In the 3D point cloud semantic segmentation algorithm based on deep learning， to enhance the fine-grained ability to extract local features and learn the long-range dependencies between different local neighborhoods， a neural network based on attention mechanism and global feature optimization was proposed. First， a Single-Channel Attention （SCA） module and a Point Attention （PA） module were designed in the form of additive attention. The former strengthened the resolution of local features by adaptively adjusting the features of each point in a single channel， and the latter adjusted the importance of the single-point feature vector to suppress useless features and reduce feature redundancy. Second， a Global Feature Aggregation （GFA） module was added to aggregate local neighborhood features to capture global context information， thereby improving semantic segmentation accuracy. The experimental results show that the proposed network improves the mean Intersection?over?Union （mIoU） by 1.8 percentage points compared with RandLA-Net （Random sampling and an effective Local feature Aggregator Network） on the point cloud dataset S3DIS， and has good segmentation performance and good adaptability.

Graph Neural Network （GNN） is vulnerable to adversarial attacks， leading to performance degradation， which affects downstream tasks such as node classification， link prediction and community detection. Therefore， the defense methods of GNN have important research value. Aiming at the problem that GNN has poor robustness when being adversarially attacked， taking Graph Convolutional Network （GCN） as the model， an improved Singular Value Decomposition （SVD） based poisoning attack defense method was proposed， named ISVDatt. In the poisoning attack scenario， the attacked graph was able to be purified by the proposed method. When the GCN was attacked by poisoning， the connected edges with large different features were first screened and deleted to keep the graph features smooth. Then， SVD and low-rank approximation operations were performed to keep the low rank of the attacked graph and clean it up. Finally， the purified graph was used for training GCN model to achieve effective defense against poisoning attack. Experiments against Metattack and DICE were conducted on the open source datasets such as Citeseer， Cora and Pubmed， and compared with the defense methods based on SVD， Pro_GNN and Robust Graph Convolutional Network （RGCN）， respectively. The results show that ISVDatt has relatively better defense effect， although the classification accuracy is lower than that of Pro_GNN， but it has low complexity and negligible time overhead. Experimental results verify that ISVDatt can resist poisoning attack effectively with the consideration of both the complexity and versatility of the algorithm， and has a high practical value.

Group activity recognition is a challenging task in complex scenes， which involves the interaction and the relative spatial position relationship of a group of people in the scene. The current group activity recognition methods either lack the fine design or do not take full advantage of interactive features among individuals. Therefore， a network framework based on partitioned attention mechanism and interactive position relationship was proposed， which further considered individual limbs semantic features and explored the relationship between interaction feature similarity and behavior consistency among individuals. Firstly， the original video sequences and optical flow image sequences were used as the input of the network， and a partitioned attention feature module was introduced to refine the limb motion features of individuals. Secondly， the spatial position and interactive distance were taken as individual interaction features. Finally， the individual motion features and spatial position relation features were fused as the features of the group scene undirected graph nodes， and Graph Convolutional Network （GCN） was adopted to further capture the activity interaction in the global scene， thereby recognizing the group activity. Experimental results show that this framework achieves 92.8% and 97.7% recognition accuracy on two group activity recognition datasets （CAD （Collective Activity Dataset） and CAE （Collective Activity Extended Dataset））. Compared with Actor Relationship Graph （ARG） and Confidence Energy Recurrent Network （CERN） on CAD dataset， this framework has the recognition accuracy improved by 1.8 percentage points and 5.6 percentage points respectively. At the same time， the results of ablation experiment show that the proposed algorithm achieves better recognition performance.

Calligraphy appreciation is widely considered to require sequence restoration, while the sequence restoration of painting is ignored in long time. Moreover, the detail feature of brush strokes is considered to enhance the perception of sequentiality. In order to quantify the sequentiality perception and to explore the influence of detail features of strokes on sequentiality perception, a sequentiality perception quantization method based on Markov chain entropy rate was proposed. Firstly, the perceived sequentiality of an individual to the markers on the artwork was modeled as a Markov chain. Then, the entropy rate of the Markov model was calculated to measure the uncertainty of the perceived sequentiality. Finally, the negentropy was used to measure the order of the perceived sequentiality and was normalized to obtain the measurement index:sequentiality perception. The feasibility of this method was verified through the actual measurement of the sequentiality perception of a group of artworks. And based on the proposed sequentiality measurement, the effect of graph transformation on the sequentiality perception of artworks was studied. Experimental results show that the sequentiality consistency keeps high level when rotation or mirror transformation is performed to the original image, while the correctness varies significantly. This means that the feature of brush strokes is not the primary factor in forming the sequentiality experience, and the viewer's own experience of order of strokes is more important in the formation, but this conclusion needs to be further verified.

Aiming at autonomous positioning of wheeled warehous robots, an indoor positioning method based on visual landmark and odometer data fusion was proposed. Firstly, by establishing a camera model, the rotation and translation relationship between the beacon and the camera was cleverly solved to obtain the positioning information. Then, based on the analysis of the characteristics of the angle difference between the gyroscope and the odometer, a method of angle fusion based on variance weight was proposed to deal with low update frequency and discontinuous positioning information problems. Finally, to compensate for a single sensor positioning defect, the odometer error model was designed to use a Kalman filter to integrate odometer and visual positioning information. The experiment was carried out on differential wheeled mobile robot. The results show that by using the proposed method the angle error and positioning error can be reduced obviously, and the positioning accuracy can be improved effectively. The repeat positioning error is less than 4 cm and the angle error is less than 2 degrees. This method is easy to operate and has strong practicability.

A fuzzy multi-objective software reliability allocation model was established, and bacteria foraging optimization algorithm based on estimation of distribution was proposed to solve software reliability redundancy allocation problem. As the fuzzy target function, software reliability and cost were regarded as triangular fuzzy members, and bacterial foraging algorithm optimization based on Gauss distribution was applied. Different membership function parameters were set up, and different Pareto optimal solutions could be obtained. The experimental results show that the proposed swarm intelligence algorithm can solve multi-objective software reliability allocation effectively and correctly, Pareto optimal solution can help the decision between software reliability and cost.

A new Particle Swarm Optimization (PSO) algorithm with linearly decreasing and dynamically changing inertia weight named L-DPSO was presented to solve the problem that the linearly decreasing inertia weight of the PSO cannot match with the nonlinear changing characteristic. The linear strategy of linearly decreasing inertia weight and the nonlinear strategy of dynamically changing inertia weight were used in the algorithm. The weights were given to two methods separately. Using the test functions of Griewank and Rastrigin to compare L-DPSO with linearly decreasing inertia weight (LPSO) and dynamically changing inertia weight (DPSO), the experimental results show that the convergence speed of L-DPSO is obviously superior to LPSO and DPSO, and the convergence accuracy is also increased. At last, the test functions of Griewank and Rastrigin were used to compare L-DPSO with several commonly used inertia weights, and results show that L-DPSO has obvious advantage too.

In order to deal with problem in private cloud environment caused by computing tasks with large amount of data, intensive computing and complex processing, an implementation of private cloud system based on dynamic service was proposed on the basis of public cloud computing and the characteristics of private cloud environment, which was able to adapt large-scale data processing. In this implementation, computing tasks were described by job files, processing workflows were constructed dynamically by job logic, service requests were driven by data streams and the large-scale data processing could be reflected more efficiently in MapReduce parallel framework. The experimental results show that this implementation offers a high practical value, can deal with computing tasks with large amount of data, intensive computing and complex processing correctly and efficiently.

Through analyzing and simulating human visual perception of stereo image, an objective quality evaluation method of stereo image was proposed. The method combined the characteristics of Human Visual System (HVS) with Structural Similarity, using steerable pyramid to simulate multi-channel effects. Meanwhile, the proposed method used stereo matching algorithm to assess the stereo sense. The experimental results show that the proposed objective method achieves consistent stereoscopic image quality evaluation result with subjective assessment and can better reflect the level of image quality and stereo sense.

The current Magnetic Resonance (MR) image reconstruction algorithms based on compressed sensing (CS-MR) commonly use global regularization parameter, which results in the inferior reconstruction that cannot restore the image edges and smooth the noise at the same time. In order to solve this problem, based on adaptive regularization and compressed sensing, the reconstruction method that used the sparse priors and the local smooth priors of MR image in combination was proposed. Nonlinear conjugate gradient method was used for solving the optimized procedure, and the local regularization parameter was adaptively changed during the iterative process. The regularization parameter can recover the image's edge and simultaneously smooth the noise, making cost function convex within the definition region. The prior information is involved in the regularization parameter to improve the high frequency components of the image. Finally, the experimental results show that the proposed method can effectively restore the image edges and smooth the noise.

In this paper, concerning the complexity of calculation, maintenance and matching ambiguity, a Deep Web query interface identification approach based on label coding was proposed after studying the current identification approach of query interface thoroughly. This approach coded and grouped labels by the directivity and the irregularity of arrangement of the query interface. The identification approach of simple attributes and composite attributes and the processing approach of isolated texts were proposed, taking each label group as an independent unit to identify the feature information. The texts matching the elements were determined by the constraints on the label subscript, which greatly reduced the number of texts considered in matching an element and avoided the problem of matching ambiguity caused by massive heuristic algorithm, and the presentation of nested information was solved by twice clustering effectively and efficiently.

Based on the concepts of primitive production process and eigenword of sequences, the index in primitive production process (IPP) of sequences was defined for measuring the complexity of chaotic pseudorandom sequences. In comparison with bifurcation curves and approximate entropy of simulation results, the efficiency and advantages of the proposed measure are obvious. IPP is superior to approximate entropy for distinguishing sequence complexity.

Complexity stability of sequences is one of the important characteristics of chaotic pseudorandom sequences. Based on the definition of the index in primitive production process (IPP) of a sequence, the concept of weight IPP was proposed. Moreover, the absolute change and the relative change of IPP were recommended to measure the complexity stability of chaotic pseudorandom sequences and some conclusions were drawn. Numerical simulations on several chaotic pseudorandom sequences indicate that the proposed indices are effective in measuring complexity stability of short chaotic sequences.

With network speed and the number of rules constantly increasing, pattern matching is becoming the bottleneck in Network Intrusion Detection System （NIDS）. This paper proposed a fast Wu-Manber-like multi-pattern matching algorithm for intrusion detection, called FWM. By subdividing the pattern group into two subgroups and dealing with the two subgroups in different methods, the FWM algorithm enhanced the efficiency of pattern matching. Experimental results show that, when pattern group contains the pattern that is less than three bytes, the FWM algorithm improves average performance by 29%~44% compared to the original NIDS pattern matching algorithm.