中文摘要：

3D模型摆正方向的确定,有利于模型对齐、功能恢复等应用.但已有方法只关注于人造模型的处理,且不少方法还需人工干预,工作效率不高.提出一种可处理任意模型的方法,具有很好的适应性.这主要是基于以下的观察：无论是人造模型还是自然模型,其底部往往是表面内容很贫乏的区域;而主要能观察到这些区域的视点,其视点评分一般很低.因此,根据视点评分,就能有效地寻找到模型的底部朝向,即,模型的摆正方向.为了提高计算的可靠性,进一步引入了物体摆放稳定性和人们观察物体习惯性的度量.实验结果表明：新方法能够高效地处理不同种类的模型,包括一些已有方法中难以处理的情况,并具有很好的计算效率.

英文摘要：

Upright orientation of a 3D object is conducive to applications such as model alignment and function recovery. However, state of art methods mainly focus on man-made models only, and some of them require much user interaction, thus lowering the processing efficiency. In this paper, a new method is proposed to effectively attain the upright orientation of an object, no matter whether it is man-made or natural. It is based on the observation that there are very few surface features on the base of the object, and so the views for watching the base will be given very low scores in view evaluation. Thus, using to the view scores, the base of the model can be found effectively to get the upright orientation of the model. To improve computation reliability, two metrics are further integrated into the method to measure the stability of model placement and the habitual viewing direction of human beings to effectively watch a model. Experimental results show that the presented method can effectively handle a variety of models, including some man-made models that cannot be well handled with existing methods, and it can run much faster than existing methods.