3D Reconstruction Of The Great Wall Relics Based On LiDAR Unorganized Point-cloud

Since appeared in1990s, laser scanning technology has greatly promoted thedevelopment and application of3D reconstruction technology, from the early CAD(Computer Aided Design) to the reverse engineering, urban construction planning,geological exploration, protection of cultural relics, etc. Especially in the field of GIS(Geographic Information System), due to its fast scanning, flexibility, high-precision andwide applicability,3D laser scanning technology makes it possible to acquire mass dataimmediately, which makes it popular in3D-GIS system. However, the introduction of the3D reconstruction into GIS also brings texture-mapping problem. At present,3D laserscanner can not obtain the realistic texture of the target object as well as high-precisiongeometric information at the same time, and most models use monochrome or false colorto represent the texture. Texture information can be implemented by software, but onlyworks for object with regular geometry, such as digital city model. For the non-regulargeographical landscape, such as the Great Wall ruins, it fails. What’s more, the texture-mapping is designed by computer, which means it can not reflect the real world. Thisdefects the sense of reality of GIS system. At the meantime, digital camera can obtainhigh-resolution image of the target object. How to merge together the advantages of bothto generate a realistic model in GIS system is still a hot research topic.Besides the advantage of3D laser scanning technology,3D laser scanners on themarket are still expensive and cumbersome, especially for UAV (unmanned aerial vehicle)system. This dissertation presents a low-cost and portable scheme for UAV-based3Dreconstruction in GIS. For the hardware, we use a cheap2D laser scanner mounted witha rotary table to simulate3D laser scanner, and a CCD (Charge-coupled Device) camerato obtain high-resolution texture images at the same time. For the Software, we use thePower-Crust algorithm instead of the more popular ones such as growth method, divide-conquer method, etc. to adapt to the scattered point-cloud better and use the cameraprojection mapping algorithm based on sparse TIN (Triangulated Irregular Network)mesh model to complete the texture mapping with multiple texture mapping technologyto solve the texture fusion problems such as texture fragmentation.In this dissertation, we briefly talk about the issues of3D modeling for geographicallandscape and realistic texture-mapping in3D-GIS system and simply implement theexperimental prototype.