多视觉环境下联合多约束的三维目标体重建.docVIP

多视觉环境下联合多约束的三维目标体重建 　　摘 要： 　　提出了一种联合多种约束条件的体元重建方法。首先将场景空间离散为体元，并从多视获取的二维图像中提取目标轮廓信息；其次利用侧影轮廓约束及颜色一致性约束，恢复目标三维信息；针对重建模型存在漂浮体元和毛刺现象，提出加入平滑约束解决该问题；最后提出模型修复算法对模型表面空洞进行填补，解决重建模型空洞问题。实验结果表明，该方法可重建颜色纹理信息准确的目标外形，且模型光滑细腻，表面空洞得到有效填补。 　　?ス丶?词： 　　体元重建；侧影轮廓；体元一致性；外形获取；空洞填补 　　?ブ型挤掷嗪牛? TP391.41 　　 文献标志码：A 　　英文标题 　　?? 　　3D voxel reconstruction under multiple constraints in multi??view environment 　　?び⑽淖髡呙? 　　CHEN Ning, LI Xiu??xiu, ZHENG Jiang??bin 　　?び⑽牡刂?( 　　School of Computer Science, Northwestern Polytechnical University, Xi’an Shaanxi 710129, China 　　英文摘要 　　)?? 　　Abstract: 　　A new voxel reconstruction under multiple constraints was proposed. Firstly, the scene was discreted to voxel, and target silhouette information was extracted in multi??view??obtained 2D image. Secondly, two constraints were used to recover target 3D information, which were silhouette constraint and the color consistency constraint. Smooth constrict was used to solve the problem that reconstructed model had float voxel and glitch. Based on the composition of the three forces, the floating voxels and the burrs could be restrained efficiently. Finally, a model recovering algorithm was proposed to fill the holes on the surface of the reconstructed model. The experimental results demonstrate that the method can reconstruct the target appearance with correct color and texture information, also the model is smooth and delicate, thus the hole can be filled effectively. 　　英文关键词 　　??Key words: 　　voxel reconstruction; silhouette; voxel consistency; shape acquisition; hole filling 　　?? 　　0 引言?? 　　多视环境下真实目标的三维重建作为计算机视觉领域中一项关键技术，已成为当今世界的一大研究热点，广泛应用于影视、动画、游戏等数字媒体制作，生物力学分析，虚拟现实等多个领域，具有极高的社会价值、巨大的经济效益以及高速增长的市场需求。?? 　　基于体表示的真实目标获取技术［1-2］是三维重建领域的一个重要分支，体重建技术将整个场景空间离散为体元，通过原始图像上目标轮廓或图像间的相关性等约束，判断体元是否属于目标表面。基于侧影轮廓的方法［3］在体重建技术中有重要应用，该方法提取目标在输入图像上的侧影轮廓线，通过摄像机投影中心和投影轮廓线的对应关系来判断体元属性，从而获取目标粗糙的外形估计，但无法获取颜色纹理信息。体元着色算法［4］是体重建技术中的经典算法之一，通过计算体元在输入图像投影像素集的颜色一致性，判断体元是否属于目标表面。该算法能够有效地获取三维目标物体的颜色纹理信息，避免了图像间的相关特征匹配问题