Ontology fusion has become an important way to reconstruct knowledge map and share knowledge in the subject field. In order to solve the problem of lack of standard fusion framework in ontology fusion field, we propose an ontology fusion model based on international standard MFI4OR. This model provides a standard ontology information partition standard, namely ontology-ontology component-ontology atomic component, and realizes the management and mapping of ontology information. In the process of fusion calculation, we choose editing distance algorithm and introduces external resources WordNet dictionary to calculate similarity. Finally, based on the application background of learner model construction requirements in provincial fund projects, the FOAF (Friend-Of-A-Friend)and RELATIONSHIP ontologies are fused. The results show that the standardized fusion model can achieve fusion and has good recall rate and accuracy.

Real natural images are often contaminated by various kinds of image noise. Traditional denoising methods are generally designed for only one type of noise, so the denoising effect is not good when dealing with mixed noise. To solve this problem, a fast two-step cross non-local hybrid filtering algorithm is proposed. Firstly, the extremum point of pixel gray level is located. The non local median filter is used to remove salt and pepper noise, and the difference integral image of the image is obtained. And then the improved non mean filter is used to further remove the noise. Finally, the non local median filter is used to further remove the noise. The experimental results show that the proposed algorithm achieves higher measurement index and better visual effect in the case of high intensity mixed noise pollution.

When the current privacy protection algorithm for network users is implemented, the noise points in the data are not removed, which leads to long time-consuming encryption, low security index, and poor integrity, which seriously affects the application experience of network users. Therefore, a network based on probability statistics is proposed using privacy protection algorithm. First, the wavelet transform threshold method is used to reduce the noise of the data to eliminate the uncertain factors generated during the data collection, so that the algorithm consumes less time in the encryption process. In order to improve the security index of the algorithm, the analytic hierarchy process is used in probability statistics. On the basis of knowledge, a privacy protection hierarchical analysis model is established, and the model is used to encrypt the obtained privacy protection target value through the Logistic chaotic mapping system to realize the privacy protection of network users. The experimental comparison results show that the proposed algorithm has short time-consuming encryption, high security index and better integrity, and can be widely used.

 In order to more accuratey locate the positioning accuracy of Android terminal, the mobile can be better applied in fan mapping scene, we select two typical measuring points, through 48 hours of continuous observation and using GAMIT / GLOBK software to calculate the “real" coordinates of the measuring points. Two smart phones are used to collect GNSS(Global Navigation Satellite System) raw observation data, and conduct static and dynamic positioning accuracy and accuracy comparison experiments. The results show that the positioning results of the Samsung Galaxy S8 are generally better than those of the Huawei P10; the static post- processing, the LGO(Leica Geo Office) solution result is better than the RTKLIB solution; the dynamic real- time solution of Samsung Galaxy S8 achieves a fixed solution and has better accuracy ( better than 1dm). The results of Huawei P10 are poor, with accuracy ranging from 3. 18 m to 5. 52 m, accuracy ranging from 2. 67 m to 4. 88 m, and no fixed solution is achieved. The positioning accuracy and accuracy of consumer-grade GNSS mobile terminals are fully recognized.

Aiming at the large number of defects at the electron transport layer/ perovskite layer interface of perovskite solar cells, an optimized strategy employing an inorganic salt for interface passivation is proposed. This strategy selects low-cost LiCl(Lithium Chloride) as the interface passivation material between the electron transport layer and perovskite layer, and prepares the perovskite solar cell with the device structure of ITO/ TiO₂ / LiCl / CH₃NH₃PbI₃ / spiro-OMeTAD/ Ag. After the introduction of LiCl with an optimized concentration, the short- circuit current density and fill factor of the perovskite solar cell achieve 21. 05 mA/ cm ² and 72. 55% , and the energy conversion efficiency is 16. 95% , which shows an increase of 23. 00% compared with the device without the introduction of LiCl. After characterizing the device and the film, it is found that LiCl can passivate the defects and traps at the interface and increase the conductivity of TiO₂ , thereby reducing the interface recombination loss and promoting the charge transport.

In order to alleviate the new load peak situation caused by the existing time-of-use pricing mechanism, a power consumption incentive mechanism which can contribute to the load regulation is proposed. At first, all kinds of household electric equipment load model, load model for the battery electric vehicle charging and discharging model are set up. Secondly the load side of indirect adjustment grid peak valley incentive mechanism model is established. Finally to electricity cost and user comfort as the objective function, based on niche chaotic particle swarm optimization algorithm is used for solving multi-objective Pareto solutions. The simulation results show that the proposed incentive mechanism can meet the comfort requirements, and significantly reduce the cost of electricity consumption and the peak-valley difference of electricity consumption, which can relieve the power grid pressure and improve its operation stability to a certain extent.

In order to solve the problems of misjudgment, oscillation and tracking velocity of traditional perturbation observation method, a control strategy based on power prediction variable step size perturbation observation method is proposed. When the sampling frequency is fixed and the light intensity remains constant and the temperature changes, the output power of the photovoltaic cell can be approximately linear in unit time. The rapidness of the traditional perturbation observation method with large step size is used to track the maximum power point, and then the linear power prediction is carried out near the maximum power point, so as to accurately track the maximum power point. A simulation model is built in the simulation platform, and the results show that the new control strategy improves the tracking speed and accuracy of the system, and optimizes the output performance of the system.

An insulator fault detection method based on ensemble convolution wavelet extreme learning neural network is proposed because the traditional methods can not accurately and efficiently identify the faults of insulators due to the remote distribution position and complex background. Firstly, the insulator images data is collected and preprocessed by industrial camera installed on UAV (Unmanned Aerial Vehicle). Secondly, the ensemble convolution wavelet extreme learning neural network is constructed by combining the advantages of convolution neural network, auto encoder, extreme learning machine and wavelet function. Finally, the insulator images samples are fed into multiple deep neural networks for automatic feature learning. The prediction results are assembled and the final fault detection results are output. The experimental results show that the average fault detection accuracy of the proposed method reaches 98. 49% and the standard deviation is only 0. 20. Compared to other methods, it has more advantages in image feature extraction and fault detection accuracy, and is suitable for automatic identification of insulator faults.

In order to solve the poor performance of traditional manifold learning method in feature extraction from non-Euclidean space of bearing data, a Riemann manifold learning method is proposed. Under the framework of Riemannian manifold, the Riemannian manifold is constructed by using the original data set, and based on this manifold, a Riemannian graph embedding feature extraction method is proposed. The preliminary dimensionality reduction is realized by coding the local structure. Then, based on the low-dimensional Riemannian manifold, a classifier is designed to cluster the bearing data by combining the principal component analysis algorithm (PCA: Principal Components Analysis) and the linear discriminant analysis algorithm (LDA: Linear Discriminant Analysis ). Finally, the ability of this method to extract features is analyzed through experiments on two bearing data sets. Compared with the existing fault diagnosis methods, this algorithm has stronger fault diagnosis ability.

With the development of information technology, a large amount of text classification is needed in many industries. In order to improve the accuracy and the efficiency of classification at the same time, a kind of CNN-BiLSTM/ BiGRU mixed text classification model based on the attention mechanism(CBLGA) is proposed, in which parallel CNN(Convolution Neural Networks) with different window sizes to extract a variety of text characteristics, then input the data in BiLSTM/ BiGRU parallel model. BiLSTM/ BiGRU combination model is used to extract global characteristics relate with the whole text context, finally the characteristics of two models are fused and the Attention mechanism is introduced. Secondly, another kind of Attention of CNN-BiLSTM/ CNN mixed text classification model based on the attention mechanism(CBLCA) is proposed, and its feature is divided CNN's output into two parts. One part is input to the BiLSTM network, another is integrated to the output of BiLSTM network. Successfully retaining the partial text features extracted by CNN and the global text features extracted by BiLSTM. After several experiments, the CBLGA model and CBLCA model is achieved effective improvements in accuracy and efficiency. Finally, a set of preprocessing methods for texts with different lengths is established, so the model can improve the accuracy and efficiency of text classification target in long text and short text.

Many governments released some mobile phone app to fight the spread of novel diseases, such as for the COVID-19 pandemic to assist health officials in tracking down exposures after an infected individual is identified. However, there are important privacy implications of the existence of such tracking apps. In order to solve the influence of current contact tracking on user privacy, this paper analyzes, discusses and summarizes the privacy risks and protection issues in current contact tracking for the first time, aiming at improving privacy issues without reducing the usefulness to public health. We hope in writing this document to ensure that privacy is a central feature of conversations surrounding mobile contact tracing apps and to encourage community efforts to develop alternative effective solutions with stronger privacy protection for the users.

Current PHR( Personal Health Records) have the challenges of insecure storage and difficulty in sharing. In order to improve the security and sharing of PHR, a secure sharing model based on Hyperledger and FV( Finger Vein) feature authentication is proposed. The PHR provider combines the FV feature with the clinician's private key signature when entering the medical record to generate a PHR index. Then store the PHR index in Hyperledger Fabric, and store the real private data in Filecoin, reducing storage costs and on-chain storage pressure. And a PHR access control contract is designed to prevent malicious node attacks. Experimental results show that this model has significant advantages in performance and storage.

In order to adapt to the centralized IT system operation and maintenance management form, to improve the user retrieval accuracy and enhance the user service quality, a centralized IT operation and maintenance information retrieval algorithm based on Bayesian network is proposed. We analyze the IT operation and maintenance architecture from the aspects of operation and maintenance strategy, mode and process, defines the overall process from the user submitting the retrieval application to the result feedback; preprocesses the text information to realize the structured display of user browsing content and calculate the user feature vector; uses the directed graph to represent the Bayesian network topology, and obtains the prior probability of the term node and the file node for reasoning In order to complete the construction of operation and maintenance information retrieval model, the probability relationship between file and retrieval is used to filter redundant information. The sample space is established to transform the information retrieval problem into the concept matching problem in the sample space, and the correlation function expression of file and retrieval is obtained and simplified. Simulation results show that the method can improve the recall and precision of information retrieval, and reduce the network load.

The spam message senders continually try to modify spam content for cheating filter system, causing the recognition accuracy to decrease. Aiming at this problem, a new recognition method based on self-clustering and self-learning algorithm is presented. First, the spam relation chain is built by the minimum edit distance to realize self-clustering function using MeanShift algorithm is used on the chain. Second, the core of each cluster is computed, and the weight of each sample is computed by the distance from the cluster core. Then train the classifier by the samples with its weights. When the new spam is recognized by the classifier, it will be classified to a cluster. The core and sample weights of this cluster will be recomputed, and update the classifier to realize the self-learning function this process is repeated. Experiment results demonstrate that the new method can improve the recognition accuracy by 2. 51% ~ 5. 14% , and can keep the high accuracy for a long time.

At present, most of the existing question answering system models use template matching for reasoning, which is not enough for question reasoning. Therefore, a question answering system reasoning model is proposed based on cognitive map. Firstly, the ontology is constructed based on the domain knowledge as the knowledge source, and then the question relation one-to-one cognitive map question answering system model is constructed based on the cognitive map. Finally, the question and answer is divided into simple question and complex question, and the simple question is matched by BERT + CRF(Bidirectional Encoder Representations from Transformers+Conditional Random Field) model. For the complex question, node2vec is used to generate subgraph, then GCN(Graph Convolutional Network) reasoning model is used for reasoning, and the answer is taken as the output result. The proposed model is tested in the field of underground operation, and the results show that the cognitive map Question answering model is better than other algorithm models.

Because the existing algorithms fail to incorporate the principle of minimum confidence and big data technology in the static analysis process, the static analysis results are inaccurate and costly. A static analysis algorithm for embedded software maturity based on minimum confidence and big data is proposed. The minimum confidence is used as the subordination criterion. The static analysis ideas of embedded software maturity are evaluated by weight and satisfaction, and evaluation indicators are constructed. Big data technology is used to standardize and non-dimensionalize various software maturity evaluation indicators to obtain effective indicators. Combining the evaluation dimensions of embedded software maturity, using attribute mathematics principles, the evaluation index measurement is calculated and analyzed in the form of expert questionnaires, a static evaluation model of embedded software maturity is built, and the static analysis of software maturity is completed through the model. The simulation experiment results show that the proposed algorithm can obtain high-accuracy static analysis results of embedded software maturity, and the static analysis cost can also be effectively reduced.

There are many defects in the mobility and convenience of the handwritten digit recognition system running on the computer. In order to make improvement for these defects, a handwritten digit recognition system based on the visual library OpenCV is designed, which transplants the digit recognition algorithm into the flexible and small high-performance embedded equipment. By adjusting the shooting Angle of the steering gear and using the technology of picture splicing and digital segmentation, the handwritten digit recognition of short distance and large area is realized. The recognition speed, recognition accuracy and model volume of the models trained by KNN(K-Nearest Neighbor), support vector machine and artificial neural network are compared. After testing, the identification time of Raspberry Pi by using the artificial neural network algorithm can be as low as 0. 115 s, and the recognition accuracy can reach 72% , which has a certain application value.

In order to let the undergraduates understand biomedical image processing technology and master deep learning methods, combined with the innovation and entrepreneurship training program of Jilin University college students, the experimental project “Research on automated analysis of corneal images based on deep learning" is completed. To assist the medical research and development of effective drugs for the treatment of CNV(Corneal Neovascularization) diseases, it is necessary to observe and obtain data on the growth of mouse corneal blood vessels under the influence of drugs. Therefore, an automated corneal image analysis program based on deep learning is designed, where the gel-processed mouse corneal image provided by the cooperative hospital is used as the research object, and the corneal features are completed through MATLAB tools and deep learning algorithms such as neural networks. SegNet semantic segmentation network and SVM(Support Vector Machine)- based image segmentation are used to achieve automatic extraction of mouse corneal images. The accuracy and reliability of corneal extraction under the two methods are analyzed. The results show that the use of SegNet semantic segmentation network is high in accuracy, and its accuracy rate can reach 97. 75% .