With the continuous development of information technology, the scale of time series data has grown exponentially, which provides opportunities and challenges for the development of time series anomaly detection algorithm, making the algorithm in this field gradually become a new research hotspot in the field of data analysis. However, the research in this area is still in the initial stage and the research work is not systematic. Therefore, by sorting out and analyzing the domestic and foreign literature, this paper divides the research content of multidimensional time series anomaly detection into three aspects: dimension reduction, time series pattern representation and anomaly pattern detection in logical order, and summarizes the mainstream algorithms to comprehensively show the current research status and characteristics of anomaly detection. On this basis, the research difficulties and trends of multi-dimensional time series anomaly detection algorithms were summarized in order to provide useful reference for related theory and application research.

In order to solve the problem of insufficient available training data in the classification task of breast mass and calcification, a multi-view model based on secondary transfer learning was proposed combining with imaging characteristics of mammogram. Firstly, CBIS-DDSM (Curated Breast Imaging Subset of Digital Database for Screening Mammography) was used to construct the breast local tissue section dataset for the pre-training of the backbone network, and the domain adaptation learning of the backbone network was completed, so the backbone network had the essential ability of capturing pathological features. Then, the backbone network was secondarily transferred to the multi-view model and was fine-tuned based on the dataset of Mianyang Central Hospital. At the same time, the number of positive samples in the training was increased by CBIS-DDSM to improve the generalization ability of the network. The experimental results show that the domain adaption learning and data augmentation strategy improves the performance criteria by 17% averagely and achieves 94% and 90% AUC (Area Under Curve) values for mass and calcification respectively.

The existence of heterogeneous information systems in colleges and universities hinders data assets integration and information interaction. The emergence of Service Oriented Architecture (SOA) and its widespread adoption in enterprises provide ideas for solving this problem, while it is difficult to implement SOA and form an SOA-based informational ecosystem in universities. In response to these problems, an SOA construction scheme driven by master data management was proposed. Firstly, a master data management platform was used to model and integrate the core data assets at the data level. In order to realize data synchronization and consumption, and solve the problem of protocol conversion and service authentication in the process, an enterprise service bus based solution was proposed. Then, in order to the transform the legacy "information island" systems to SOA, a construction solution driven by master data was proposed. The experimental results show that the average latency with concurrency single user, 10 users, 100 users and 10000 users is 8, 11, 59 and 18 ms respectively, which indicates that the performance of the proposed scheme meets the need in different concurrent scenarios. The implementation results show that the data assets integration and information interaction problems have been solved, which proves that the scheme is feasible.

The basic scoring index weight distribution of the Common Vulnerability Scoring System (CVSS) relies too much on expert experience, which leads to the lack of objectivity. In order to solve the problem, a vulnerability basic scoring index weight distribution method was proposed. Firstly, the relative importances of scoring elements were sorted. Then, the index weight combination optimal search method was used to search the weight combination scheme. Finally, combined with the grey relation analysis method, the multiple weight distribution schemes based on expert experience decision were used as the input to obtain the weight combination scheme. The experimental results show that, compared with CVSS, from the quantitative point of view, the proposed method has more gentle score distribution of scoring results than the CVSS, which effectively avoids the excessive extreme values, and the discretization of score distribution can effectively distinguish the severity of different vulnerabilities objectively and effectively. The comparative analysis from the qualitative point of view show that, while the vast majority of vulnerabilities (92.9%) in CVSS are designated as the high level of severity, the proposed method can achieve more balanced characteristic distribution in grade distribution of vulnerability severity.

Considering the road network constraint and the uncertainty of moving object location, a new reverse-kNN query on road network termed Probabilistic Bichromatic Reverse-kNN (PBRkNN) was proposed to find a set of uncertain points and make the probability which the kNN of each uncertain point contains the given query point be greater than a specified threshold. Firstly, a basic algorithm called Probabilistic Eager (PE) was proposed, which used Dijkstra algorithm for pruning. Then, the Pre-compute Probabilistic Eager (PPE) algorithm which pre-computes the kNN for each point was proposed to improve the query efficiency. In addition, for further improving the query efficiency, the Pre-compute Probabilistic Eager External (PPEE) algorithm which used grid index to accelerate range query was proposed. The experimental results on the road networks of Beijing and California show that the proposed pre-computation strategies can help to efficiently process probabilistic bichromatic reverse-kNN queries on road networks.

In the video based in-air signature verification system, the existed methods cannot meet the requirement of accuracy, real time, robustness for fingertip tracking. To solve this problem, the Tracking-Learning-Detection (TLD) method based on temporal context was proposed. Based on the original TLD algorithm, the temporal context massage, namely the prior knowledge that the movement of fingertip is continuity in two adjacent frames, was introduced to narrow the search range of detection and tracking adaptively, thereby improving tracking speed. The experimental results on 12 public and 1 self-made video sequences show that the improved TLD algorithm can accurately track fingers, and tracking speed can reach 43 frames per secend. Compared with the original TLD tracking algorithm, the accuracy was increased by 15% and the tracking speed was increased more than 100%, which make the proposed method meet the real-time requirements for fingertip tracking.

In the bridge crane payload space positioning system based on monocular vision, in order to solve the problem of matching performance degradation caused by the rotation and tilt of the target template, a real-time bridge crane payloadspace positioning method based on circle detection with vertical gradient direction lines and linear pre-interpolation was proposed. Aspherical target was attached to the top of the payload, and the spherical target was not sensitive to the rotation and tilt when it was detected. First, the spherical target in the Region of Interest (ROI) was detected accurately and fastly by the circle detection method based on vertical gradient direction lines. Secondly, the space coordinates of the spherical target center was confirmed by the space geometry method. And then, the space coordinates were fed back by the method of the linear pre-interpolation. In the contrast experiment with traditional method, space coordinates were transformed to the cable lengths and the payload-swing angles. The experimental results show that the payload-swing angles transformed by this method are more accurate than that of traditional method, and the method can meet the real-time requirement. Moreover, the maximum length measurement error between this method and traditional method is 2.49%, which meets the accuracy requirement.

Focusing on the problem that PrePost algorithm needs to build complex Pre-order and Post-order Code tree (PPC-tree) and Node list (N-list), an improved frequent itemset mining algorithm based on the Interval list (I-list) was proposed. Firstly, data storage structure with more compression compared to Frequent Pattern tree (FP-tree), called Interval FP-tree (IFP-tree), was adopted, which mined frequent itemsets without iteratively establishing conditional tree. Secondly, the more concise method called I-list was used to replace the complex N-list in PrePost so as to improve mining speed. Finally, in the case of single branch path, some frequent itemsets were directly obtained by the method of combination. The experimental results prove the correctness of the proposed algorithm by getting the same results for the same dataset under same minimum supports, the proposed algorithm is superior to PrePost algorithm by about 10 percent in terms of time and space which has a good application in sparse database or intensive database.

Since Markov chain model based software testing does not consider the software structural information and has low ability of path coverage and fault detection, a new software testing model called software hierarchical testing model was proposed based on the combination of statistical testing and Markov chain model based testing. The software hierarchical testing model contains the interaction between software and external environment, and also describes the internal structural information of software. Besides, the algorithm for generating test data set was put forward:firstly, the test sequences conforming to the actual usage of software were generated; then the input data which covered software internal structure was generated for the test sequences. Finally, in the comparison experiments with software testing based on Markov chain, the new model satisfies the software testing sufficiency and improves the test data set's ability of path coverage and fault detection.

Concerning that the high performance requirement of face pose estimation system which could not run on mobile phone in real time, a real-time face pose estimation system was realized for Android mobile phone terminals. First of all, one positive face image and one face image with a certain offset angle were obtained by the camera for establishing a simple 3D face model by Structure from Motion (SfM) algorithm. Secondly, the system extracted corresponding feature points from the real-time face image to 3D face model. The 3D face pose parameters were got by POSIT (Pose from Orthography and Scaling with ITeration) algorithm. At last, the 3D face model was displayed on Android mobile terminals in real-time using OpenGL (Open Graphics Library). The experimental results showed that the speed of detecting and displaying the face pose was up to 20 frame/s in the real-time video, which is close to 3D face pose estimation algorithm based on the affine correspondance on computer terminals; and the speed of detecting a large number of image sequences reached 50 frame/s. The results indicate that the system can satisfy the performance requirement for Android mobile phone terminals and real-time requirement of detecting the face pose.

Concerning the problem in the conventional neural network expert system that can not provide explaining facility and reasoning process, the hot rolling products quality test system based on Radial Basis Function (RBF) neural network with function of qualitytrack can overcome the shortcoming. In the part of quality test, it provided detailed explanations and tracking process of the output of the expert system. According to the characteristics of steel industry, it used multiRBF neural network in the part of physical properties test. Firstly it used RBF neural network to forecast physical properties, and then dealt with the input parameters with the multiquadratic function. Finally it made longitudinal tracing and horizontal tracing to the result. The practical application result shows this system improves the degree of automation and the efficiency quality test in iron and steel enterprise. Compared with the previous way, it saves 60% of the time.

Since the functional form in consideration is of non-convex variational nature, the calculation results of the image segmentation model often fall into local minimum. Based on the global vector-valued image segmentation of active contour, the global vector-valued image segmentation and image denoising were integrated in a new variational form within the framework of global minimum. The new model was easy to construct and of less computation. Compared to the classical level set method, tedious repetition of the level set could be avoided. With the analyses on artificial images and real images, the new method is verified to have better segmentation results.

Concerning that the lesion areas of leukoaraiosis in Magnetic Resonance (MR) image present hyper intense signal on T ₂ flair sequence, a level set segmentation method based on C-V model was proposed. First, the C-V model was improved to avoid the re-initialization; second, the Otsu threshold method was used for image's pre-segmentation, and then the image's pre-segmentation result was directly used as the initial contour for the improved C-V model; finally, the segmentation result was obtained by curve evolution. The results show that the proposed segmentation method can get better separation effects, and realize fast auto-segmentation. It has certain application value for clinical diagnosis and prognosis on leukoaraiosis.

A method was proposed for estimating building heights via registering catadioptric omni-directional image and remote sensing image, which can be applied to large-scale 3D city reconstruction. Firstly, the top edges of building roof were extracted from the catadioptric omni-directional image by using omnidirectional Hough transform. Then the catadioptric omni-directional image and the remote sensing image were registered based on the extracted top edges where the angle consistency nature of horizontal lines in catadioptric omni-directional imaging was used as evidence. Finally, according to the model of catadioptric omnidirectional camera, the building heights were estimated by using the registration results. The proposed method is simple and easy to implement. The experimental results show that the method is effective and the error of estimated building height is fairly small.