The image caption model needs to extract features in the image, and then express the features in sentence by Natural Language Processing (NLP) techniques. The existing image caption model based on Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) have the problems of low precision and slow training speed during the extraction of key information from the image. To solve the problems, an image caption generation model based on convolutional attention mechanism and Long Short-Term Memory (LSTM) network was proposed. The Inception-ResNet-V2 was used as the feature extraction network, and the full convolution operation was introduced in the attention mechanism to replace traditional full connection operation, reducing the number of model parameters. The image features and the text features were effectively fused together and sent to the LSTM unit for training in order to generate the semantic information to caption image content. The model was trained by the MSCOCO dataset and validated by a variety of evaluation metrics (BLEU-1, BLEU-4, METEOR, CIDEr, etc.). The experimental results show that the proposed model can caption the image content accurately and perform better than the method based on traditional attention mechanism on various evaluation metrics.

Focused on the problem that when traditional Non-negative Matrix Factorization (NMF) is applied to clustering, robustness and sparsity are not considered at the same time, which leads to low clustering performance, a sparse Non-negative Matrix Factorization algorithm based on Kernel technique and HyperGraph regularization (KHGNMF) was proposed. Firstly, on the basis of inheriting good performance of kernel technique, L _2,1 norm was used to improve F-norm of standard NMF, and hyper-graph regularization terms were added to preserve inherent geometric structure information among the original data as much as possible. Secondly, L _2,1/2 pseudo norm and L _1/2 regularization terms were merged into NMF model as sparse constraints. Finally, a new algorithm was proposed and applied to image clustering. The experimental results on six standard datasets show that KHGNMF can improve clustering performance (accuracy and normalized mutual information) by 39% to 54% compared with nonlinear orthogonal graph regularized non-negative matrix factorization, and the sparsity and robustness of the proposed algorithm are increased and the clustering effect is improved.

Concerning the low accuracy of existing software birthmark extraction algorithms in detecting code theft problem, a new static software birthmark extraction algorithm was proposed. The birthmark generated by this algorithm covered two kinds of software features. The source program and the suspicious program were preprocessed to get the program meta data, which was used to generate Application Programming Interface (API) call set and instruction sequence as two features. These two features were synthesized to generate software birthmarks. Finally, the similarity of source program and suspicious program was calculated to determine whether there was code theft between the two programs. The experimental result verifies that the birthmark combined by API call set and instruction sequence has credibility and resilience, and has stronger resilience compared with k-gram birthmark.

Due to the downward closure property over uncertain data, existing solutions of mining all the frequent itemsets may lead an exponential number of results. In order to obtain a reasonable result set with small size, frequent closed itemsets discovering over uncertain data were studied, and a new algorithm called Normal Approximation-based Probabilistic Frequent Closed Itemsets Mining (NA-PFCIM) was proposed. The new method regarded the itemset mining process as a probability distribution function, and mined frequent itemsets by using the normal distribution model which supports large databases and can extract frequent itemsets with a high degree of accuracy. Then, the algorithm adopted the depth-first search strategy to obtain all probabilistic frequent closed itemsets, so as to reduce the search space and avoid redundant computation. Two probabilistic pruning techniques including superset pruning and subset pruning were also used in this method. Finally, the effectiveness and efficiency of the proposed methods were verified by comparing with the Possion distribution based algorithm called A-PFCIM. The experimental results show that NA-PFCIM can decrease the number of extending itemsets and reduce the complexity of calculation, it has better performance than the compared algorithm.

Integrity is a critical parameter for Global Positioning System (GPS)/ Strapdown Inertial Navigation System (SINS) tight coupling system. In order to reduce satellites' failure detection time, an innovation extrapolation method based on the innovation test method was proposed. By disposing the innovation produced in the extrapolation process, the innovation extrapolation method's test statistics that has been used for failure detection was formed. Applying the proposed method in GPS/SINS tightly coupled system, the simulation results show that innovation extrapolation method can detect slowly growing failure faster than the innovation test method, and innovation extrapolation method can undermine the effect of outliers for failure detection.