Advanced Materials Research Vols. 756-759

Abstract: The relevant properties of the selected classical threshold function and selected threshold were compared and analyzed for the methods of noisy-image wavelet threshold de-noising. On the basis of this, a new wavelet threshold de-noising function was given in this paper. The new defined threshold function overcomes the shortcoming which the hard threshold function and the soft threshold function on higher derivatives are discontinuous by adding a variable. Theoretical analysis and experimental results show that, the constructed wavelet threshold de-noising function possess the better adaptability and de-noising effect. In the case of strong Gauss noise for the image, relatively soft threshold method, the PSNR of a new threshold de-noising method can be improved from 4dB to 6dB.

Abstract: This article focuses on heat conduction problems in the process of steel industry. Modeling the direct problems of heat transfer, establish heat conduction and thermal radiation model. Model discretization method are used, discussion process from one dimension to two. We give the difference schemes, and the numerical example. Through the results we compare differences between one and two dimensional models, and the impact to the results of the two heat transfer mode.

Abstract: Compressive sensing (CS) technique can capture and represent compressible signal at a rate below the Nyquist rate of sampling. In this paper, the compressive sensing principles are studied and a new wavelet-based coding method is proposed. Two-dimentional discrete wavelet transform (DWT) is applied for sparse representation. Then the wavelet coefficients are divided into four blocks and are compressed by four different sensing matrixes respectively. The recovery quality depends on the number of received CS measurements. Experimental results show that the proposed coding method achieving higher performance compared with the conventional wavelet-based compressive sensing coding system. With 50% coefficients retained, the image can still be recovered with considerably objective and subjective quality.

Abstract: In order to meet the user requirement of the dynamic customization and configuration of the changeable and complicated page functionality on the client, when constructing the web application platform, an advanced rich-client method and technology based on DOM ( Document Object Model ) was designed and used to develop the client module. The client module with rich-client technology was in the traditional J2EE (Java 2 Enterprise Edition ) architecture which was the Client-Centric and MVC ( Model-View-Control ) mode. On the client side, according to the dynamic page generation algorithm, developers wrote JavaScript scripting language based on DOM and Ajax (Asynchronous JavaScript and XML) for user customization and choose the part of the third-party open-source Extjs ( Extendable JavaScript ) components as the page elements to generate client-side dynamic configuration interface. From a user experience perspective, The good performance test results of the advanced rich-client method effectively examine the distinguishing features of the new method.

Abstract: Traditional scientific research project cost estimating method cannot meet accuracy and practicability at the same time. Aiming at this problem, scientific research project cost estimating method based on neural network was built. Firstly, the construction and influencing factors of scientific research project cost were analyzed. Secondly, an estimating model based on improved BP neural network was built; a nonlinear expression between influencing factors (input) and cost (output) was created. Finally, an estimating system with the model was implemented by Java. The effectiveness of the method was tested. Testing experiment showed the estimating model based on improved BP neural network is reliable and the precision is high.

Abstract: MapReduce is a widely adopted parallel programming model. The standard MapReduce model is designed for data-intensive processing. However, some machine learning algorithms are computation-intensive and time-consuming tasks which process the same data set repeatedly. In this paper, we proposed an improved MapReduce model for computation-intensive algorithms. The model is constructed from a service combination perspective. In the model, the whole task is divided into lots of subtasks taking account into the algorithms parameters, and the datagram with acknowledgement mechanism is used as the communication channel among cluster workers. We took the multifractal detrended fluctuation analysis algorithm as an example to demonstrate the model.

Abstract: In order to transmit the image which has captured by the distal thermal imager rapidly and real-time to the PC for processing, but due to the unreliability of the UDP protocol, It may occur data loss and "chaotic" and so on when each frame of data is transmitted after segmentation. This article uses segmentation transmission, packet head structure, buffer, timer, and the improved sliding window which not only ensures the UDP protocol transmission timeliness in LAN, but also ensures the transmission of reliability of thermal image with the UDP protocol in LAN.

Abstract: The risk evaluation of empty mine mined-out area is of great significance to the security and stability of the power facilities. But influence evaluation of empty mine mined-out area factor multitudinous, this paper selected seven factors associated with the system. Based on the principle of BP neural network, we build a 3 layer BP neural network model suitable for grid risk evaluation. The BP neural network model was trained by collected samples of empty mine mined-out area and the logical parameters of BP neural network were acquired and tested by the testing samples for accuracy, and finally we proposes preventive measures based on the evaluation.

Abstract: Nowadays, it is particularly important to improve the level of backup devices management for enterprises, because a reasonable management mechanism would ensure a business operation normally. The paper presents a backup device management system in detail, including requirements analysis, architectural design, function modules and implementation mechanism, which provides basis for systems implementation and application. The system we designed can manage backup devices dynamically, which gives support to the improvement and maintenance, and also provides more convenience for users.

Abstract: The principle and structure of vessel light signaling is introduced. When LED light signaling working,the light signal is aimed by zoom camera and the signaling image is obtained by high-speed camera. The light signaling on the overall image is discriminated and disposed at real time. At last, the character is translated and shown on computer. Experimental result shows that the system is stable, safe and reliable.