针对目前各种类型航海模拟器的虚拟建模效率较低与真实感较差等问题,设计与实现了一种基于多幅图像及少量用户交互恢复物体三维几何模型的系统.本系统采用稳定性好的改进Harris算子进行角点检测,基于角点空间特征及交叉相关实现角点的初始匹配;采用基于极线约束的角点匹配快速算法完成角点的精匹配,同时引入迭代和人工辅助修正的方法解决匹配算法的退化情况.在角点匹配结果基础上进行仿射重建,通过升级矩阵实现度量重建,在OpenGL环境下绘制出物体的三维模型框架.从原图像中抽取纹理并采用多视角图像融合的方法进行纹理映射,得到“照片级”真实感的三维模型.
Aiming at the problem of low efficiency and low photorealistic models in the current navigation simulator, this paper provides one system of image-based modeling which needs images of the object and the little user inputs. Firstly,we detect the corners using improved Harris operator. Secondly, we realize the initial matching points using feature space of corner and the cross-correlation. Thirdly, we adopt the fast arithmetic of corner matching based on epipolar line to get exact image, introduce iterative and got up means matching to avoid algorithmic degradation. Fourthly, we can finish a projective reconstruction and then an Euclidean reconstruction by upgrade matrix,then we can draw the 3D model of the object by OpenGL. Finally,In order to get photorealistic models,we should map the object by the texture drawn out from original images of the object using image fusion.