基于二值图像的三维人体模型重建

Human body shape reconstruction based on binary image

为了能够基于最小的图片输入快速地重建三维人体模型,提出了一种输入人体轮廓二值图像对三维人体模型进行精准重建的新方法。这种对三维人体模型进行重建的方法首先通过主成分分析法对人体进行编码得到了人体的低维形状描述子,随后构建了一个新的深度卷积神经网络,其拥有2个分支,分别用于提取人体正视图与人体侧视图的相关特征;将分支特征融合后,通过联合训练,可以学习到一个从输入到形状描述子的全局映射;最终对形状描述子进行解码得到点云数据,完成了对三维人体模型的重建。结果表明:该方法与现有的基于图像的三维人体模型重建技术相比,在精度上提高了1.07%,并证明了多视角的预测结果优于单视角预测结果。权重共享网络的预测结果优于权重不共享网络的预测结果。

In order to quickly reconstruct the three-dimensional human body based on the minimum image input,a new method for accurate reconstruction of the three-dimensional human body by inputting binary images was proposed.The shape of the human body was first encoded via principal component analysis(PCA)to extract the low dimensional shape descriptor,then,a novel body reconstruction convolutional neural network(BRCNN)with two branches was designed,which could capture deep correlated features from front and lateral views and merge them.The BRCNN was jointly trained to learn a global mapping from the input to the shape descriptor which can be then decoded to points cloud for the reconstruction of various body shapes under neutral poses.The experimental results show that compared with the existing human reconstruction technology,the accuracy has been improved by 1.07%,and the prediction results of the two views are better than those from the single view.Further investigation also reveals that the prediction results of the weight-sharing network are better than the network without weight-sharing.