Aiming at the problem that the existing ensemble classification methods have poor learning ability for unbalanced time series data, the idea of optimizing component algorithm performance and integration strategy was adopted, and based on the heterogeneous ensemble method Hierarchical Vote Collective of Transformation-based Ensembles (HIVE-COTE), an ensemble classification algorithm IMHIVE-COTE (Imbalanced Hierarchical Vote Collective of Transformation-based Ensembles) for unbalanced time series was proposed. The algorithm mainly contains two improvements:first, a new unbalanced classification component SBST-HESCA (SMOM ( K-NN-based Synthetic Minority Oversampling algorithm for Multiclass imbalance problems) & Boosting into ST-HESCA (Shapelet Transformation-Heterogeneous Ensembles of Standard Classification Algorithm) algorithm) was added, the idea of boosting combined with resampling was introduced, and the sample weights were updated through cross-validation prediction results, so as to make the re-sampling process of the dataset more conducive to improving the classification quality of minority samples; second, the HIVE-COTE calculation framework was improved by combining the SBST-HESCA component, and the weight of the component algorithm was optimized, so that the unbalanced time series classification algorithm had higher voting weight to the classification result, as a result, the overall classification quality of the ensemble algorithm was further improved. The experimental part verified and analyzed the performance of IMHIVE-COTE:compared with the comparison methods, IMHIVE-COTE had the highest overall classification evaluation, and the best, the best and third overall classification evaluation on three unbalanced classification indexes. It is proved that IMHIVE-COTE's ability to solve the problem of unbalanced time series classification is significantly better.

In order to use computer vision technology to accurately detect the fall of the elderly, aiming at the incompleteness of existing fall detection algorithms caused by artificial designing of features and the problems in the fall detection process such as the difficulty of separating foreground and background, the confusion of objects, the loss of moving objects, and the low accuracy of fall detection, a deep learning fall detection algorithm with the fusion of human motion information was proposed to detect the fall state of human body. Firstly, foreground and background were separated by the improved YOLOv3 network, and human object was marked by minimum bounding rectangle according to the detection results of YOLOv3 network. Then, by analyzing the motion features in the process of human fall, the motion features of human body were vectorized and transformed into the motion weight information between 0 and 1 through the Sigmoid activation function. Finally, in order to classify human falls, the motion features and the features extracted by Convolutional Neural Network (CNN) were spliced and fused through the fully connected layer. The proposed fall detection algorithm was compared with human object detection algorithms such as background difference, Gaussian mixture, VIBE (VIsual Background Extractor), Histogram of Oriented Gradient (HOG) and human fall judgment schemes such as threshold method, grading method, Support Vector Machine (SVM) classification, CNN classification, and tested under different lighting conditions and the interference of mixed daily noise motion. The results show that the proposed algorithm is superior to traditional human fall detection algortihms in environmental adaptability and fall detection accuracy. The proposed algorithm can effectively detect the human body in the video and accurately detect the fall state of human body, which further verifies the feasibility and efficiency of the deep learning recognition method with the fusion of motion information in the video fall behavior analysis.

Focusing on the problem that the best neighbor nodes of the communities can not accurately be found in most local community detection algorithms, an improved local community detection algorithm was proposed based on local modularity. The concept of node intimacy was introduced to quantify the relationship between the community and the neighbor nodes by the algorithm, and the nodes were selected into the communities according to the node intimacy in descending order. In the end,the extension of the local community was terminated by the local modularity index. Compared with the four kinds of typical community detection algorithms such as the random walk algorithm based on information compression, the algorithm was applied in the real networks and the artificial simulation network. The comprehensive evaluation indexs (F1score) and Normalized Mutual Informations (NMI) of the results are better than comparison algorithms. The experiments show that the algorithm has better efficiency and accuracy, and is very suitable for community detection in a large scale network.

Focused on the problem of traditional object recognition algorithm that the artificially designed features were more susceptible to diversity of object shapes, illumination and background, a deep convolutional neural network algorithm was proposed for object recognition. Firstly, this algorithm was trained with NYU Depth V2 dataset, and single depth information was transformed into three channels. Then color images and transformed depth images in the training set were used to fine-tune two deep convolutional neural networks, respectively. Next, color and depth image features were extracted from the first fully connected layers of the two trained models, and the two features from the resampling training set were combined to train a Linear Support Vector Machine (LinSVM) classifier. Finally, the proposed object recognition algorithm was used to extract super-pixel features in scene understanding task. The proposed method can achieve a classification accuracy of 91.4% on the test set which is 4.1 percentage points higher than SAE-RNN (Sparse Auto-Encoder with the Recursive Neural Networks). The experimental results show that the proposed method is effective in extracting color and depth image features, and can effectively improve classification accuracy.

To solve the problems of synchronization error accumulation and unbalanced energy consumption in multi-hop wireless sensor networks, a cluster-based and energy-balanced time synchronization algorithm for wireless sensor networks was proposed. Based on hierarchical clustering topology, cluster heads in adjacent layers adopted pairwise broadcast mechanism instead of bidirectional pair-wire synchronization mechanism to reduce communication overhead and the synchronization error of transmission delay. Cluster members synchronized the cluster head using the combination of bidirectional pair-wise synchronization and reference broadcast synchronization. In addition, the response node was selected according to residual energy to balance energy consumption of cluster nodes. The performance of synchronization precision and energy consumption of the proposed algorithm and traditional algorithm were analyzed by theoretical analysis and simulation. The results show that the new algorithm not only ensures high synchronization accuracy, but also reduces communication overhead and balances network node energy consumption to lengthen the cycle life of the network.

As a kind of pure location routing algorithm in Wireless Multimedia Sensor Network (WMSN), Two Phase Geographic Greedy Forwarding (TPGF) helps to select the next-hop node which is of nearest distance to the destination from neighbor ones. In some cases, the distance between the next-hop node and the destination node could be farther than that of the current node and the destination node; At the same time, by numbering the nodes, TPGF solves the problem of hole and satisfies the Quality of Service (QoS) requirements. In line with the strategy of selecting the next-hop node farther than the current node, action-angle variables and DATF (Direction-Angle Greedy Forwarding) algorithm were introduced to guarantee and optimize the process of selecting the bound nodes. The simulation result indicates that DATF algorithm shows better performance than TPGF in both energy consumption and end-to-end transmission delay and also has a significant effect on solving the problem of hole.

The existing optimization algorithm for solving the Traveling Salesman Problem (TSP) easily falls into local extremum. To overcome this shortcoming, a new optimization method based on the particle filter, which regarded the searching process of the best route of TSP as a dynamic time-varying system, was brought forward. The basic ideas using particle filter principle to search the best route of TSP were enunciated, and its implementation procedures were given. In order to reduce the possibility of sinking into local extreme, the crossover and mutation operator of Genetic Algorithm (GA) was introduced into the new optimization algorithm to enhance the variety of particles in the sampling process. Finally, the simulation experiments were performed to prove the validity of the new method. The new optimization method based on particle filter can find better solutions than those of other optimization algorithms.

Using the object-oriented technology, the paper generalized and abstracted the embedded-GIS into such models as element, feature, map representation & rendering, geo-data computation & analysis, spatial-data compression & index and web services supporting. It discussed the design, organized relation, and management method, implementation technique between these models. At the end of paper, it gave some brief statements on models application and existed problem.