Research On Key Issue Of Airborne LiDAR Based True-orthophoto Production

Orthophoto is of rich orthrographic image and precise geometric information, it can be directly used in vector GIS database as well as measure work. Moreover, it plays an important role in many other applications such as digital mapping, city 3D modeling, etc. Traditional orthophoto is rectified with DEM (Digital Elevation Model) to correct the relief displacement caused by terrain ups and downs and sensor tilt. While True orthophoto not only corrects these relief displacements, but also need to correct the surface objects’ relief displacement, which requires the regional DSM (Digital Surface Model). The emergence and development of LiDAR (Airborne Detection and garages Ranging) system can provide high quality data source for DSM, which is the necessary condition for the production of true orthophoto.At present, the production of true orthophoto is still at an early stage, with many difficulties in the real production constraining its development, such as:image color dodging, the detection of occlusion, building boundaries sharpeness, the inpainting of occlusion area and the selection of mosaicking seamlines, etc.Based on the data of LiDAR system, this dissertation will study the key issues of true-orthophoto production. The main contents and innovative findings are as follows:(1) Consistency processing of Airborne image dodgingStudies will be conducted on the inconsistency of the processing of image dodging for single image and multiple images respectively. For single image dodging, a local standard deviation model based grid image dodging method will be proposed, which can solve the color difference caused by vignetting, camera angle difference, etc. While for multiple images, a grade-curve base method is proposed to resolve the color difference between multiple images.(2) occlusion detectionOcclusion detection is the key step in true orthophoto production. Based on the study of current detection methods, a new multi-scale method is proposed in this study, which detects the occlusion area roughly at first, then based on the result a TIN (Triangulated Irregular Network) model is constructed adaptively for the occlusion detection. The method can not only ensure the quality of detection, but also improve the efficiency.(3) building boudary sharpenessSince the building boudaries in true orthophoto are always not sharp, a new method which combines the LiDAR data and images for precise boudaries extraction is proposed in this study. The precise 3D boudaries can improve the quality of DSM to sharpen the boudaries of buildings in the result images.(4) inpainting of occlusion areaIn order to preserve the integrity of image information, the occlusion area should be repaired. For overlapped occlusion area, Amhar et al.,1998, Sheng et al.,2003 proposed several methods for image inpainting with assistance of overlapped images. In current study, an emphasis is laid on the occlusion area which has no information to inpainting at all. A vector constrained image inpainting method is proposed to preserve the integrity of building roofs.(5) seam-line network optimizationThe seam-line network refers to the network which is constructed by multiple seam-line between images during the masaiccking. In this dissertation, seam-line network generation method is studied to generate the holistic seam-line and effective enchase polygon separately, which will increase the efficiency of enchase processing. The optimization of the network is processed on the basis of LiDAR data, which can improve the mosaiccking quality.Due to the limitation of time, only the above aspects will be covered in this study. However, there may still exist many other questions in the true orthophoto production, which can be the research contents for future study.