function set, some sufficient conditions were provided that there were three balanced Boolean functions in the set of four plateaued functions. Then, based on three balanced disjoint spectra plateaued functions, a special Boolean permutation and a balanced Boolean function with high nonlinearity, a method of constructing balanced Boolean functions with high nonlinearity was proposed on a small number of variables. The analysis results show that the proposed method can construct the 2 k-variable balanced Boolean functions with the optimal algebraic number and the nonlinearity is not less than 2 ^2k-1-2 ^k-1-2 ^k/2-2 ^{⌈(k-1)/2⌉}.

In order to solve the security threat of resource sharing and privilege existed in cloud service environment, a new cloud trust model based on non-interference theory, namely NICTM, was proposed. The elements existed in cloud service such as domains, actions, situations, and outputs were abstracted to formally define the trusted domain in cloud services. Besides, the theorem of trusted user domain behavior was proved, and the user domain which followed the theorem could be proved to be trusted. Finally the prototype system was built on Xen virtualization platform, and the feasibility of the model was verified by experiments.

The Coding Unit (CU) size in High Efficiency Video Coding (HEVC) is more times than that in previous coding standards, so the error concealment in HEVC has a poor result in video decoding when packet loss occurs. An error concealment method based on block-merging of segmentation block under CU was proposed. Firstly, the correlation between residual energy and block segmentation was analyzed. Secondly, the reference frame residual energy was compared with a set threshold, and the lost CU segmentation block was merged based on the comparison information. Then, the vector extrapolation method was optimized by weights to ensure the applicability of the proposed algorithm in HEVC error concealment. Finally, the optimized vector extrapolation method was used for error concealment of merged block. The experimental results show that,compared with the classic error concealment methods such as the methods of copy, motion compensation, the proposed method guarantees the Structural Similarity Index Measurement (SSIM) of decoded video and improves the Peak Signal-To-Noise Ratio (PSNR) of decoded video in different motility, and the feasibility of the proposed algorithm is verified.

Concering that the traditional access control methods fail to support dynamic authorization and file evaluation, and suffer from malicious re-sharing issue, an Access Control Mechanism with Dynamic Authorization and File Evaluation (DAFE-AC) was proposed. DAFE-AC adopted a dynamic authorization mechanism to monitor authorized users in real-time and allowed users to supervise each other. The file evaluation mechanism in DAFE-AC could dynamically update the access threshold of files. Based on the Hash/index database, DAFE-AC can ensure the uniqueness of files in the system. In DAFE-AC, a user' authorization value can dynamically change with behaviors of other users, and users can perform file evaluation to eliminate malicious re-sharing of files.