Applied Mechanics and Materials Vols. 263-266

Abstract: Airdrome line recognition system can recognize the deformity areas of the airdrome lines, and send their information to the redraw system to redraw them. To gain the information of deformity areas of airdrome lines, according to the characters of airdrome lanes, we expressed a algorithms that use the number of white points every two lines as the eigenvalue of a graphic, after pretreated and divide the images, with the rules predefined the algorithms can recognized the location and size information of the deformity areas, which can be useful for the redraw system.

Abstract: Color recognition plays an important role during the process of bobbin sorting. A new combined method with BPNN and SVM is proposed to establish the color recognition model for bobbin sorting machine. The comparisons are done between the model based on new method and the models based on single BPNN and single SVM. The experiments demonstrate that the new color recognition model has the best learning and generalization abilities.

Abstract: According to the characteristics of infrared imaging and rare bird species, a infrared imaging inspection method for rare bird species based on nonlinear filter algorithm has been proposed. Firstly, the CCD(charge-coupled device) sensors are used to obtain infrared video-data for rare birds, and then the captured images of rare birds are preprocessed to obtain the body axis length. Secondly, nonlinear filter algorithm, combined with physiological characteristics of rare bird species, is used to detect and identify the high-speed flying birds. Finally, the capabilities of EKF(extended Kalman filter) and PF(particle filter), the two kinds of nonlinear filter algorithm in inspecting have been researched deeply. Theoretical analysis and experiments demonstrate that the performance of PF is much better than EKF to inspect high-speed flying non-rigid object in complex environment.

Abstract: Human body detection of intelligent system is an important and crucial issue, and this work has been studied for many years. Luciano Spinello achieved the real-time detection performance of human body based on Kinect with the help of GPU to accelerate to computation of features .But when its algorithm is realized in CPU, it can’t still achieve the real-time detection performance. This paper put forword a improvement measure to accelerate the computation of features. Features can be rapidly calculated by integral images, which was proposed by Qiang Zhu to detect object rapidly in 2001, abandoning the previous procession of using three line interpolation and Gauss filter, the improved algorithm, in CPU 3.10Ghz, RAM 2.85GB, 640*480 detection window, can achieve the average detection rate of 40 FPS. Performance gets promotion greatly.

Abstract: This paper comes up with an approach for surface reconstruction from arbitrary, unorganized and intersecting curves. First of all, camera calibration determines the interior and exterior camera parameters. Based on laser triangulation measurement, the point in world coordinate system can be computed. The extracted curves are processed in a volumetric in which a vector field is built and updated to compute the neighbor-hood points. Then the surface reconstruction can be obtained with MC algorithm. This approach makes surface reconstruction and curves registration a unified procedure. This algorithm solves the problem of linear complexity respect to input curves, and makes it suitable for interactive modeling. The experiment verifies its available, reduces the effect of registration errors and has a good result.

Abstract: Bag of words algorithm is to integrate visual words, described the local features of the object, together to form a bag model. It does not consider the structural information of the object, considering only the local features. This paper presents an image classification algorithm based on structural information. Algorithm adds the structural information of the object according to the ideas of generalized Hough transform. We use the structural information on the local features of the test image with generalized Hough transform and get the possible position of the center of the object. Then, we have to analyze the template size on the basis of the dispersion degree of the voting points, and finally optimize the voting results. Experimental results demonstrate that our proposed method is superior to the method which does not consider the structural information and uses the low-level features to classification.

Abstract: For the problem that appearance of human will be changed in motion, it is a challenge to those capabilities of tracker, related to adaptation of human appearance. In the paper, an improved Tracking-Learning-Detection algorithm, including short-term-tracking, real-time-detection and online-learning, is proposed. Furthermore, scale-invariant feature is applied to the short-term-tracking to determine the location of target in consecutive frames. The video stream of the PETS2006 database is utilized to evaluate the proposed algorithm. Experiments results demonstrate the effectiveness, especially on the short-term-tracking performance.

Abstract: Pedestrian detection from a video sequence is a challenging problem. In this paper, we mainly use active basis model which consists of a small number of Gabor wavelet elements at selected locations and orientations to detect pedestrian. In order to enhance the detection rate, we propose an adaptive background modeling method for background subtraction method to detect objects in video sequence which emphasize background and remove foreground using mask technology. After acquiring many pedestrian templates of different poses and orientations, active basis method can be applied in pose estimation and human action analysis. The detection results indicate that our approach is capable of obtaining better detecting effects and pose estimation even under conditions of noise and illumination changes.

Abstract: A recognition method of offline handwritten Chinese characters of amount in words is presented. The method uses elastic mesh strategy to divide character images written by special men into meshes, and extracts directional element and key point features in every mesh to produce a vector. Based on independent Hidden Markov Model classifiers, this paper uses voting rule to integrate the Hidden Markov Model classifiers. The experimental results show that this method has a relative high recognition rate.

Abstract: Snake model is one of classical active contour models, and is widely used in the image segmentation. But it has defects of sensitive to the initial contour and lack of curvature constraints. This paper presents an improved Snake model algorithm, which initializes contour by maker control watershed algorithm and adds an external force related with the curve shape into the energy function of Snake model to solve the problem about curvature constraints. Testing on the Magnetic Resonance (MR) images of the abdomen for liver image segmentation proves that the new algorithm can achieve good results.