"True" Orthophoto Generation Based On Tilt Images

Compared with the traditional orthophoto,the "true" orthophoto is more accurate,real and comprehensive,it is an important information source,which can be used to study the city's population,resources,environment,disaster and other urban sustainable development laws and major issues in society.However,being limited by current treatment of theories and techniques,the existing large-scale city "true" orthophoto generation are of high costs and long cycle,consequently the needs of urban high-speed development and construction cannot be met.In this paper,five-direction tilted aerial imagery is used as data source to carry out the research about urban "true" orthophoto generation.The main contents are as follows:(1)The research background and significance of "true" orthophoto generation will be firstly mentioned in detail.Systematically statement of domestic and abroad studies and development trend follows after that.(2)Considering the problems that the traditional SGM intensive matching process is unable to take the impact of occlusion into account,and with the disadvantage that image occlusion area cannot be worked out,an improved algorithm based on SGM is proposed.It consists of three steps:(a)The occlusion model is built.The pixel occlusion restrict which is applied to the epipolarline is introduced to the 0° and 180° direction matching paths of the SGM.Then building a new function of new matching cost which consider occlusion,based on the similarity measurement of pixel mutual information and the structure of parallax spatial image;(b)The piecewise smooth constraint.The SGM matching path except 0° and 180° is segmented by the occlusion area,which makes the parallax search being limited to the local path with current pixel but without occlusion.thus overcoming the damage of occlusion to the matching path continuity and as a result,SGM smooth constraints and its existing calculation can be effectively used;(c)Pyramid image matching.Firstly,based on the new matching cost function,the initial occlusion area(occlusion diagram)and the disparity map are obtained by matching with the maximum scale pyramid image.Then transfer of two map layer-by-layer and re-match,update based on them,and output the correct occlusion and parallax.(3)Based on the parallax information of the improved SGM,the digital differential correction principle of orthoimage imaging is introduced and digital surface model DSM is generated.Then occlusion detection information of the improved SGM is introduced to the process of "true" orthophoto generation.Finally,the pixel is filled with the masking area to complete the generation of the "true" orthophoto.The feature of the method in this article is that,pixel masking limit which applied to the epipolar line is naturally introduced into the SGM multipath matching cost calculation process,and the occlusion area is clearly given at the same time as the disparity map,and the three-dimensional model and priori knowledge of the image space parameter are not needed.Moreover,because this method can effectively solve the impact of the occlusion phenomenon toward the non-occlusion area pixel matching,reduce the pixel matching error accumulation and propagation,and the output parallax more uniform,the target boundary is more completely,clearly and the quality is improves further.The method has high practical value for the photogrammetry 4D product high quality generation and tilt image application processing,such as computer automatic three-dimensional reconstruction and so on.