Aiming at the problem of low generalization of the license plate detection model， which makes it difficult to reuse in different application scenes of smart transportation， a license plate detection algorithm in unrestricted scenes based on adaptive confidence threshold was proposed. Firstly， a multi-prediction head network model was constructed， in it， the segmentation prediction head was used to reduce the model reuse pre-processing work， the adaptive confidence threshold prediction head was used to improve the model detection ability， and the multi-scale fusion mechanism and bounding box regression prediction head were used to improve the model generalization ability. Secondly， a differentiable binary network training method was adopted to learn model parameters through differentiable binary transformation combined with the training of classification confidence and confidence threshold. Finally， the Connectivity Aware Non-Maximum Suppression （CANMS） method was used to improve the post-processing speed of license plate detection， and the lightweight network ResNet18 was introduced as the backbone network of feature extraction to reduce the model parameters and further improve the detection speed. Experimental results show that in 6 scenes with different constraints in Chinese City Parking Dataset （CCPD）， the proposed algorithm can achieve the average precision of 99.5% and the recall of 99.8%， and achieves the efficient detection rate of 70 frames per second， which are better than the performance of anchor-based algorithms such as Faster Region-Conventional Neural Network （Faster R-CNN） and Single Shot MultiBox Detector （SSD）. On the three supplementary scene test sets， the license plate detection accuracy of the proposed algorithm is higher than 90% in unrestricted scenes with different resolutions， different shooting distances， and different shooting angles of pitch. Therefore， the proposed algorithm has good detection performance and generalization ability in unrestricted scenes， and can meet the requirements of model reuse.

Three-phase imbalance and load peak-valley difference in the distribution network were caused by single-phase charging of Electric Vehicle(EV) in residential area. Therefore, amulti-objective charging scheduling strategy for EV considering load balance was proposed. Based on the three-phase network, the total delay time and charge balance were used as the objective function, and constraints such as load peak-valley difference and three-phase imbalance were taken into account to establish the scheduling model of EV charging for static and online scheduling problems. The multi-objective solution was obtained by the improved Non-dominated Sorting Genetic Algorithm-Ⅱ(NSGA-Ⅱ), and the results were optimized by designing crossover operators, adaptively adjusting mutation probability and local optimization. The Pareto optimal frontier was obtained by setting a certain volume of external archives and crowding distance, and the fuzzy membership method was used to obtain the compromise optimal solution. The influence of number of simultaneously active charging points and three-phase imbalance value on the optimization results was analyzed through an example.The proposed strategy was compared with the disorderly charging strategy so that the validity of the proposed model and strategy was proved.

In the current large-scale data retrieval tasks, learning to hash methods can learn compact binary codes, which saves storage space and can quickly calculate the similarity in Hamming space. Therefore, for approximate nearest neighbor search, hashing methods are often used to improve the mechanism of fast nearest neighbor search. In most current hashing methods, the offline learning models are used for batch training, which cannot adapt to possible data changes appeared in the environment of large-scale streaming data, resulting in reduction of retrieval efficiency. Therefore, the adaptive hash functions were proposed and learnt in online hashing methods, which realize the continuous learning in the process of inputting data and make the methods can be applied to similarity retrieval in real-time. Firstly, the basic principles of learning to hash and the inherent requirements to realize online hashing were explained. Secondly, the different learning methods of online hashing were introduced from the perspectives such as the reading method, learning mode, and model update method of streaming data under online conditions. Thirdly, the online learning algorithms were further divided into six categories, that is, categories based on passive-aggressive algorithms, matrix factorization technology, unsupervised clustering, similarity supervision, mutual information measurement, codebook supervision respectively. And the advantages, disadvantages and characteristics of these algorithms were analyzed. Finally, the development directions of online hashing were summarized and discussed.

Due to the limitation of driving mileage of pure electric vehicles, it is difficult to realize the long-distance transportation service of pure electric vehicles in a short time to meet the commercial requirements. However, due to the characteristics such as small distribution area, small quantity per batch and large batch number of urban logistics, the pure electric vehicles can be considered to complete the urban distribution tasks. In order to meet the requirements of multiple distribution tasks of the vehicle on the same day and consider the specific impact of vehicle load on real-time energy consumption, a distribution model considering the impact of vehicle load on real-time energy consumption was established to meet the customers' service time requirements in a timely manner. Taking city A as an example, an ant colony algorithm was designed to solve the model, so as to make the reasonable path planning and charging strategy arrangement for the distribution tasks of pure electric vehicles. Finally, the feasibility of pure electric vehicles in urban distribution and logistics in the future was analyzed by comparing to the operation with fuel vehicles.

Considering the data security problems of service sharing in SWIM (System Wide Information Management), the risks in the SWIM business process were analyzed, and a malicious data filtering method based on Latent Dirichlet Allocation (LDA) topic model and content mining was proposed. Firstly, big data analysis was performed on four kinds of SWIM business data, then LDA model was used for feature extraction of business data to realize content mining. Finally, the pattern string was searched in the main string by using KMP (Knuth-Morris-Pratt) matching algorithm to detect SWIM business data containing malicious keywords. The proposed method was tested in the Linux kernel. The experimental results show that the proposed method can effectively mine the content of SWIM business data and has better detection performance than other methods.

To help users to quickly retrieve the interesting game strategies, a knowledge driven automatic annotating algorithm for game strategies was proposed. In the proposed algorithm, the game domain knowledge base was built automatically by fusing multiple sites that provide information for each game. By using the game domain vocabulary discovering algorithm and decision tree classification model, game terms of the game strategies were extracted. Since most terms existing in the strategies in the form of abbreviation, the game terms were finally linked to knowledge base to generate the full name semantic tags for them. The experimental results on many games show that the precision of the proposed game strategy annotating method is as high as 90%. Moreover, the game domain vocabulary discovering algorithm has a better result compared with the n-gram language model.

For improving recovery performance of signals sampled by sub-Nyquist sampling system with Compressed Sensing (CS), the block Simultaneous Iterative Hard Thresholding (SIHT) recovery algorithm for joint sparse model based on ε-closure was proposed. Firstly, The CS synthesis model for Multiple Measurement Vector (MMV) of sampling system was analyzed and the concepts of ε-coherence and Restricted Isometry Property (RIP) were proposed. Then, according to the block coherence of redundant dictionaries, the SIHT algorithm was improved by optimizing the support sets in iterations. In addition, the iterative convergence constant was given and the algorithm convergence property was analyzed. At last, the simulation experiments show that, compared with traditional method, the new algorithm can achieve recovery success rate of 100% with enough sampling channels, while the noise suppressing ability was increased by 7 dB to 9 dB and the total execution time was brought down by at least 37.9%, with higher convergence speed.

To solve the problem of small fault detection of instruments in process industry, independent components were extracted by Independent Component Analysis (ICA) from instruments recorded data. And independent component subspaces were established according to the contribution matrix. Fault detection model was constructed in each independent component subspace with statistical variables. A proper ensemble strategy was chosen by combining all the fault detection results. Finally, the instrument with fault was located by contribution algorithm. The simulation results with TE (Tennessee Eastman) process show that this method has higher precision on small fault detection and more flexibility with proper ensemble strategy.

With the development of computer technology and Internet, more and more office hosts have been connected to Internet, the threat of sensitive information leakage becomes serious. Therefore, it is extremely necessary to detect whether documents contain sensitive information. In order to solve the low precision and low recall problems caused by the traditional query expansion retrieval methods, this paper built an ontology of sensitive information for users interest, proposed a concept similarity query expansion algorithm based on the interest ontology, and described an experimental case to verify the feasibility of algorithm. The experimental results show that the proposed algorithm can improve the precision and recall of the traditional methods.

Concerning the premature convergence of invasive weed optimization algorithm, a new invasive weed optimization with multi-population based on chaotic sequence (CMIWO) was proposed. Firstly, chaotic sequence was adopted to initialize population at the initialization of algorithm, which improved the quality of the initial solution. Secondly, threshold was used to estimate the cluster degree of individuals in iterations and if cluster degree was less than threshold, initializing population with chaotic sequence was implemented again, thus the algorithm could effectively jump out of local minima. Thirdly, the weed population was divided into five groups to collaborate so as to discourage premature convergence, thus improving the algorithm's precision and increasing the convergence speed. In the end, the test results on eight test functions show that the proposed algorithm improves the accuracy by 25% to 300% than basic algorithm in terms of optimal value and 50% to 100% for standard deviation.

The character and meaning of agile virtual enterprise was analyzed, and gray relational theory is applied in the partner selection. The paper firstly brought forward gray relational mathematic model, and then lays emphasis on the illustration of agile virtual enterprise evaluation infrastructure, and on the basis of constructing of system digitized model, presented principle, method and approach of muti-levels gray relational selection by means of instance analysis.