Catálogo de publicaciones - libros

When detecting ridge-valley lines on 3D mesh model, estimation of the curvature and curvature derivatives may often yields to squiggly and noisy result, because the estimation is sensitive against unwanted surface noises. We present two algorithms to obtain smooth and noiseless ridge-valley lines. First, we apply an iterative procedure on ridge and valley vertices and their previous and next neighbors on connected feature lines, which leads to smooth lines. Secondly, we propose an algorithm to distinguish noises from meaningful feature lines based on graph theory model. Each separate feature line is considered as an undirected weighted graph which is called as . We can reasonably get rid of most noises and preserve meaningful feature lines through optimizing the minimal spanning tree of each feature graph.

Developing wireless internet service and mobile devices, web access mechanism is various. There are various methods for developing wireless internet services, mobile devices, and web access mechanism. However, the existing web infrastructure and content are designed for desktop computers and not well-suited for other types of accesses, e.g, PDA or mobile phone that have less processing power and memory, small screens, limited input facilities, or network bandwidth etc. Thus, there is a growing need for transcoding techniques that provide ability to browse the web through mobile devices. In this paper, we present personalized XML documents transcoding techniques for mobile device users.

Visualization in scientific computing and engineering design is receiving wide attention. It can assist engineers, scientists, and technicians to access, analyze, manage, visualize, and present large and diverse quantities of data from the information gained from raw technical data. Clothing engineering design is a complex iterative-decision–making competitive process. Large amount of dynamic data are generated by simulation. The visualization application in clothing biomechanical design is presented is presented in this paper. The simulation results can be described by graph. The subdivision method is used to create the pressure distribute change animation.

The Robotic Space is the space where many intelligent sensing and tracking devices, such as computers and multi sensors, are distributed. According to the cooperation of many intelligent devices, the environment, it is very important that the system knows the location information to offer the useful services. In order to achieve these goals, we present a method for representing, tracking and human following by fusing distributed multiple vision systems in Robotic Space, with application to pedestrian tracking in a crowd. And the article presents the integration of color distributions into SOM based particle filtering. Particle filters provide a robust tracking framework under ambiguity conditions. We propose to track the moving objects by generating hypotheses not in the image plan but on the top-view reconstruction of the scene. Comparative results on real video sequences show the advantage of our method for multi-motion tracking. Simulations are carried out to evaluate the proposed performance. Also, the method is applied to the intelligent environment and its performance is verified by the experiments.

A robust motion tracking algorithm based on color and motion information was presented. Color is an effective feature in visual object tracking because of its robustness against rotation and scale variation. Nevertheless, the color of an object may change with varying illuminations, different image capture devices and different visual positions. Here, the color and motion information were fused in our visual tracking applications. Particle filter was employed as the essential framework because of its capacity of dealing with Non-linear/Non-Gaussian models by randomly sampling in state space. A particle filter can generate several hypotheses simultaneously in state space by randomly sampling and evaluate the states by weighing them respectively. The similarity between prediction data and observation information depends on the integration of Bhattacharyya distance and spacial Euclidean distance. Experimental results show the effectiveness of the proposed approach.

Motion Estimation (ME) is an important part of video encoding systems, since it can significantly affect the output quality of an encoded sequences. However, ME requires a significant part of the encoding time, because ME is a combination of techniques such as motion starting point, motion search pattern, etc. For this reason, low complexity motion estimation algorithms are viable solutions. In this paper, we propose a motion estimation algorithm to find the most accurate motion vectors(MVs) with the aim to maximize the encoding speed as well as the image quality. The proposed algorithm takes advantage of spatio-temporal correlations to decide the search pattern and the search start point adaptively and to avoid unnecessary motion vector search. Experiments show that the speedup improvement of the proposed algorithm over Motion Vector Field Adaptive Search Technique (MVFAST) and Predictive Motion Vector Fiekd Adaptive Search Technique (PMVFAST) can be up to 1.5 ~ 8 times faster while maintaining very similar image quality.

This paper presents a new method of counting the bi-directional passing people on crowded situations. It deals with an application of image sequence analysis. In particular, it addresses the problem of determining the number of people who get into and out of a surveillance zone when it’s crowded, and background and/or illumination changes. The proposed method analyzes image sequences and processes them using an algorithm based on the use of least squares and hausdorff distance. The method’s accurate degree will not be influenced by light, sunlight and shade of the passing people. Experimental results show that the new method is robust and more efficient than classic ones.

In this paper, a moving object detection method is proposed based on a level set algorithm of which speed function employs three properties based on human visual characteristics. The speed function is composed of three factors: directional filtered difference, proximity weighted spatial edgeness, and directional intensity consistency. For the directional filtered difference factor, directional filtering of the difference image between background and current images is introduced to utilize temporal edgeness along a detected contour. The edgeness in the current image is also employed for an initial estimation of moving object regions. The last factor, directional intensity consistency, is based on the continuity assumption of gray-level intensities along an estimated contour. The effectiveness of the proposed algorithm is shown with four real image sequences in terms of objective detection accuracies for various experimental conditions.

In this paper, we propose a new integrated robot vision system designed for multiple human tracking and silhouette extraction using an active stereo camera. The proposed system focuses on robustness to camera movement. Human detection is performed by camera egomotion compensation and disparity segmentation. A fast histogram based tracking algorithm is presented by using the mean shift principle. Color and disparity values are combined by the weighted kernel for the tracking feature. The human silhouette extraction is based on graph cut segmentation. A trimap is estimated in advance and this is effectively incorporated into the graph cut framework.

A vision-based real-time human detection and tracking capability is one of the key components of surveillance systems, human computer interfaces and monitoring systems. In this paper, we propose a method which uses color and disparity information obtained with a stereo camera. In order to achieve optimal performance with respect to detection or tracking of objects, it is better to consider multiple features together. We have developed a tracking method in which color and disparity information can be combined in a histogram. We used skin color and disparity distribution information to distinguish between different people. For human tracking, we propose a color histogram that is weighted by the disparity distribution of the target. The proposed method is simple and robust for moving camera environments and overcomes the drawbacks of conventional color histogram-based tracking methods. Experimental results show the robustness of the proposed method in environments with changing backgrounds and the tracking capabilities of targets which have similar color distributions as backgrounds or other targets. The proposed method can be used in real-time mobile robot applications.