Catálogo de publicaciones - libros

A segmentation algorithm in 3d medical data is proposed based on boundary model and local character structure in this paper. We found out inner voexls and outer voexls by pre-appointed voxel based on boundary model. And then, boundary voexls are correctly classified into different tissues by their eigenvalues of Hessian matrix based on the local character structure. Only eigenvalues of the boundary voxels are computed, so little time is used compared with other algorithms based on local character structure. It can quickly and effectively realize the segmentation of single tissue.

In the opening on the ripe aging society, in an effort to maintain their healths are improving through with their prevention. The internal organs of human body have good relationship with each other. Being the kidney keeps with best relationship in the internal organs, that the kidney function filtering the wastes to pick out into the urine on the processing to replace the old with the new blood. In this paper among the great four diagnosis, using the way of Ocular inspection & Auscultation, we would like to propose in the way of the kidney trouble shooting. To do these, through the assistance of the input image, extract the value of the color with optimization output, analizing the color of the face with related to the kidney.

The objective of this study was the application of driving comfort for the prediction of the driver’s posture. At the first, triangular type membership function was introduced to depict the driver’s feeling of comfort for the uncertain meaning of “comfort”, and then a model was built to assess the relation between design parameters and comfort assessment based on Fuzzy SVM. To get an appropriate driving posture, the comfort result was feed back to the design phase and some design parameter changing rules were built. At the last, a computer-aided driving posture prediction system was formed. This method can also be used to solve other relative problems.

Virtual colon unfolding is a method to produce an unfold image from volume data. The previous method generates only 2D images casting rays from a central path toward the colon surface and maps the color value into 2D regular grid. However, since rays cannot be reached behind folds, some regions are not represented on the final image. In order to represent those areas adequately, we exploit a height field derived from ray casting. Since the problematic areas have high gradient in comparison to neighboring regions, the differences of height value in those areas are relatively large. Therefore, we can find problematic areas using the height field. To visualize those areas, we exploit supersampling method that casts additional rays toward perpendicular direction of the original rays. Experimental results show that our method represents colon folds accurately.

With the popularization of digital museum, effective retrieval in huge image databases of special domain has attracted much research attention. As an effect approach, corresponding semantic information is adopted in many retrieval systems. However, most cultural relic retrieval systems only supply the explicit use of semantic information. It requires the user to be professional in this domain. In this paper, we propose a novel relevance feedback method which combines semantic annotation to visual feature implicitly. With this method, the user can unknowingly use professional semantic information to retrieve images. We also do some improvement in feature extraction and similarity measurement methods to fit the retrieval basis of unprofessional user better. The experiment results show that our approach is effective and efficient.

The problem of the existing constructing process for virtual biology model is that it is not analyzed whether the enacted initial condition of the judgment model is reasonable or satisfying with the natural biological rule. Based on this, we propose an idea to construct the virtual biology model using validity confirmation mechanism method. The novel mothod considers both the complexity of the system model, and the characteristics of alternation and cooperation between various essential factors. The feasibility of the new mechanism is confirmed in the biology teaching.

With the development of simulation technology, the numerical methods can help to reconstruct the traffic accidents. Based on the dynamic models of collisions, a Momentum/Energy method to predict the pre-impact motion was introduced through the optimization of the rest positions of the vehicles. With the help of Pc-Crash, the method of traffic accident reconstruction can be achieved, and several simulation runs will be performed, to find a solution, where post impact trajectories and rest positions correspond with the real accidents. In the paper, two kinds of collision accidents, the car crash to the wall and the car crash to another car, are analyzed based on the method. The simulation results show that applying the tire marks in the accident scene and the Momentum/Energy method can provide the scientific and numerical references for the analysis and judgment of the traffic accidents.

A seamless & adaptive visualization model of global DEM based on the QTM (Quaternary Triangular Mesh) is made to approximate to the Earth terrain in this paper. The approaches start with an Octahedron QTM partition based on the spherical surface. In order to improve the operation efficiency, the global DEM array points are organized as a hierarchy of s and the corresponding algorithms on indexing mechanism, coding scheme, and neighbor finding based on linear quadtree are given in details. Then, a dynamic operation method of the multi-resolution datasets in visualization is approached. Furthermore, an adaptive simplification rule of hierarchical triangles in data blocks is presented, in which an idea of the is introduced to form a continuous DEM mesh on the edges between different resolutions. In the end, the experiment and analysis are done with the GTOPO30 dada. The results is smoothly and receivable.

An approach is proposed in this paper for aligning the photographs, which are taken with varying amounts of exposure towards the same scene, and fusing into a high dynamic range image. Firstly, the median threshold images of the photograph-sequence are computed, and their image pyramids are constructed using multi-scale techniques. Comparing the image pyramids level-by-level completes shift-alignments of photographs. Based on the shift-alignments, rotation-alignments of photographs can be achieved. According to least-squares principle, the response function curve of the imaging process can be recovered; the mapping relationship between pixel values and exposures is acquired with the sample pixels from the aligned photograph-sequence. A satisfactory high dynamic range image can be constructed from the photographs. The experimental results show that the proposed approach is fast and efficacious.

Elastic Algorithm proposed in this paper is a new and fast path planning algorithm to solve the problem of auto-navigation in 3D virtual scene. First, with a grid model based representation of virtual environments, different obstacles are assigned different color in the projection map of scene and their encasing boxes are stored to provide local path. Then, Elastic Algorithm is designed to optimize the basic path. It can be applied to complex condition and moving object, and has high stability and efficiency in solving the path planning problem.