模型驱动的快速三维场景构建  

Model Driven Rapid Scene Building

作　　者：谭小慧[1,2] 樊亚春[2] 王学松[2] 周明全[2] 

机构地区：[1]首都师范大学信息工程学院,北京100048 [2]北京师范大学信息科学与技术学院,北京100875

出　　处：《系统仿真学报》2013年第10期2397-2402,共6页Journal of System Simulation

基　　金：国家自然科学基金(61170203;61104035);国家科技支撑计划(2012BAH33F04)

摘　　要：提出了一种基于模型驱动的快速三维场景构建方法,利用基于图像、基于语义和基于草图三种方式检索三维模型数据并合成虚拟三维场景。首先,利用图割算法,语义分析,草图轮廓分割出待检索三维模型;其次,根据基于尺度不变特征点的轮廓形状局部特征描述方法,用量化后的视觉词汇向量描述三维模型轮廓及对象轮廓,从三维数据库中检索出与对象相似的三维模型;最后,通过对象模型和背景场景模型的坐标变换,合成渲染虚拟场景。实验证明,本方法用户操作简单,能够快速有效实现三维场景构建。The new 3D scene building approach was proposed which utilized the visual information of image scene, natural language and sketch to retrieval 3D models from database. The virtual scene was built with the retrieved 3D models. First, graph cutting method, semantics analysis or sketch drawing was used to segment objects and extract contour information; second, a new local shape description method was proposed to describe the contour of 3D model projections and 2D image by using quantized visual vocabulary vectors based on scale-invariant feature transform; finally, coordinate position was transferred to combine object models to build the visual scene. Experiments show that the proposed approach performs quite well and can effectively build 3D virtual scene rapidly.

关 键 词：三维场景构建 特征描述 模型驱动 视觉词汇 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]