Advanced Materials Research Vols. 889-890

Abstract: In the recent years, the feature extraction algorithms based on manifold learning, which attempt to project the original data into a lower dimensional feature space by preserving the local neighborhood structure, have drawn much attention. Among them, the Marginal Fisher Analysis (MFA) achieved high performance for face recognition. However, MFA suffers from the small sample size problems and is still a linear technique. This paper develops a new nonlinear feature extraction algorithm, called Kernel Null Space Marginal Fisher Analysis (KNSMFA). KNSMFA based on a new optimization criterion is presented, which means that all the discriminant vectors can be calculated in the null space of the within-class scatter. KNSMFA not only exploits the nonlinear features but also overcomes the small sample size problems. Experimental results on ORL database indicate that the proposed method achieves higher recognition rate than the MFA method and some existing kernel feature extraction algorithms.

Abstract: Edge detection is the basic problem in the field of image processing. Various image edge detection techniques are introduced. Using various edge detection techniques different images are analyzed and compared by MATLAB7.0. In order to evaluate the effect of edge segmentation, the root mean square error is used. The experimental results show that no an edge detection technique works well for all types of images.

Abstract: Hyperspectral images have been widely used in earth observation. However, there are some problems such as huge amount of data and high correlation between bands. An application of particle swarm optimization algorithm based on B distance was proposed to band selection of hyperspectral images. First of all, bands are grouping by the correlation coefficient of the band and adjacent bands. B distance was used as separability criterion between classes and the fitness function comes into being. Finally, the classification results illustrate that the total classification accuracy of the proposed method is higher than the traditional method.

Abstract: The BP neural network is a classifier commonly used in partial discharge type recognition, but the traditional BP algorithm with defects cannot satisfy the actual need. So the optimization algorithm of BP network was studied intensively. DPSO algorithm was used for optimizing the network, and DPSO-BP algorithm is applied to analyze typical defects of GIS, which can be identified by the types of partial discharge. Compared with traditional BP algorithm, DPSO-BP algorithm occupied obvious advantage in recognition effect. It has improved the learning speed of the algorithm, effectively avoid network training going into local minimum point, and maintain the generalization ability and fault tolerance of BP neural network at the same time.

Abstract: A triangulation Algorithm of Depth Images based on Self-adaptive Quadrangle mesh is proposed in this paper. Depth images will be triangulated in two steps. Firstly, self-adaptive quadrilateral mesh is generated based on depth images. Row and column pixels of quadrilateral mesh can be obtained using uniform sampling method. Quadrilateral mesh is generated by connecting adjacent row and column pixels. Secondly, each quadrangle consists of two diagonals and triangulation is generated by selecting diagonals with larger directional angle difference.

Abstract: A novel model which is about the image denoising and enhancement is proposed in this article, the image denoising and enhancement increasingly becomes a bottleneck restricting the follow-up image of a series of processing On the basis of anisotropic diffusion model, an edge stopping function is introduced, which can make up the drawback that solely relies on detecting the gradient information to control the diffusion process .Using the edge stopping function position accurately on the edge so as to achieve the purpose of the noise reduction fully in the non-edge zone, but it inevitably will blur the edge information. Therefore, the further use of the shock filter in the edge enhancement is essential. Experiments show that the model can well remove the image noise and achieve good visual effect.

Abstract: In the TV goniometer detection system, to play the signal and field of view points line extraction is a key link in the process of parameter detection. Combination of target processing requirements, this article will target extraction algorithm based on gray level threshold and edge detection algorithm is studied, and through the experimental analysis to select the optimal algorithm was applied to the detection of TV goniometer; According to the characteristics of the standard signal and view points, lines, and put forward the corresponding methods of target recognition, and is verified through experiments

Abstract: In this paper a improved algorithm of median filtering based on extremum detection is introduced. According to the deficiencies of the extreme median filtering method in the two stages of noise detection and noise filtering, Adds a false detection noise pixel gray value correction recovery process. Experimental results show that the algorithm in this paper are superior than traditional median filtering algorithm and some improved image filtering algorithm in noise removal and edge retention.

Abstract: Image Fusion is an important and useful subject in Image Processing and Computer Vision. The traditional image fusion algorithm could not provide satisfactory fusion results. Aiming to solving this problem, in this paper, we proposed an algorithm based on shearlet and multi-decision. First we discussed the application of the shearlet transform. Then we use difference decision rules for image decomposition high-frequency coefficients. Finally, the fused image is obtained through inverse Shearlet transform. Experimental results show that comparing with traditional image fusion algorithms, the proposed approach can provide more satisfactory fusion outcome.

Abstract: Thread features of the traditional measuring method mainly adopts working gauge measurement, due to limitations in the traditional thread features measurement accuracy is relatively low, the efficiency is low, the cost is high. The thread features detection method based on digital image processing techniques using CCD to obtain basic image of thread, processing the thread image, extracting thread outline, calculating thread features through the computer, improves the efficiency, saves the cost.