Aiming at the problem of vehicle objection detection in daily road scenes， a real?time detection method of traffic information based on lightweight YOLOv4 （You Only Look Once version 4） was proposed. Firstly， a multi?scene and multi?period vehicle object dataset was constructed， which was preprocessed by K?means++ algorithm. Secondly， a lightweight YOLOv4 detection model was proposed， in which the backbone network was replaced by MobileNet?v3 to reduce the number of parameters of the model， and the depth separable convolution was introduced to replace the standard convolution in the original network. Finally， combined with label smoothing and annealing cosine algorithms， the activation function Leaky Rectified Linear Unit （LeakyReLU） was used to replace the original activation function in the shallow network of MobileNet?v3 in order to optimize the convergence effect of the model. Experimental results show that the lightweight YOLOv4 has the weight file of 56.4 MB， the detection rate of 85.6 FPS （Frames Per Second）， and the detection precision of 93.35%， verifying that the proposed method can provide the reference for the real?time traffic information detection and its applications in real road scenes.

Causality extraction is a kind of relation extraction task in Natural Language Processing (NLP), which mines event pairs with causality from text by constructing event graph, and play important role in applications of finance, security, biology and other fields. Firstly, the concepts such as event extraction and causality were introduced, and the evolution of mainstream methods and the common datasets of causality extraction were described. Then, the current mainstream causality extraction models were listed. Based on the detailed analysis of pipeline based models and joint extraction models, the advantages and disadvantages of various methods and models were compared. Furthermore, the experimental performance and related experimental data of the models were summarized and analyzed. Finally, the research difficulties and future key research directions of causality extraction were given.

Natural Language Generation (NLG) technologies use artificial intelligence and linguistic methods to automatically generate understandable natural language texts. The difficulty of communication between human and computer is reduced by NLG, which is widely used in machine news writing, chatbot and other fields, and has become one of the research hotspots of artificial intelligence. Firstly, the current mainstream methods and models of NLG were listed, and the advantages and disadvantages of these methods and models were compared in detail. Then, aiming at three NLG technologies:text-to-text, data-to-text and image-to-text, the application fields, existing problems and current research progresses were summarized and analyzed respectively. Furthermore, the common evaluation methods and their application scopes of the above generation technologies were described. Finally, the development trends and research difficulties of NLG technologies were given.

Aiming at the security problem of remote control brought by illegal device access in Internet of Things (IoT), a contextual authentication method based on device fingerprint was proposed. Firstly, the fingerprint of IoT device was extracted by a proposed single byte analysis method in the interaction traffic. Secondly, the process framework of the authentication was proposed, and the identity authentication was performed according to six contextual factors including device fingerprint. Finally, in the experiments on IoT devices, relevant device fingerprint features were extracted and decision tree classification algorithms were combined to verify the feasibility of contextual authentication method. Experimental results show that the classification accuracy of the proposed method is 90%, and the 10% false negative situations are special cases but also meet the certification requirements. The results show that the contextual authentication method based on the fingerprint of IoT devices can ensure that only trusted IoT terminal equipment access the network.

With the rapid growth of medical literature, it is difficult for physicians to maintain up-to-date knowledge by reading biomedical literatures. An algorithm named MedRank can be used to recommend influential medications from literature by analyzing information network, based on the assumption that "a good treatment is likely to be found in a good medical article published in a good journal, written by good author(s)", recomending the most effective drugs for all types of disease patients. But the algorithm still has several problems:1) the diseases, as the inputs, are not independent; 2) the outputs are not specific drugs; 3) some other factors such as the publication time of the article are not considered; 4) there is no definition of "good" for the articles, journals and authors. An improved algorithm named HIC-MedRank was proposed by introducing H-index of authors, impact factor of journals and citation count of articles as criterion for defining good authors, journals and articles, and recommended antihypertensive agents for the patients suffered from Hypertension with Chronic Kidney Disease (CKD) by considering published time, support institutions, publishing type and some other factors of articles. The experimental results on Medline datasets show that the recommendation drugs of HIC-MedRank algorithm are more precise than those of MedRank, and are more recognized by attending physicians. The consistency rate is up to 80% by comparing with the JNC guidelines.

To address the issue that the non-well-formed components in a component-based system may lead to the whole system working abnormally, an algorithm for checking the well-formedness of a component was proposed based on its Interface Automaton (IA) model, and a relevant prototype tool was developed. Firstly, the reachability graph isomorphic with the given IA was constructed. Secondly, an ordered set including all the transitions of the reachability graph relevant to the IA was obtained by depth-first-searching the reachability graph. Finally, the well-formedness check of a given IA was completed by checking whether each action belonging to a method in the IA could autonomously reach its return action without exception according to the ordered set under the condition that the external environment meets the input hypothesis. As a realization of the proposed algorithm, a relevant prototype tool was developed on Eclipse platform, namely T-CWFC (Tool for Component Well-Formedness Checking). The prototype tool can model the given component, set up its reachability graph, check its well-formedness and output check result message. The validity of the proposed algorithm was verified by running the tool on a set of components.

Due to random movement of the segmented target, rapid change of background, arbitrary variation and shape deformation of object appearance, in this paper, a new unsupervised video segmentation algorithm based on multiple spatial-temporal feature representations was presented. By combination of salient features and other features obtained from pixels and superpixels, a coarse-to-fine-grained robust feature representation was designed to represent each frame in a video sequence. Firstly, a set of superpixels was generated to represent foreground and background in order to improve computational efficiency and get segmentation results by graph-cut algorithm. Then, the optical flow method was used to propagate information between adjacent frames, and the appearance of each superpixel was updated by its non-local sptatial-temporal features generated by nearest neighbor searching method with efficient K-Dimensional tree (K-D tree) algorithm, so as to improve robustness of segmentation. After that, for segmentation results generated in superpixel-level, a new Gaussian mixture model based on pixels was constructed to achieve pixel-level refinement. Finally, the significant feature of image was introduced, as well as segmentation results generated by graph-cut and Gaussian mixture model, to obtain more accurate segmentation results by voting scheme. The experimental results show that the proposed algorithm is a robust and effective segmentation algorithm, which is superior to most unsupervised video segmentation algorithms and some semi-supervised video segmentation algorithms.

Focusing on the issues in review helpfulness prediction methods that training datasets are difficult to construct in supervised models and unsupervised methods do not take sentiment information in to account, an unsupervised model combining semantics and sentiment information was proposed. Firstly, opinion helpfulness score was calculated based on opinion support score of reviews and replies, and then review helpfulness score was calculated. In addition, a review summary method combining syntactic analysis and improved Latent Dirichlet Allocation (LDA) model was proposed to extract opinions for review helpfulness prediction, and two kinds of constraint conditions named must-link and cannot-link were constructed to guide topic learning based on the result of syntactic analysis, which can improve the accuracy of the model with ensuring the recall rate. The F1 value of the proposed model is 70% and the sorting accuracy is nearly 90% in the experimental data set, and the instance also shows that the proposed model has good explanatory ability.

With the development of remote sensing technology, the data type and data volume of remote sensing data has increased dramatically in the past decades which is a challenge for traditional storage mode. A combination of quadtree and Hilbert spatial index was proposed in this paper to solve the the low storage efficiency in HBase data storage. Firstly, the research status of traditional terrain data storage and data storage based on HBase was reviewed. Secondly the design idea on the combination of quadtree and Hilbert spatial index based on managing global data was proposed. Thirdly the algorithm for calculating the row and column number based on the longitude and latitude of terrain data, and the algorithm for calculating the final Hilbert code was designed. Finally, the physical storage infrastructure for the index was designed. The experimental results illustrate that the data loading speed in Hadoop cluster improved 63.79%-78.45% compared to the single computer, the query time decreases by 16.13%-39.68% compared to the traditional row key index, the query speed is at least 14.71 MB/s which can meet the requirements of terrain data visualization.

In order to resolve the low precision and complexity problem of the existing MapReduce model caused by the reasonable phase partitioning granularity, a multi-phase MapReduce Model (MR-Model) with 5 partition granularities was proposed. Firstly, the research status of MapReduce model was reviewed. Secondly, the MapReduce job was divided into 5 phases of Read, Map, Shuffle, Reduce, Write and the specific processing time of each phase was studied. Finally, the MR-model prediction performance was tested by experiments. The experimental results show that MR-Model is suitable for the MapReduce actual job execution process. Compared with the two existing models of P-Model and H-Model, the time accuracy precision of MR-Model can be improved by 10%-30%; in the Reduce phase, its time accuracy precision can be improved by 2-3 times, the comprehensive property of the MR-Model is better.

For the image processing, computer vision and 3D rendering have the feature of massive parallel processing, the programmability and the flexible mode of parallel processing on the Polymorphic Array Architecture for Graphics (PAAG) platform were utilized adequately, the parallelism design method by combing the operation level parallelism with data level parallelism was used to implement the OpenVX Kernel functions and 3D rendering pipelines. The experimental results indicate that in the parallel implementation of image processing of OpenVX Kernel functions and graphics rendering, using Multiple Instruction Multiple Data (MIMD) of PAAG in parallel processing can obtain a linear speedup that the slope equals to 1, which achieves higher efficiency than the slope as nonlinear speedup that less than 1 of Single Instruction Multiple Data (SIMD) in traditional parallel processing of the Graphics Processing Unit (GPU).

In order to improve the robustness and accuracy of relative orientation, an approach combining direct resolving and iterative refinement for relative orientation was proposed. Firstly, the essential matrix was estimated from some corresponding points. Afterwards the initial relative position and posture of two cameras were obtained by decomposing the essential matrix. The process for determining the only position and posture parameters were introduced in detail. Finally, by constructing the horizontal epipolar coordinate system, the constraint equation group was built up from the corresponding points based on the coplanar constraint, and the initial position and posture parameters were refined iteratively. The algorithm was resistant to the outliers by applying the RANdom Sample Consensus (RANSAC) strategy and dynamically removing outliers during iterative refinement. The simulation experiments illustrate the resolving efficiency and accuracy of the proposed algorithm outperforms that of the traditional algorithm under the circumstance of importing varies of random errors. And the experiment with real data demonstrates the algorithm can be effectively applied to relative position and posture estimation in 3D reconstruction.

Object-oriented analysis of polarimetric Synthetic Aperture Radar (SAR) has been used commonly, while the polarimetric decomposition is still based on pixel, which is inefficient to extract polarimetric information. A object-based method was proposed for polarimetric decomposition. The coherent matrix of object was constructed by weighted iteration of scattering coefficient of similarity, and the convergence of coherent matrix was analyzed, therefore polarimetric information could be obtained through the coherent matrix of object instead of pixel， which can improve the efficiency of obtaining polarimetric features. To more fully reflect the terrain target, spatial features of object were extracted. After feature selection, polarimetric SAR image classification experiments using Support Vector Machine (SVM) demonstrate the effectiveness of the proposed method.

Evolutionary Algorithm based on State-space model (SEA) is a novel real-coded evolutionary algorithm, it has good optimization effects in engineering optimization problems. Global convergence of crossover SEA (SCEA) was studied to promote the theory and application research of SEA. The conclusion that SCEA is not global convergent was drawn. Modified Crossover Evolutionary Algorithm based on State-space Model (SMCEA) was presented by changing the comstruction way of state evolution matrix and introducing elastic search operation. SMCEA is global convergent was proved by homogeneous finite Markov chain. By using two test functions to experimental analysis, the results show that the SMCEA are improved substantially in such aspects as convergence rate, ability of reaching the optimal value and operation time. Then, the effectiveness of SMCEA is proved and that SMCEA is better than Genetic Algorithm (GA) and SCEA was concluded.

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.

Evolutionary Algorithm based on State-space model (SEA) is a new evolutionary algorithm using real strings, and it has broad application prospects in engineering optimization problems. Global convergence of SEA was analyzed by homogeneous finite Markov chain to improve the theoretical system of SEA and promote the application research in engineering optimization problems of SEA. It was proved that SEA is not global convergent. Modified Elastic Evolutionary Algorithm based on State-space model (MESEA) was presented by limiting the value ranges of elements in state evolution matrix of SEA and introducing the elastic search. The analytical results show that search efficiency of SEA can be enhanced by introducing elastic search. The conclusion that MESEA is global convergent is drawn, and it provides theory basis for the application of algorithm in engineering optimization problems.

In view of the existence of a lot of spam comments in microblog, a new method based on AdaBoost was proposed to identify spam comments. This method firstly extracted feature vectors which consisted of eight feature values to represent the comments, then trained several weak classifiers which were better than random prediction on these features via AdaBoost algorithm, and finally combined these weighted weak classifiers to build a strong classifier with a high precision. The experimental results on comment data sets extracted from the popular Sina microblogs indicate that the selected eight features are effective for the method, and it has a high recognition rate in the identification of spam comments in microblog.

In order to avoid the problem of blind discretization of traditional classification "discretize first learn second", a new method of associative classification based on lazy thought was proposed. It discretized the new training dataset gotten by determining the K-nearest neighbors of test instance firstly, and then mined associative rules form the discrete dataset and built a classifier for predicting the class label of test instance. At last, the results of contrastive experiments with CBA (Classification Based on Associations), CMAR (Classification based on Multiple Class-Association Rules) and CPAR (Classification based on Predictive Association Rules) carried out on seven commonly used quantitative datasets of UCI show that the classification accuracy of the proposed method can be increased by 0.66% to 1.65%, and verify the feasibility of this method.

The de-aliasing Super-Resolution (SR) algorithm in image frequency domain demands some limits on frame numbers and sub-pixel shifts between frames of input Lower Resolution (LR) images, which limits the application range of this algorithm. Using single frame super-resolution method and resample function, the authors produced 16 frames images which have the same resolution as input images from each input LR image and then selected images that meet the requirements from these produced images. Therefore, a novel de-aliasing SR algorithm in image frequency domain from two to many frames of input LR images was proposed. Three simulation experimental results indicate that the proposed algorithm removes the limits on frame numbers and sub-pixel shifts between frames and makes the PSNR of SR images to be increased by about 5dB.

Based on profound analysis and conclusion of several typical indexing strategies towards traditional search, the unique characteristics of similar graph containment search was discussed and a coverage and support based frequent pattern filtering approach for constructing index for this kind of search was proposed as well. Experimental results show this approach is effective.

A construction method of detector and its relevant dynamic recognition algorithm were put forward by introducing the data field theory to computer immunology. Antibodies are brought up based on self set. By recognizing the unknown self set, the algorithm can decrease the rate of self-immunity, and also improve the antibody set dynamically and overcome the limitations of traditional IDs that have high requirement for self set, thus simplify the way to implement cloning, mutation and memory. The results of experiments show that the new dynamic recognition algorithm makes IDs possess a higher self adaptability and dynamic equilibrium capability.

In order to judge the poisoning quickly through poisoning diagnosis expert system,after analyzing the failure of previous model of inference,a novel model of inference was put forward with which the expert system applies,and the particular method of implementation was introduced.This system firstly set up a decision tree according to the symptoms doctor known and thinking enough of its diagnostic contribution and diagnostic expense,and then found the result(the cause of poisoning) through repeatedly heuristic deduction.System test shows that it achieves the desired result.

A new windows platform wireless emulator was proposed based on the ideas of NIST Net emulator. By means of analyzing and modifying the NDIS module of windows, this emulator was implemented by adding high accuracy timer based on RTC and wireless module. This emulator could emulate a large scale network by using some emulated nodes and some measurement nodes. Experiments validate the functions of the emulator and the accuracy of the emulator.

This thesis took apart the unordered label tree matching into tree structure matching and tree label semantic matching. By combined with the tree structure matching and semantic matching, the thesis changed the traditional tree matching algorithms into approximately matching, and a metadata retrieval method based on three-level tree approximately matching model was put forward. According to this new retrieval method, the accuracy and recall rates would be adjusted by different requirement of users. In the end, this thesis brought forward the retrieval process of XML-oriented metadata, and gave out the applied design of metadata approximately match. The results of experiments prove that the approximately matching model is feasible and efficient in the application of retrieval XML metadata.