Applied Mechanics and Materials Vols. 333-335

Abstract: The main challenge of Topic Detection and Tracking (TDT) for Blog is the insufficient information in a topic description and the lack of key words input by users. We propose a Two-layer KL Distance approach which combines the KL distance model with a lexical semantic association matrix model. First, the KL Distance model captured the weights of Initial feature words. Second, the KL Distance model was used again to estimate weights of words linked with initial feature words in the lexical Semantic Association Matrix. Extensive experiments show the advantages of our method over the baselines as well as the effectiveness of the two-layer of KL Distance.

Abstract: With the Web 2.0 development, microblog has become the current hot spots.To HowNet Sentiment dictionary-based, and the current dictionary of commonly used network terminology adds microblog emotional, Sina API access to data, to build the microblog sentiment dictionary, for the Chinese microblog emotional information analysis, the experimental results show that the analytical methods microblog text emotional tendency analysis with good results.

Abstract: Nowadays, mobile WEB video application got rapid development, and the video service quality evaluation and the analysis of the network performance have become an urgent demand. In this paper, we propose a new framework and method to assess the subjective quality of mobile web video based on no-reference (NR) model. Through discuss the structure of TD mobile network and the application of web video, the paper built a set of testing process for mobile video on demand service and fault location method based on the result of network structure and the business of video quality evaluation. The actual test results from TD network shows the proposed way can measure the web video quality and find the bottleneck location of the transmitting network. The evaluation result of video on demand quality is in accordance with the subjective feeling of the user. The benchmark contrast testing method proved to be practical and feasible in the real test, thus providing readily available tools for video service quality assessment.

Abstract: With the development of economy, there are an increasing number of cars as well as traffic accidents, thus intensifying the need to take measures to reduce traffic accidents and protect the safety of life and property. Vehicle distance is one of the most important indexes of traffic safety. The measurement of safety vehicle distance has become an increasingly hot research area of intelligent transportation. Through analyzing the basic principle of stereo vision and calibrating the parameters of the CCD sensors both inside and outside, this paper comes up with a method to measure the former vehicle distance based on stereo vision and DSP. Once the vehicle speed and distance form a non-security association, it will give a warning, and upload data or force speed-limiting. According to the different coordinates of the obtained images of the target vehicle from the left and the right sensor, this method can identify feature points, calculate distance to the target vehicle, and analyze the security of vehicle distance. The experimental results show that this method has wide measurement range, high measurement accuracy, and fast operation rate, thus it can meet the actual needs of the measurement of safe vehicle distance in intelligent transportation.

Abstract: Matching Algorithms are explored by many experts, lots of good methods are used for photography, computer vision, 3D reconstruction etc. Such as SIFT and some other improved algorithms that are robust to translation, rotation, illumination, angle of view, and scale different. However, it is a bottleneck for their application in some in time system because of their high time cost. A parallel framework for automatic matching for high resolution imagery with large size is introduced, the most time-consuming parts while matching are devised by parallel in a server with many cores, for example, the feature extracting, matching, and the quick quasi-dense matching on the epipolar resampled images. Experiments show that, it works efficiently and robustly.

Abstract: To adapt the contradiction between the increasing information quantity of highway traffic network monitoring and the limited network bandwidth resources, this paper proposes an object detection algorithm based on Bayesian compressed sensing. Video are sparse in a wavelet base, and a partial Hadamard measurement matrix is adopted to compress the video. The object detection method combines background difference and Bayesian compressed sensing of wavelet tree structure. To get more accurate foreground, an adaptive background model is proposed. Experiments results show the accuracy and effectiveness of the method, and can robustly detect the targets under changing light and reduce the price of video transmission.

Abstract: It is known that wavelet transform provides very useful feature values in analyzing various types of images. This paper presents a novel approach for content-based textile image retrieval which uses composite feature vectors of low-level color feature from spatial domain and second-order statistic features from wavelet-transformed sub-band coefficients. Even though color histogram itself is efficient and most used signature for CBIR, it is unable to carry local spatial information of pixel and generate inaccurate retrieval results especially in large image data set. In this paper, we extract texture features such as contrast, homogeneity, ASM(angular-second momentum) and entropy from decomposed sub-band images by wavelet transform and utilize these multiple feature vector to retrieve textile images combining with color histogram. From the experimental results it is proven that the proposed approach is efficiently retrieve the desired images from a large set of textile image database.

Abstract: Road extraction is the recurring important application of high-resolution remote sensing images. In order to achieve the goal of road extraction, the various characteristics of geographic information of high-resolution remote sensing images as well as the application and models of road extraction are analyzed, then an effective way of extracting roads from high-resolution remote sensing images is found, and then the high-resolution remote sensing image road extraction algorithm based on texture characteristics assisted by other characteristic information is put forward. The specific process of road extraction in the algorithm is introduced, and the function of road extraction of urban high-resolution remote sensing image based on texture characteristics is also tested practically, the result shows that this method has a higher degree of accuracy in extracting roads from urban high-resolution remote sensing images.

Abstract: Proposed An images denoising method by using variable coefficient in filtering. The key point is using different coefficient to minimize the influence of each gray values of the pixels as is regarded as black noise and white noise and choosing a threshold to decided which filter will be used as the most suitable one. Usually,a image is polluted by Gaussian noise and impulse noise simultaneously, It cannot be obtained if only using one of them,though both median and average filter are used inimage denoising in obtain the edge as well as detail characteristics of impulse noise image and smoothing Gaussiannoise respectively when denoising in spatial space. Experimental results show that the variable coefficient method can improve the quality of images than the classical methods.

Abstract: To obtain image enhancement by getting detail texture features rich through the subtraction between the original graph and low-frequency graph, utilizing the wavelet decomposition technique, separating the components of high frequency and low frequency in the 2D (two-dimensional) aerial image, filtering the low frequency subgraph, and retaining only the low-frequency part. Practice has proved that the unsharp masking algorithm of wavelet transforming has obvious superiority over the traditional algorithm. The algorithm can separate characteristics of different resolution details, adjust the filter window size, make the wavelet components of different scales enhanced, the detail more clear, get a strong sense of layer and good enhancement effect.