Papers by Keyword: Feature Extraction

Abstract: This paper presents a new algorithm to retrieve 3D model on distance classification histogram. First, we select the certain number of random points on the model surface and compute the distance between two random points. Secondly, we sort the distance into two types which is based on the different geometry properties of these distance and construct the distance classification histogram. Finally, we measure the similarity of 3D models by comparing distance classification histogram. The experimental results on PSB show that our method has a good performance in precision and computational complication.

Abstract: In the field of printing research, color replication based on spectrum need to solve the issues that how to get cyan, magenta and yellow ink spectrum feature extraction. In this study, based on spectral derivative and spectral image deviation, establish a method to choose special wavelength positions for making up the corresponding relation between ink value and ink spectral reflectivity. The ink scales are printed on papers in the way that the value of the ink dot increase 5% from 0-100%. Ink spectrum reflectivity curves are made up through measuring ink scales. The first-order derivatives of cyan, magenta and yellow ink spectrum reflectivity curves in the range of 380-730nm, in sequence, change smooth at 650nm, 530-560nm, 450nm, and change violently at 530nm, 600nm, 500nm wavelength position. Based on spectral image deviation, the cyan ink spectral wavelength in the range of 630-730nm, the magenta ink spectral wavelength in the range of 500-540nm, include the most information and the least correlation. For the reason that yellow ink spectral images intersect with others, the analysis is combined with first-order derivative of ink spectrum curves. The method proposed in this paper is propitious to get primary ink spectral feature for color spectrum separation.

Abstract: Aiming at multi directions analysis problem of surface feature extraction from point cloud data, Curvelet transform is introduced to multi directions analysis of point cloud data. Based on the preprocessing of location and expansion, second-generation discrete Curvelet transform is used to analyze point cloud data. Curvelet transform coefficients are processed to enhance the contour of point cloud data. Nonlinear function is used to process Curvelet transform coefficients of coarse layer. Compromise for soft and hard thresholds is used to process Curvelet transform coefficients of detail layer. Piecewise nonlinear function is used to process Curvelet transform coefficients of fine layer. The data point is reconstructed from the enhanced Curvelet transform coefficient with Curvelet inverse transformation. Initial surface feature is achieved with edge detection. The precise surface feature is achieved with morphological dilation and erosion to filter edge without real shape significance. Example of part point cloud data of brake shell shows the proposed surface feature extraction method can accurately extract surface feature from data point.

Abstract: Automatic orange quality classification based on computer image processing is accurate and efficient. In this paper, we discuss the orange feature extraction and description method based on image processing. Design an orange image edge detection method based on Canny operator, color characteristics description methods based on HIS model and shape characteristics description methods based on Fourier descriptor operator. The experiment result proof that Canny operator is high SNR,high accuracy and low computation;HIS model is more accord with human vision and low computation also; shape characteristics description methods based on Fourier descriptor operator is more easy to shape classification.

Abstract: Smooth Support Vector Regression (SSVR) is new modified edition of traditional support vector regression for better performance. To further improve the modeling capability of SSVR, it is necessary to take into account the feature extraction based on Independent Component Analysis (ICA) before SSVR. Simulation on the example of function approximation shows that the result of SSVR based on ICA feature extraction is better than that of SSVR without ICA preprocess.

Abstract: In order to improve the diagnosis efficiency of faulty circuits with tolerance and to ameliorate characteristic description of faulty symptom, the faulty features of the circuit with tolerance are created by information entropy method at first, and then feature extraction method based on discrete Particle Swarm Optimization algorithm (PSO) is proposed to obtain the optimal feature subset of faulty feature. At last, the optimal feature subset is used to train the classifier to diagnosis the faults of the circuit with tolerance. Experiments results show the validity of this proposed method.

Abstract: Synthetic aperture radar (SAR) can obtain remote sensing data under all-weather and all-time, but the imaging principle is very complex and the right interpretation is more difficult. In this paper, using the characteristics of non-subsampled Contourlet transform (NSCT), including multi-scale, multi direction, anisotropy and shift invariant, the microscopic analysis and extraction of multi-scale features of SAR images is fully discussed. The purpose is to supply right interpretation for SAR image applications. The practical SAR image data is decomposed by NSCT and the decomposition coefficient features are extracted and discussed.

Abstract: In order to improve classification ability and diagnostic accuracy of centrifugal fan signals, a new feature extraction method from fault signals of centrifugal fan vibration based on manifold learning method (MLM) that is a kind of reduction method of data dimension is proposed in this paper.The MLM is able to remain nonlinear information of original signal, to improve the classification and diagnostic ability of fault better than traditional reducing dimension methods. The results in this paper show that, fault feature information of centrifugal fan vibration is extracted effectively by the MLM and the fault feature information of different types are separated effectively in themselves areas. The diagnostic accuracy by feature extracted by the MLM is significantly higher than by the wavelet packet analysis method.

Abstract: Ocean front is a narrow transitional zone that the penetration of sea is obviously different between two or more waters there. It is an important feature of geophysical turbulence which plays an important role in ocean dynamics. Ocean fronts become visible on radar images because they are associated with a variable surface current which modulates the sea surface roughness and thus the backscattered radar power. This paper propose a new integrated method to extract ocean fronts based on two-dimensional Empirical Mode Decomposition (EMD), image edge detection and mathematical morphology processing. Experimental results show that this integrated method can be effective in ocean front feature extraction.

Abstract: To investigate the brain default mode network (DMN) of healthy young people, a novel hierarchical clustering method was proposed to detect similarities of low-frequency fluctuations between any two out of 160 regions of interest (ROI) all over the brain. Feature of these ROIs were firstextractedand analyzed the feature using hierarchical clustering approach.Combining with the strongest connected network node identified by network centric criterion, the default mode network which presented the strongest connectivity in resting state was then determined. The results demonstrated that cingulate had the highest value of average degree, making it the most suspectof where the centrality indices of DMN lay.The comparative results between nodes included by DMN returned by our method and these given by Dosenbach’s research showed quite high coincidence rates,indicating the proposed method of combining complex network theory and hierarchical clustering analysis feasible method to parse brain regions.