In order to solve the problem of feature information loss caused by the introduction of a large number of pooling layers in traditional convolutional neural networks, based on the feature of Capsule Network (CapsNet)——using vector neurons to save feature space information, a network model 3DSPNCapsNet (3D Small Pooling No dense Capsule Network) was proposed for recognizing 3D models. Using the new network structure, more representative features were extracted while the model complexity was reduced. And based on Dynamic Routing (DR) algorithm, Dynamic Routing-based algorithm with Length information (DRL) algorithm was proposed to optimize the iterative calculation process of capsule weights. Experimental results on ModelNet10 show that compared with 3DCapsNet (3D Capsule Network) and VoxNet, the proposed network achieves better recognition results, and has the average recognition accuracy on the original test set reached 95%. At the same time, the recognition ability of the network for the rotation 3D models was verified. After the rotation training set is appropriately extended, the average recognition rate of the proposed network for rotation models of different angles reaches 81%. The experimental results show that 3DSPNCapsNet has a good ability to recognize 3D models and their rotations.

To solve the sparse and cold start problems of recommendation system, a Trust Regularization Euclidean Embedding (TREE) algorithm by fusing trust information was proposed. Firstly, the Euclidean embedding model was employed to embed the user and project in the unified low-dimensional space. Secondly, to measure the trust information, both the project participation degree and user common scoring factor were brought into the user similarity calculation formula. Finally, a regularization term of social trust relationship was added to the Euclidean embedding model, and trust users with different preferences were used to constrain the location vectors of users and generate the recommendation results. In the experiments, the proposed TREE algorithm was compared with the Probabilistic Matrix Factorization (PMF), Social Regularization (SoReg), Social Matrix Factorization (SocialMF) and Recommend with Social Trust Ensemble (RSTE) algorithms. When dimensions are 5 and 10, TREE algorithm has the Root Mean Squared Error (RMSE) decreased by 1.60% and 5.03% respectively compared with the optimal algorithm RSTE on the dataset Filmtrust.While on the dataset Epinions, the RMSE of TREE algorithm was respectively 1.12% and 1.29% lower than that of the optimal algorithm SocialMF. Experimental results show that TREE algorithm further alleviate the sparse and cold start problems and improves the accuracy of scoring prediction.

The key link of air defense command and control system is to evaluate the threat degree of air target according to target situation information, the accuracy of the assessment will have a significant impact on air defense operations. Aiming at the shortcomings of traditional evaluation methods, such as poor real-time performance, heavy workload, low evaluation accuracy, and unable to evaluate multiple objectives simultaneously, a method of air target threat assessment based on Adaptive Crossbreeding Particle Swarm Optimization (ACPSO) and Least Squares Support Vector Machine (LSSVM) was proposed. Firstly, according to the air target situation information, the framework of threat assessment system was constructed. Then, ACPSO algorithm was used to optimize the regularization parameter and kernel function parameter in LSSVM. In order to overcome the disadvantages of the traditional crossbreeding mechanism, an improved cross-hybridization mechanism was proposed, and the crossbreeding probability was adjusted adaptively. Finally, the training and evaluation results of the systems were compared and analyzed, and the multi-target real-time dynamic threat assessment was realized by the optimized system. Simulation results show that the proposed method has the advantages of high accuracy and short time required, and can be used to evaluate multiple targets simultaneously. It provides an effective solution to evaluate the threat of air targets.

Since the Genetic Algorithm (GA) is easy to produce premature phenomenon and has slow convergence rate, an improved GA based on Traditional GA (TGA), called HLGA (Halton-Levenshtein Genetic Algorithm), was proposed for path planning of real restaurant service robots. Firstly, the similarity method based on edit distance optimized the initial population of quasi-random sequence; secondly, the improved crossover probability and mutation probability adjustment formula based on the adaptive algorithm were adopted to cross and mutate the individuals after they had been selected. Finally, the individual fitness values of the safety evaluation factor functions were calculated, and the global optimal solution was obtained by comparison and iteration. Through theoretical analysis and Matlab simulation, the running time of HLGA was decreased by 6.92 seconds and 1.79 seconds compared with TGA and Adaptive Genetic Algorithm based on Improved independent Similarity (ISAGA), and the actual path of planning was more secure and smooth. The simulation results show that HLGA can effectively improve the quality of path planning in practical applications, meanwhile reduces the searching space and the planning time.

In recent years, facial expression recognition has received extensive attention in education, medicine, psychoanalysis and business. Aiming at the problems of not systematic enough and fuzzy concept of facial expression recognition method, the steps and methods of facial expression recognition were reviewed and discussed. Firstly, the commonly used facial expression databases were introduced and the development of facial expression recognition was reviewed. Then, two aspects of facial expression recognition were introduced, such as facial expression coding and facial expression recognition. The four processes of face facial expression recognition were summarized. The classical algorithms, the basic principles of these algorithms and the comparisons of their advantages and disadvantages were summarized emphatically in the two processes of feature extraction and facial expression classification. Finally, the existing problems and possible development trends in the future of the current facial expression recognition were pointed out.

Intelligent air combat is a hot research topic in military aviation field and missile hit prediction is an important part of intelligent air combat. Aiming at the shortcomings of insufficient research on missile hit prediction, poor optimization ability of the algorithm, and low prediction accuracy of the model, a missile hit prediction model based on Adaptively-Mutated Chaotic Particle Swarm Optimization (AMCPSO) and Support Vector Machine (SVM) was proposed. Firstly, feature extraction of air combat data was carried out to build sample library for model training; then, the improved AMCPSO algorithm was used to optimize the penalty factor C and the kernel function parameter g in SVM, and the optimized model was used to predict the samples; finally, comparison tests with classical PSO algorithm, the BP neural network method and the method based on lattice were made. The results show that the global and local optimization ability of the proposed algorithm are both stronger, and the prediction accuracy of the proposed model is higher, which can provide a reference for missile hit prediction research.

Concerning ranging difficulties with 60GHz signals in Non Line of Sight (NLOS) conditions, a new positioning algorithm based on beamforming in Wireless Local Area Network (WLAN) was proposed. Firstly, the beamforming technology was applied to search the strongest path by adjusting receiving antennas along the channel path with the maximum power.The searching robustness was enhanced and the location coverage was expanded. Secondly, the time delay bias in NLOS conditions was modeled as a Gaussian random variable to reconstruct the NLOS measurements. Finally, to further improve the positioning accuracy, the outlier detection mechanism was introduced by setting a reasonable detection threshold. The localization simulation experiments were conducted on Matlab using STAs-STAs (STAtions-STAtions) channel model, the Time of Arrival (TOA) localization algorithm based on traditional coherent estimation method achieved the average positioning error at about 2m, and the probability of 1m localization accuracy was just 0.5% under NLOS conditions, while the proposed algorithm achieved the average positioning error at 1.02cm, and the probability of 1m localization accuracy reached 94%. Simulation results show that the beamforming technology is an effective solution to 60GHz localization in NLOS conditions, and the localization accuracy and the probability of successful localization are effectively improved.

To further reduce the great computational complexity for High Efficiency Video Coding (HEVC) intra prediction, a novel algorithm was proposed in this paper. First, in Coding Unit (CU) level, the minimum Sum of Absolute Transformed Difference (SATD) of current CU was used to decide an early termination for the split of this CU at each depth level: if the minimum SATD of this CU is smaller than the given threshold value. Meanwhile, based on statistical analysis, the probabilities of each candidate prediction modes being optimal mode were used to further reduce the number of candidate modes which have almost no chance to be selected as the best mode. The experimental results show that, the proposed algorithm can save an average of 30.5% of the encoding time with negligible loss of coding efficiency (only 0.02dB Y-PSNR(Y-Peak Signal-to-Noise Ratio) loss) compared with the reference model HM10.1. Besides, the proposed algorithm is easy to provide software and hardware implementations, and it is also easy to be combined with other methods to further reduce the great computational complexity for HEVC intra coding.

Key management is the core security issue in Wireless Sensor Network (WSN). Based on bivariate symmetric polynomials of random key pre-distribution scheme, a safety mechanism was provided for the nodes communication; however, it has the "t-security" problem and vulnerable to the node capture attack. In order to well solve this problem, and to improve the safety threshold value of network and node ability of anti-trapping, the authors used the convertible ternary polynomial instead of binary polynomial to establish inter-node communication and introduce the key distribution node to do key distribution of the cluster network. The new scheme improved the network anti-trapping ability. Moreover, decoding the sensor node key became almost impossible because of the use of the one-way Hash function. The analytical results show that this scheme has higher anti-destruction ability, scalability and security, and also reduces the overhead of storage and computation among common sensor nodes.

Based on the research of two-dimensional code and recognition of mobile phone, the authors have developed a two-dimensional code recognition system on Android platform. In the system architecture design, the two-dimensional code recognition system platform support layer, client application layer, cloud handle layer were hierarchically designed to ensure the system security. According to the requirements, the system function models and cases analyses were given. Finally, the two-dimensional code core encode and decode functions of mobile phone were achieved.

Traditional image texture mapping can increase abundant details for virtual object.As virtual lighting condition varies,textures lack realistic representation with dynamic lighting changes.Bump mapping can get basic shadow varying by disturbing normal vectors.But it is difficult for a realistic photo to create a Bump map.Some concepts and theories of BRDF & BTF were described.An image-based method was offered,which could dynamically represent surface detail of diffuse objects.Using it,the mapped textures can represent dynamical effect in different virtual lighting.

A secure model was proposed which not only relied on prevention controls but also incorporated detect intrusion and tolerant intrusion. The model can detect intrusions and locate and repair the damage caused by intrusion based on transactions, which protect the database system and provide secure services to legitimate clients. The techniques of both redundancy and diversity were utilized for secret data storage. Threshold secret share approach was presented to protect confidential data based on Chinese remainder theorem.