高级检索

    基于序列图像建模的多视图模型合并策略

    Strategy of Combining Multiple Models Based on Image Sequence

    • 摘要: 通过基于单幅图像的建模方法或者基于立体像对的建模方法,可以恢复图像序列中每一幅图像或者立体图像对对应的模型.但所得的是局部模型,并非物体的完整模型.针对这一问题,提出了一种基于图像建模的多视图模型合并策略.根据建筑物模型的规整性,在从单幅图像恢复出建筑物模型的基础上,给出了一种简单灵活、易于实现的针对建筑物的多视图合并方法,从而得到完整的场景模型.该方法大致分为建立视图公共坐标系链表、坐标转换、顶点合并、模型合拢等几部分.从整个过程来看,该方法系统、完善,实验结果显示每一步的操作都是可行的,并且是有效的.

       

      Abstract: A single image or a couple of images always include enough geometry information for reconstructing a portion of model. But it is not sufficient when it comes to the whole model reconstruction, and the error accumulation is also a problem to get the intact model. Based on the structural regularity, a new strategy is proposed, which integrates the whole model by combining separate portions of model reconstructed from each of series images. The strategy is composed of four procedures: constructing the linked list of the view coordinates, converting the coordinates, merging the vertex, and integrating the whole model. The linked list of the view coordinates offers the method to construct the relationship for all the view models. More important, the epipolar line restriction is used to merge the vertex and integrate the whole model. The experimental results show that each step of the strategy is effective.

       

    /

    返回文章
    返回