In consideration of the problems of low utilization rate of action information and insufficient attention of temporal information in video human action recognition, a human action recognition model based on tightly coupled spatiotemporal two-stream convolutional neural network was proposed. Firstly, two 2D convolutional neural networks were used to separately extract the spatial and temporal features in the video. Then, the forget gate module in the Long Short-Term Memory (LSTM) network was used to establish the feature-level tightly coupled connections between different sampled segments to achieve the transfer of information flow. After that, the Bi-directional Long Short-Term Memory (Bi-LSTM) network was used to evaluate the importance of each sampled segment and assign adaptive weight to it. Finally, the spatiotemporal two-stream features were combined to complete the human action recognition. The accuracy rates of this model on the datasets UCF101 and HMDB51 selected for the experiment and verification were 94.2% and 70.1% respectively. Experimental results show that the proposed model can effectively improve the utilization rate of temporal information and the ability of overall action representation, thus significantly improving the accuracy of human action recognition.

For the problems such as poor unmixing results and sensitivity to noise of traditional Nonnegative Matrix Factorization (NMF) applied to hyperspectral unmixing, a Manifold Regularized Sparse NMF with superpixel (MRS-NMF) algorithm for hyperspectral unmixing was proposed. Firstly, the manifold structure of hyperspectral image was constructed by superpixel segmentation based on entropy. The original image was divided into k-superpixel blocks, and the data points in each superpixel block with same property were labeled the same label. Weight matrices were defined between any two data points with the similar label in a superpixel block, and then the weight matrices were applied to the objective function of NMF to construct the manifold regularization constraint. Secondly, a quadratic parabola function was added to the objective function to complete the sparse constraint. Finally, the multiplicative iterative update rule was used to solve the objective function to obtain the solution formulas of endmember matrix and abundance matrix. At the same time, maximum iteration times and tolerate error threshold were set to get the final results by iterative operation. The proposed method makes full use of spectral and spatial information of hyperspectral images. Experimental results show that on synthetic data the unmixing accuracies of endmember and abundance based on proposed MRS-NMF are 0.016-0.063 and 0.01-0.05 respectively higher than those based on traditional methods like Graph-regularized L1/2-Nonnegative Matrix Factorization (GLNMF), L1/2NMF and Vertex Component Analysis-Fully Constrained Least Squares (VCA-FCLS); while on real hyperspectral images, the average unmixing accurary of endmember based on proposed MRS-NMF is 0.001-0.0437 higher than that of traditional GLNMF, Vertex Component Analysis (VCA) and Minimum Volume Constrained Nonnegative Matrix Factorization (MVCNMF). This proposed algorithm improves the accuracy of unmixing effectively with good robustness to noise.

Nodes in the Content-Centric Network (CCN) cache all the passed content by default, without selective caching and optimally placing of the content. In order to solve the problems, a new Cooperative Caching strategy based on User Preference (CCUP) was proposed. Firstly, user's preference for content type and content popularity were considered as user's local preference indexes to realize the selection of cached content. Then, the differentiated caching strategy was executed on the content that needed to be cached, the globally active content was cached at the important central node, and the inactive content was cached according to the matching of the local preference and the distance level between node and user. Finally, the near access of user to local preference content and the quick distribution of global active content were both achieved. The simulation results show that, compared with typical caching strategies, such as LCE (Leave Copy Everywhere)、Prob(0.6) (Probabilistic caching with 0.6)、Betw (cache "less for more"), the proposed CCUP has obvious advantages in average cache hit rate and average request delay.

To solve the problem of that although Intra String Copy (ISC) improved the effect of the screen content coding, but it transformed the 2D image to 1D by Coding Unit (CU), making adjacent regions in an image segmented and spatial correlation not to be used, a new algorithm called 2D Intra String Copy (2D ISC) was proposed. Almost without additional memory in encoder and decoder, the algorithm realized arbitrary 2D shape searching and matching without boundary restriction of CU for pixels in current CU, by using dictionary coding tool in High Efficiency Video Coding (HEVC) reconstruction cache. Also adopted technologies of color quantization preprocessing and horizontal vertical search order self-adaption to enhance coding effect. Experiments on common test for typical screen content test sequences show that compared with HEVC, 2D ISC can achieve bit-rate saving of 46.5%, 34.8%, 25.4% for All Intra(AI), Random Access(RA) and Low-delay B(LB) configurations respectively in lossless coding mode, and 34.0%, 37.2%, 23.9% for AI, RA and LB configurations respectively in lossy coding mode. Even compared with ISC, 2D ISC can also achieve bit-rate saving up to 18.3%, 13.9%, 11.0% for AI, RA and LB configurations in lossless coding mode, and 19.8%, 20.5%, 10.4% for AI, RA and LB configurations in lossy coding mode. The experimental results indicate that the proposed algorithm is feasible and efficient.

The existing similarity search algorithms do not consider the time factor. To address this problem, a meta path-based dynamic similarity search algorithm named PDSim was proposed for the heterogeneous information network. Firstly, PDSim calculated the link matrix of object under the given meta-path, thus obtained the instances ratio of meta-path between different objects. Meanwhile, the differences of establishing time were calculated. Finally, the dynamic similarity was measured under the given meta-path. In multiple instances of the similarity search, PDSim kept up with the interest variation of object which dynamically changed with time. Compared with the PathSim (Meta Path-Based Similarity) and PCRW (Path-Constrained Random Walks) methods, the clustering accuracy of Normalized Mutual Information (NMI) could be increased by 0.17% to 9.24% when applied to clustering. The experimental results show that, compared to the traditional similarity search algorithm based on link, the efficiency of dynamic similarity search and the satisfaction of user of PDSim are significantly improved, and it is a dynamic similarity search algorithm for object changes with time.

The methods of image super-resolution via sparse representation achieve good quality image reconstruction, but the CPU-based implementation of the methods hardly satisfies the requirement of real-time video super-resolution because of high computational complexity. Then, the method of real-time video super-resolution via sparse representation based on GPU acceleration was proposed. It focused on optimizing data parallel processing and improving resource utilization of GPU, including utilizing queues for video sequences, improving memory concurrent access rates, employing Principal Component Analysis (PCA) dimensionality reduction and optimizing dictionary querying operation. As a result, compared with the CPU-based implementation, the speed of data processing is increased two orders of magnitude, and the speed of playing a video with the size of 669×546 reaches 33 frames per second.

Concerning the poor localization precision of Distance Vector-Hop (DV-Hop), a calibration based DV-Hop algorithm with credible neighborhood distance estimation (CDV-Hop) was proposed, which defined a new measure to estimate the neighborhood distances by relating the proximity of two neighbors to their connectivity difference, and then calculated the more accurate neighborhood distances. According to the unique location relationship between the unknown nodes and their nearest anchor nodes, this algorithm added the calibration step, which took the credible neighborhood distances as the calibration standard to correct the position of unknown nodes. The simulation results show that the CDV-Hop algorithm works stably in different network environment. With the ratio of anchor nodes increasing, there is an improvement of 4.57% to 10.22% in localization precision compared with DV-Hop algorithm and 3.2% to 8.93% in localization precision compared with Improved DV-Hop (IDV-Hop) algorithm.

In Mobile Ad Hoc Network (MANET), each node forwards a message in the traditional routing protocol such as flooding and expanding ring search, which results in heavy overhead and long latency of routing. In order to improve the performance of routing protocol, a scheme of probabilistic forwarding algorithm was provided based on directional prediction. The information such as ID and time was extracted from data packets by monitoring network, and a table was established to store these records which can hint the distance to the destination node. Based on these records, the node's forwarding probability was calculated and adaptively adjusted according to the network. Whether some node should forward a packet depended on the forwarding probability, which was high enough only for sustaining the routing process towards the destination. The simulation results show that the routing overhead declined up to 70% compared with flooding algorithm and 20% compared with the classical probabilistic forwarding algorithm. The new scheme significantly improved the performance of the network.

With the explosive growth of unstructured data, the existing storage technology in the aspects of I/O throughput, scalability and manageability needs improving urgently. Based on cloud storage and reliability theory, a model of distributed storage for unstructured data was created, and reliability function was also proposed. The distributed Relational Database Management System (RDBMS) was adopted as the bottom storage facilities, so unstructured data could be stored directly in the data table. Separated storage and unified management for unstructured data and metadata was realized, and thus storage system performance was promoted. Relative to the centralized storage, new system has superior availability. The simulation results show that the storage system has higher reliability and it is easy to expand. The distributed storage system can be applied to dynamic open computing environment, and it provides cloud storage service with better performance.

Concerning the issue of extracting features more efficiently from a sequence of gait frames and real-time recognition, an effective human recognition method based on Mean Gait Energy Image (MGEI) was described, which utilized Kernel Principal Component Analysis (KPCA). A pre-processing technique was used to segment the moving silhouette from the walking figure. The algorithm obtained the gait quasi-periodicity through analyzing the width information of the lower limbs' gait contour edge, and the MGEI was calculated from gait period. KPCA extracted principal component with nonlinear method and described the relationship among three or more pixels of the identified images. In this paper, KPCA could make use of the high correlation between different MGEIs for feature extraction by selecting the proper kernel function, and Euclidean distance weighted by variance reciprocal was designed as the classifier. The experimental results show that the proposed approach has better recognition performance and the computation time is greatly reduced.

Mining the hidden knowledge in the spatial data of Geographic Information System (GIS) is an important direction in the study fields of GIS and data mining. The technique of concept lattice is very important in finding association rules. In this paper, an algorithm of building concept lattice based on incremental method was proposed to improve the speed of building lattice by comparing the extent of the concept and introducing support constraint. It could also simplify the lattice and make it easier to mine rules. To expand the application of the algorithm, it was applied in finding association rules of the spatial data in GIS and got practicable application result.

The multicast tree in MAODV is reconstructed frequently because of the nodes mobility, making the cost of routing and delay of transmission increase significantly. The neighbor change ratio based stable path selection method was proposed to overcome the shortcomings. And a new neighbor change ratio calculation method was put forward, which did not need to send Hello messages timely. Based on this new method, a stable path based MAODV (SP-MAODV) multicast routing protocol was given with a stable path selection and less hops. The simulation results on data packet transmission rate, routing overhead, average end-to-end delay and delay jitter show that the new protocol reduces interruption probability of the path and improves performance of the protocol.