复杂高层建筑结构图像三维重建技术研究

On Three-Dimensional Reconstruction Technology of Structural Image of Complex High-rise Buildings

目前利用几何造型对复杂高层建筑结构图像进行三维重建的方法,根据几何操作组建三维模型,并利用光照和光线跟踪绘制场景,构建漫游场景,然而该方法自动化的程度较低,开发的周期较长,且创造过程复杂.为此,提出一种基于动态规划的复杂高层建筑结构图像三维重建方法.该方法首先利用三维世界至二维图像之间的映射,以及相机坐标系至成像平面之间的映射完成建筑成像,并采用分形小波对图像的边缘进行提取,旨为通过分解系数实现图像边缘阈值的提取,然后根据平均误差分形法,对图像曲线进行分段,最后对图像的曲线轮廓分段所对应的特征点进行标记,通过相关系数NCC实现这些点的匹配,将标记点分为多对多匹配点对,并将曲线段势能当作度量的准则,根据动态规划法完成匹配点对之间的曲线段优化匹配.并通过实验证明,所提方法能够很好地提取出图像轮廓,增强了复杂高层建筑结构图像三维重建的逼真效果,减少了三维重建耗时.

At present,the method of three-dimensional reconstruction of complex high-rise building structure image by geometric modeling is used.According to geometric operation,three-dimensional model is built,and the scene is drawn by light and ray tracing.However,this method has low degree of automation,long development period and complex creation process.For this reason,a three-dimensional reconstruction method based on dynamic programming is proposed.Firstly,the method uses the mapping between the three-dimensional world and the twodimensional image,and the mapping between the camera coordinate system and the imaging plane to complete the building imaging,and uses fractal wavelet to extract the edge of the image.The purpose of this method is to extract the image edge threshold by decomposition coefficient,and then,according to the average error fractal method,the curve is segmented.At last,the feature points corresponding to the curve contour segments of the image are marked.The matching of these points is realized by the correlation coefficient NCC.The marker points are divided into multiple pairs of matching points,and the potential energy of the curve segments is regarded as the criterion of measurement.The curve segment matching between matching points is optimized according to the dynamic programming method.Experiments show that the proposed method can extract image contour well,enhance the realistic effect and reduce the time-consuming of three-dimensional reconstruction.