Applied Mechanics and Materials Vols. 556-562

Abstract: By analyzing and studying the most current algorithms about mining association rule, the rules evaluated by minimum confidence could not ensure the validity of the rules and will generate unrelated rules which will affect the intrusion detection work. This paper proposes CF measure based on the previous work and applies the association rule algorithm based on CF to intrusion detection technology to detect the intrusion behaviors in the network. Finally, experiments show that improved algorithm is more efficient.

Abstract: IEEE1394 is a high performance serial bus, which physical layer directly to transmit and receive data with other nodes. when the ports of physical layer connect to other nodes, The connection which is not allowed may be activated. so there may be a loop formed in a bus topology ,which may influence the normal work of system. This paper analyzes the causes of the physical layer interface to form a loop, and introduces the method of port loop detection.

Abstract: As human-computer interaction gradually shifting to people-centered, head movement recognition technology in Augmented Reality application is becoming a hot spot of research. This paper presents an improved fast face detection algorithm based on skin color model, to distinguish the face region and the non-face regions, uses the feature localization algorithm based on color features to locate the facial features, and defines a series of simple head movements and associates them with operations.

Abstract: This paper analyzes the present situation of intelligent transportation industry in China and proposes a solution based on web map compatible interface which is aimed at satisfying the various requirements of the vehicle monitoring system on map. Firstly the technical feasibility of the solution is analyzed and 36 interface methods are listed. Then design principles and key techniques of vehicle monitoring and tracking, historical trajectory playback, regional function and monitoring are provided. Finally the development environment of the system and the screenshot of running results are presented. This system has already passed the platform compliance testing of the Department of Transportation and has achieved good results in management benefit in several enterprises.

Abstract: Introduce a complete system of license plate recognition: using morphological processing and priori knowledge of license plate to discern the location of license plate, accomplishing tilt correction through Radon transform, then fulfilling character segmentation of accurate positioning license plate by projection, finishing character recognition through BP neural network which was improved by the use of adaptive learning rate and momentum factor. With the programming and verification on Matlab experimental platform, experimental results show that we can have a preferable recognition speed and accuracy.

Abstract: In order to solve the problem of high time complexity of Linear Regression Classification algorithm,we propose a Mapreduce-based parallel linear regression classification algorithm. The map task uses the test image vector and the vector subspace to predict the response vector for one class, then calculates the distance measure between the predicted response vectory and the original response vector. The reduce task processes the data which are generated by the mappers, the test image is assigned to the nearest class. The experiments shows that the MapReduce-based parallel linear regression classifier can significantly improve the efficiency of Face Recognition.

Abstract: With the satellite development of our country, higher accuracy and stability are requires, which makes the control systems becoming more complex and requiring more telemetry parameters. Data mining techniques do not consider the physical relationship between the various components, but use of satellite telemetry parameters of the satellite states the purpose of fault identification. In this paper, we give a model based on multiple support vector machines (MM-SVM) technology satellite fault diagnosis method. The experiment shows that our method is effective in satellite equipment fault diagnosis

Abstract: An algorithm for multiple moving targets tracking in computer vision is put forward in this paper. The association between the tracking chains and new targets detected is mainly discussed. There are four situations. Among them, the most complicated is that several tracking chains exist and several new targets are detected. The distance between new targets and old targets is used to determine whether they are the same targets. When the tracking chain is found broken, it is seen that the target has passed the monitoring area, and begins to count. In the paper, the program code is shown, too. In the end of the paper, real images on the spot are used to test the algorithm. Its good effect can be proved from practical spot.

Abstract: Real-time multi-task embedded operating system is widely used in intelligent electronic devices. But the software design method with embedded OS is not familiar to developers in power system. So this paper depicts a sophisticated solution to overcome the difficulty in embedded software design. An overall software design is achieved in detail, and the emphasis is software architecture design, interrupt and task allocation and communication interface. After field validation, the measurement-control devices designed by this design solution have been proven that it is well adapted to the new requirements of Smart Grid.

Abstract: Compressed Sensing is likely to provide an effective way for lowering the extremely high sampling speed of UWB signal while the design of CS measurement matrix is of great significance for reducing the number of observations and hardware costs as long as improving the reconstruction accuracy. In this paper, with the combination of the structural features of the Fourier matrix and the idea of entry permutation of determined matrices, we propose a new measurement matrix of which the Fourier transformed entries are randomly permuted. Simulation results show that the same algorithm has a better reconstruction performance with the proposed measurement matrix rather than Gaussian/ Bernoulli matrix.