近景数码影像中墙面纹理自动拼接方法的研究

影像拼接是从近景数码影像中生成墙面纹理需要解决的问题。然而由于现有的拼接软件通常不适用于处理几何变形大的近景影像,但通过对从近景数码影像中自动拼接墙面纹理的方法进行的研究发现,在相机的位置离建筑物相对较远,且大致沿一条直线进行拍摄时,当原始影像预先纠正到同一个铅垂面上后,由于各影像之间的大小比例接近于1,因此对于此种情况的拼接而言,就只需要通过选取拼接点来确定相邻影像之间沿X轴和Y轴方向的平移量,即可实现拼接。然而对于比较狭窄的街道(如步行街)来说,则由于摄站与建筑物之间的距离比较小,使影像的倾角增大,导致影像的几何变形增大,所以不适于采用选取拼接点的方法进行拼接。针对此种情况的拼接,提出了一种类似于航带法生成正射影像的条带法,用来进行墙面纹理的自动拼接,并且结合影像间同名水平直线的约束和加密匹配同名点两种方法来对拼接结果进行优化。最后针对地面实拍的序列影像进行了实验。实验结果证明,该方法能够显著消除相邻影像间的几何差异,从而真正实现了影像的无缝拼接。

The Research on Approaches of Automatic Mosaic for Building Facade Texture from C lose-range Digital Imagery

Image mosaic is a problem inevitable faced in reconstruction for facade texture from close-range digital imagery.The approach of automatic mosaic for building facade texture from close-range digital imagery is studied in this paper.When the camera is distant from building and the images are taken along a linear route,after the rectification process,only mosaic point pair is acquired to implement mosaic process.However,this approach won't be applicable when the imagery is taken too close to the facade,since the geometric distortion cannot be eliminated perfectly after rectification process.Therefore,strip method,similar to strip block adjustment,is presented to generate the whole facade texture.Two refining strategies are presented to optimize the mosaic result.One is to refine mosaic region by constraining correspondingly horizontal lines.The other is to refine by choosing more corresponding points according to the relative spatial relationship of the image sequence.Experiments are carried out on ground-leveled image sequences.It is proved that seamless mosaic is realized by the method presented because the geometric distortion between each image is eliminated perfectly.