Search

Because the optical images obtained in geological hazard monitoring area can't display the Gaussian coordinates visually and the Synthetic Aperture Radar (SAR) images aren't easy to be identified. The Digital Elevation Model (DEM) data and Global Navigation Satellite System (GNSS) monitoring data are applied in this paper. Firstly, interpolation is used in SAR image, and the missing height Gaussian coordinates are obtained by matching with DEM data. Secondly, the cross-range and height Gaussian coordinates of each target point on the optical image are obtained by applying the pixel scale, and the range Gaussian coordinates are got by matching with DEM data. Finally, the fusion information is obtained by matching the SAR image with the optical image. The experimental results show that this method has certain feasibility. It proves that the method can provide simple and effective help for landslide hazard monitoring.

Since the building room consists mainly of three brick walls and one glass wall, Unmanned Aerial Vehicle (UAV) equipped with optical sensor and radar sensor detects targets only from a specific view during monitoring the building fire explosion site. The unknown height information of combustible objects in the two dimensional (2-D) radar image and the unknown range information of combustible objects in the 2-D optical image makes it difficult to get comprehensive information. By fusion of the range and azimuth information extracted by the 2-D radar image and the height and azimuth information extracted by the 2-D optical image, in this paper, a new algorithm is proposed to reconstruct a three dimensional (3-D) image with the correct range, azimuth and height information of combustible objects. The simulation results validate the correctness and feasibility of the algorithm.

The effects of pressure on the structural and elastic properties of orthorhombic TiAl are investigated using first-principles calculations based on density functional theory within the projector augmented wave method. The calculated lattice parameters at 0 GPa are in good agreement with the available experimental data. The pressure dependence of the normalized lattice parameters and the single crystal elastic constants are investigated. By the elastic stability criteria under pressure, it is found that orthorhombic TiAl is mechanically stable under pressure up to 100 GPa. The elastic moduli and Poisson’s ratio under pressure up to 100 GPa are calculated using the Hill average method. The ductility/brittleness under pressure are evaluated, and a critical pressure for brittle-to-ductile transition is found to be 40 GPa. The elastic anisotropy and Debye temperature under different pressure are estimated from the calculations.

A novel face recognition method using wavelet transform and Bidirectional Principal Component Analysis (BDPCA) was presented. In the proposed method, the logarithm transform and wavelet transform were calculated for face preprocessing. BDPCA algorithm was used for face feature extraction. Finally, the nearest neighborhood classifier using Cosine distance was adopted for feature classification. The experimental results on Yale B frontal face database show that the face recognition rate of the proposed approach can attain 100% when wavelet type and wavelet decomposing levels were selected properly, and the proposed algorithm can alleviate face uneven illumination efficiently.

A novel illumination compensation algorithm for face recognition under variable illumination conditions is presented. The method calculates 2-dimensional multiwavelets transform for face illumination compensation in logarithm domain, and then conducts the face recognition experiment directly. The proposed approach is compared with other method in logarithm domain on Yale B face database. The experimental results show that the average error rate of the proposed approach is only 0.18%, which is superior to the most of existing face illumination compensation methods.

In this paper, a new method to estimate the intrinsic dimensionality of high dimensional dataset is proposed. Based on neighborhood graph, our method calculates the non-negative weight coefficients from its neighbors for each data point and the numbers of those dominant positive weights in reconstructing coefficients are regarded as a faithful guide to the intrinsic dimensionality of dataset. The proposed method requires no parametric assumption on data distribution and is easy to implement in the general framework of manifold learning. Experimental results on several synthesized datasets and real datasets have shown the facility of our method.

It is generally thought that semantic and grammatical information was very significant to better understanding and processing of text. But in simple text categorization task, absence of this information does not always lead to the degradation of classifier performance. In this paper, we discuss the application of the character-level statistical method in text categorization, which extract character-level frequent pattern rather than consider the semantic and grammatical information. Compared with traditional n-gram model, the presented method is easy and convenient. Then by casting character-level statistical method in Bayesian theory framework, the proposed method was applied to spam detection. At last, we discuss the multiclass problem in short message categorization based on combination strategies. Effectiveness of the models and feasibility of the present method are verified.

Geodesic distance calculation is a crucial stage in the distance-based manifold learning algorithms. Owning to the nonparametric of Isomap, however, it is not economic and convenient to compute the geodesic distances for an unseen new sample via neighborhood graph. In this paper, we proposed a Geodesic Metric Learning (GML) scheme based on Iterative Majorization, to obtain a parametric geodesic metric function. Once the parametric geodesic metric is produced, one can calculate the geodesic distance directly and have no longer use for reconstructing the neighborhood graph for those unseen samples. Integrating distance-based triangulation with the parametric geodesic metric, it is easy to accomplish the task of feature extraction based on geodesic distance. Experiments on benchmark database have demonstrated the encouraged results.

In this paper, we present an effective scheme combining speech enhancement with feature post-processing to improve the robustness of speech recognition systems. At front-end, minimum mean square error log-spectral amplitude (MMSE-LSA) speech enhancement is adopted to suppress noise from noisy speech. Nevertheless this enhancement is not perfect and the enhanced speech retains signal distortion and residual noise which will affect the performance of recognition systems. Thus, at back-end, the MVA feature postprocessing is used to deal with the remaining mismatch between enhanced speech and clean speech. We have evaluated recognition performance under noisy environments using NOISEX-92 database and recorded speech signals in continuous speech recognition task. Experimental results show that our approach exhibits considerable improvements in the degraded environment.

Spam has been a serious problem to global email users. In this paper, a two-layered spam detection flow was used, which showed the trade-off between accuracy and efficiency. Then we discussed Naive Bayes classifiers ensemble based on Bagging. By casting spam detection in a decision theoretic framework, a Naive Bayes Bagging spam detection model based on embedded decision tree is proposed. Then this model was reduced by strict likelihood score bound limitation of the Naive Bayes classifiers. Finally, an improved method based on classifier error weighted is presented. The experiment results show that the modification is effective.

Recently several manifold learning algorithms have been presented for nonlinear dimensionality reduction. Isomap is one of them. However, Isomap suffers from a deficiency that it does not give an explicit mapping function, which is from high dimensional space to low dimensional target space. In this paper, a version of Isomap with explicit mapping, called E-Isomap, is proposed. In E-Isomap, the geodesic distance matrix is fed into a cost function and then iterative majorization is adopted to solve an optimization problem for obtaining both the low dimensional configuration and the nonlinear mapping. Owing to the existence of explicit mapping, this version of Isomap can be more easily used in pattern recognition than the original ones. The experiments on two benchmark data sets are given to demonstrate the performance of the presented method

A novel method for face gender discrimination was proposed. The method got 27 intrinsic mode functions (IMFs) by calculating empirical mode decomposition (EMD) for 3 mean faces. For face gender feature extraction, the method used these IMFs as projection vectors. Finally, kernel Fisher discriminant analysis (KFDA) and support vector machine (SVM) were used for classification, respectively. With the same performance for face gender discrimination, computational results show that the efficiency of EMD+KFDA method is more than 3.7 times as that of direct KFDA method, and the EMD+ SVM method is at least 1.5 times faster than the PCA+ SVM method.

Recently, SVMs are wildly applied to regression estimation, but the existing algorithms leave the choice of the kernel type and kernel parameters to the user. This is the main reason for regression performance degradation, especially for the complicated data even the nonlinear and non-stationary data. By introducing the ‘empirical mode decomposition (EMD)’ method, with which any complicated data set can be decomposed into a finite and often small number of ‘intrinsic mode functions’ (IMFs) based on the local characteristic time scale of the data, this paper propose an important extension to the SVM method: multi-scale support vector machine based on EMD, in which several kernels of different scales can be used simultaneously to approximate the target function in different scales. Experiment results demonstrate the effectiveness of the proposed method.