Applied Mechanics and Materials Vols. 333-335

Abstract: Palm vein pattern recognition is one of the newest biometric techniques researched today. This paper presents a palm vein recognition instrument that uses blood vessel patterns as a personal identifying factor. The instrument uses the recognition algorithm of two dimensional Fisher linear discriminant for classification. The experiment has been done in a self-build palm vein database. Experimental results show that the designed instrument achieves an acceptable level of performance.

Abstract: We present a 3D(three-dimensional)-modeling disparity-map optimization algorithm using a neural network and image segments for stereo navigation. We decompose the optimization algorithm problem into two sub-problems: initial stereo matching and depth optimization. A two-step procedure is proposed to solve the sub-problems sequentially. The first step is a region based NCC(normalized cross-correlation) matching process. But we use fast Fourier transformation and inverse fast Fourier transformation to eliminate redundant calculations in NCC, and we create a high-confidence disparity map by cross checking. In the second step, the reference image (the left image of the inputted stereo pair) is segmented into regions according to homogeneous color. A neural network is then built to model the three dimensional surface and applied to refine disparities in each image segment. The experimental results obtained for Middlebury test datasets and real stereo road images indicate that our method is competitive with the best stereo matching algorithms currently available. In particular, the approach has significantly improved performance for road images used in navigation and the disparity maps recovered by our algorithm are similar to ground truth data.

Abstract: Palm vein pattern recognition is one of the newest biometric techniques researched today. This paper proposes project the palm vein image matrix based on independent component analysis directly, then calculates the Euclidean distance of the projection matrix, seeks the nearest distance for classification. The experiment has been done in a self-build palm vein database. Experimental results show that the algorithm of independent component analysis is suitable for palm vein recognition and the recognition performance is practical.

Abstract: Based on the method of SVD, a novel watermark algorithm is designed in this paper. Firstly, the SVD is performed on each segment of the original host image; Secondly, the watermark bits are formed by chaotic encrypting;Thirdly, the secret bits are embedded into the S component of the matrix by quantization; Lastly, the inverse SVD transformation is performed on each block and the watermarked image is constructed. The original host image is not required in the secret watermark recovery. Experimental results show that the watermarked image has transparent feature, and the quality of the recovered watermark is satisfying. The algorithm is strongly robust to many attacks, such as JPEG compression attack, Noise Interference and so on..

Abstract: The ancient Tangka image with reflecting Tibet art has been broken by the nature and human with different degrees. The part-shed Tangka image is researched as a type of special breaking. After discussing color transfer algorithm and the existing problems, we improve the algorithm. The experiments show that the improved method can do well in inpainting the shedding and scratch of damaged TangKa image color.

Abstract: With human eyes has different insensitive to different types of texture, edged and dark area, this paper proposed an image steganography with higher embedding capacity and good stego image quality. In our method, the image blocks was divided into four kinds (texture, edged, dark and smooth), and secret message was embedded with the algorithm of modulus function coincide with wavelet transform. From the experimental results, the complexity notion can distinguish different kinds of blocks precisely, and the steganography method can hide much larger message and maintain a good visual quality of stego image.

Abstract: Vision-based measurement method can be widely used for a variety of real-time and online precision measurements, and particularly well suited for dynamic real-time precision measurement of geometry parameters of the part, which has advantages of non-contact, high-speed, big dynamic range, rich amount of information, and relatively low cost. After the study of vision-based online detection system of small gear, we propose a composite subpixel edge detection method, which combines the four-way weighted differential algorithm based on the classic Sobel operator and OFMM (Orthogonal Fourier-Mellin Moment), aiming at achieving the precision location of the subpixel edge firstly. And then detect tooth profile defects rapidly through scanning circularly the edge image, according to the structural characteristics of gears. The theoretical analysis and experimental results show that the detection method has so high accuracy and speed that it can meet the industrial online tests requirements.

Abstract: With the development of stereo vision, much more attention has paid from two-dimensional to three-dimensional (3-D) spaces, research on 3-D image/video becomes an inevitable trend presently. We present a novel research field that focused on the enhancement of 3-D videos, using two different 3-D videos and enhancing them with histogram equalization and edge sharpening algorithms. And we utilize the subjective assessment in the experiments. The experimental results show that the edge sharpening method has better effect than the histogram equalization method in 3-D video mode. But we also find some problems that both methods have blurred edges.

Abstract: The multi-sensor image fusion is the effective practices to increase the image information, highlight the detection superiority, reduce fuzzy understanding and to reduce data redundancy. Image fusion based on wavelet transform, the image wavelet decomposition processing only exists in the low-frequency, when the image contains high-frequency information, such as a large number of small edge or texture, which can not extract the feature information of the image, so resulting in the fusion is ineffective. In response to these problems, the use of image fusion algorithm based on wavelet packet transform, continue to break down, while the low-frequency further decomposition of the high-frequency of the image, extracts image feature information more effectively. In the same conditions of wavelet function, decomposition level, the fusion policy, comparative analysis has been researched on wavelet transform and wavelet packet transform on the same parameters of the information entropy, average gradient, standard deviation, spatial frequency, the results show that, image fusion of the algorithm based on wavelet packet transform are the highest and the better. In the other hand, in order to investigate the fusion effectiveness of the decomposition level on the same wavelet function conditions, fusion image parameters, such as entropy, average gradient, standard deviation, and spatial frequency, have been calculated using the db3 wavelet function corresponding to the decomposition level 1-5. The results show that the fusion effectiveness should achieve the best with wavelet decomposition level of 3 or 2.

Abstract: Human motion analysis in an intelligence surveillance system is a hot research topic in computer vision field.In this paper we proposed a motion recognition method based on silhouette information and centroid displacement for static environment.We used background subtraction method added background update and did binarization processing on foreground image by using adaptive threshold segmentation technique,then extracted moving target from the videos.Last we used silhouette information and centroid displacement as human motion features for behaviour analysis.Experiments show that the method has high accuracy rate at daily behaviour recognition.