Applied Mechanics and Materials Vols. 631-632

Abstract: Currently, video has become the most popular form of media, traditional video field is facing simulated to digital transformation. Using the computer to process the video information has a broad application prospect in many areas. This paper proposes an approach of real-time video acquisition and frame compression processing technology by using the FFmpeg and PHP function to achieve the effect of time-lapse photography. Finally, through the Apache log test the system running status.

Abstract: In this work, a kernel principle component analysis network (KPCANet) is proposed for classification of the facial expression in unconstrained images, which comprises only the very basic data processing components: cascaded kernel principal component analysis (KPCA), binary hashing, and block-wise histograms. In the proposed model, KPCA is employed to learn multistage filter banks. It is followed by simple binary hashing and block histograms for indexing and pooling. For comparison and better understanding, We have tested these basic networks extensively on many benchmark visual datasets ( such as the JAFFE [13] database, the CMU AMP face expression database, a part of the Extended Cohn-Kanade (CK+) database), The results demonstrate the potential of the KPCANet serving as a simple but highly competitive baseline for facial expression recognition.

Abstract: Over past several years, there has been a significant interest in employing terahertz (THz) imaging technology. Due to low contrast and bit noise in scanning THz pulse imaging, a novel fourth-order partial differential equation (PDE) denosing algorithm is proposed in this paper. The proposed approach is capable of overcoming THz pulse image block suppressing noise and preserving the sharp edges at the same time. A symmetric discrete algorithm based on four-neighbor system is developed to solve the new model. Median filtering is applied to alleviate the speckle effects in the processed image at last. Experimental results are provided to show that the proposed approach outperforms the image denoising approaches developed in the literature, which is useful for further image processing and recognition.

Abstract: This paper presents the design of a real-time high-definition image acquisition. The hardware platform combines Intel Xscale PXA270 processor, high-resolution camera and SAA7114H. The system is based on the embedded Linux system. Beetween the image sensor and the system memory,there is a quick capture interface.The interface receives the data from the image sensor,and converts the raw image data to a suitable format, and sends H.264 stream to the memory via the DMA channel. The result shows that the design can realize the real-time and high-definition image acquisition in a complicated environment. The advantage of this system is small volume, low power consumption and low cost. It can be widely used in agricultural and hydrological monitoring, intelligent transportation, security monitoring and intelligent home.

Abstract: the application of mobile learning is the main application of intelligent mobile terminal realizes on pattern, this paper focuses on the Android platform to realize mobile multimedia data real-time transmission, by optimizing the data transmission speed, to meet the learners' online learning, for subsequent teaching and interaction provides a good technical support.

Abstract: The method of view-dependent smoothly rendering of large-scale vector data based on the vector texture on virtual globe is presented. The vector texture is rasterized from the vector data based on view-dependent quadtree LOD. And the vector texture is projected on the top of the terrain. The smooth transition of multi-level texture is realized by adjusting the transparency of texture dynamically based on view range in two processes to avoid texture “popping”. In “IN” process, the texture’s alpha value increases when the view range goes up while In “OUT” process, the texture’s alpha value decreases. the vector texture buffer updating method is used to accelerate the texture fetching based on the least-recently-used algorithm. In the end, the real-time large-scale vector data rendering is implemented on virtual globe. The result shows that this method can real-time render large-scale vector data smoothly.

Abstract: A wealth of stream data is produced in the application of wireless sensor network (WSN). The knowledge in stream data can be extracted by data mining and it is useful for decision making. However, it is challenging to apply classical data mining methods on the scenario of WSN due to the factors such as limited power supply, on-line mining, data conversion and dynamic topology. This paper proposed a framework for distributed data mining by combining the existing approaches with the intrinsic property of WSN.

Abstract: This paper develops a pipeline leak detection software based on the vibration analysis method. It introduces the hardware designs, experimental methods and the based theory, experimental methods, and the leak detection algorithm. The system uses the generalized cross-correlation (GCC) algorithm to analyze those signals including vibration, temperature, flow and pressure. Then the system takes use of matlab to analyze those signals to choose the best weighted methods. At last, the software uses LabVIEW to generate the fronted user interface. Key words: Leak detection; Generalized Cross-correlation; Vibration analysis; LabVIEW

Abstract: For a correct judgment on the fault cause of transformers with the rich knowledge in various criteria and guideline, a method of automatic reasoning for the fault cause through fuzzy Petri net has been put forward in this paper. In this method, the knowledge in criteria and guidelines is firstly presented in the form of IF-THEN structure through the production rule, based on which the fuzzy Petri net model of fault cause is established then; and lastly possibility of each fault cause can be worked out through matrix iteration, which means the automatic reasoning is completed. Based on fuzzy Petri net imaging, this method makes the reasoning clearer and the result be got faster. The example calculation verifies that the method is correct and feasible in practical projects.

Abstract: The deformation monitoring data of the dam has the typical characters of instability and nonlinearity after being completed and impounding water. To solve the problems, this paper introduces the time series model and BP neural network model to analysis the dam monitoring data. Firstly, time series model was applied to fit and predict and then used the BP neural network model to correct the nonlinear part of residuals. Finally, we can get a series of fitting and predictive value of the monitoring data by combining of above both models. Taking the certain radial displacement value of a measuring point of a certain dam as an example, ARIMA-BP model was established to analyze the data. The result shows: fitting and predictive accuracy of ARIMA-BP model is relatively high and closed to the measured value.