Applied Mechanics and Materials Vols. 380-384

Abstract: The circuit in this paper is an effective solution to interference and noise of input signal during data acquisition and the circuit designed is based on an active low pass circuit with output frequency from 0~100Hz. The low pass circuit consisted mainly of an OPAMP TL084. Chaos theory is applied in this active nonlinear filter.The designed circuit is analysed by Matlab simulation software and the simulation results show the method can effectively improve effects of the filtering circuit.

Abstract: SELinux (Security Enhanced Linux) inherited the basic design of LINUX, and it is a high secure operating system. It is important to know how to make the IPC (Inter-Process Communication) for this kind of multi-task and multi-user system. In this paper, Finite State Automaton is used to verify the security of IPC mechanisms of SELinux. Finally, IPC mechanisms are verified with SPIN model checker.

Abstract: This paper proposes an ICA image feature extraction method without pre-whitening based on wavelet denoising. Firstly, we conduct denoising for noisy observed images with wavelet transformation; then the feature extraction for denoised observed images is done by using FastICA algorithm without pre-whitening; finally, we further remove the residual noise in extracted feature images with wavelet transformation. Simulation experiment results show that this method is apparent in the performance of denoising, meanwhile, the extracted feature images can distinguish the texture and shape feature well, which has stronger practicability and validity.

Abstract: Based on the analysis of influence factors for hull deformation data, physical models including hull curve deformation, hull torsion deformation and hull athletics deformation were established according to mechanical principle. Considering characteristics of the space instrumentation ship, systems errors identification method based on total variation constrained was provided. First order differential coefficient of deformation data was joined in a constrained equation in the method. Systems errors coefficient and exact value of deformation data were solved through alternating iterations. The simulation result of deformation measured data indicated that identification errors were in consistent with plus errors, and systems errors coefficient was also estimated accurately. This demonstrated that the proposed method is suitable for systems errors identification of hull deformation data.

Abstract: In order to improve recognition rate of human ear, a method based on point feature of image for ear recognition is proposed in this paper. Firstly force field transformation theory is applied to human ear image two times in our method. It can extract the structural feature points and contour feature points of ear respectively and compose feature point set. Then feature points described by the scale invariant feature transformation descriptor. At last nearest neighbor classifier is employed for ear recognition. Feature points extracted from ear image using force field transformation are stable, reliable and discriminative, and they are insensitive to variations in image resolution. Constructing descriptor can resolve the problems caused by lower recognition owing to illumination change, scaling transformation, rotation and minute alteration caused by pose transformation. The experimental results show that the proposed algorithm not only can effectively improve ear recognition rate but also has quite good robustness.

Abstract: A novel method for image recognition using Pulsed Coupled Neural Network (PCNN) based on Maximum Average Correlation Height (MACH) filtering is proposed. Firstly based on the property of synchronous pulses oscillation for the similar group of neurons in PCNN, segment image to extract edge information, and this effectively suppresses noise. Subsequently the synthesized MACH is adopted to excellently realize the distorted target recognition using appropriate filter parameters depending on different targets. The simulations results show that the output correlation peak is obvious, and validate the effectiveness and accuracy of the method for distortion target recognition.

Abstract: KPCA extracting principal component with nonlinear method is an improved PCA. The KPCA can extract the feature set which is more suitable in categorization than the conventional PCA. The method of KFDA is equivalent to KPCA plus LDA. KPCA is first performed and then LDA is used for a second feature extraction in the KPCA-transformed space. The KPCA and KFDA have been got widely used in feature extraction and face recognition. In this paper, the method of KPCA and KFDA is analyzed and their nature is revealed. Finally, the effectiveness of the algorithm is verified using the ORL database.

Abstract: According to the defects of classical k-means clustering algorithm such as sensitive to the initial clustering center selection, the poor global search ability, falling into the local optimal solution. A differential evolution algorithm which was a kind of a heuristic global optimization algorithm based on population was introduced in this article, then put forward an improved differential evolution algorithm combined with k-means clustering algorithm at the same time. The experiments showed that the method has solved initial centers optimization problem of k-means clustering algorithm well, had a better searching ability,and more effectively improved clustering quality and convergence speed

Abstract: This paper presents a method of detecting pedestrians side in video frames of cluttered scenes. This detection technique is based on the idea of wavelet template and SOM neutral network. In order to make detection results more accurate and reduce computation cost, we combine background subtraction and frames difference to decide where pedestrians stand in a frame.

Abstract: Considering the fact that original histogram of oriented gradients (HOG) cannot extract the body local features in large image regions, its features are improved when extracted, then more gradient information are extracted and feature description operators can be obtained which describe human detail features better in lager image regions or detection windows. Considering speed, we select support vector machine (SVM) using linear function kernel as a classifier. Combining with HOG extraction and SVM training, the process includes three steps: features extraction, training and detection. Experiments show that while maintaining a relatively satisfactory speed the human detection system improves detection accuracy.