二维动画人物的特征提取与三维重建

Feature Extraction and 3D Reconstruction of Two Dimensional Animated Characters

为了解决快速提取二维动画人物特征,完成立体动画人物形象构建的问题,提出了二维动画人物的特征提取与三维重建方法.通过分析二维动画人物特征提取,具体分析了二维动画人物Harris算子的确定、二维动画人物对极几何关系的确定、基于Harris算子和对极几何角度的特征提纯;分析了基于二维动画人物特征的三维重建方法,主要包括基于二维动画人物特征的三维重建流程、三维重建方法的哈希排序和利用CUDA框架实现三维重建方法.实验可知,所提方法的二维动画人物提取速度最大值为0.401mm/s,而传统二维动画人物提取速度最大值为0.589mm/s.动画人物形象三维构建准确性的对比中,在特殊检测方法中,本文方法对称性更好,准确性也更高.应用对比实验的方法证明,应用新型重建方法后,二维动画人物特征提取速度明显加快,动画人物形象三维构建准确性也大大提升.

To extract the feature of 2D animation characters quickly and complete the image construction of three-dimensional animation characters, a method of feature extraction and 3D reconstruction of 2D animation characters is proposed. By analyzing the feature extraction of 2D animation characters, the determination of Harris operator of 2D animation characters, the determination of the dipole geometry relation of 2D animated characters, and the character purification based on Harris operator and dipole geometry angle are analyzed in detail. Three dimensional reconstruction method based on the features of 2D animated characters is analyzed, which mainly includes the 3D reconstruction process, Hash ordering of 3D reconstruction methods, and the implementation of 3D reconstruction method using CUDA framework. The experiment shows that the maximum speed of 2D animation character extraction is 0. 401 mm/s. The maximum speed of traditional 2D animation character extraction is 0. 296 mm/s. In the comparison of 3D construction accuracy of animation characters, in the special detection method, the symmetry of this method is better. The method of contrast experiment proves that after the new reconstruction method, the speed of feature extraction of 2D animation character is obviously accelerated, and the 3D structure of animation character image is obviously accelerated. The accuracy of construction is also greatly improved.