Research Of True Orthophoto Generation Based On Unmanned Aerial Vehicle Images

Unmanned Aerial Vehicle(UAV) can take high resolution images faster and safer with the characteristics of maneuvering and flexible. However, the images of UAV are small and the spatial artifacts are very clearly in the images which will cause occlusion. True Orthophoto use DSM to eliminate the projection errors of terrain and artifacts, the artifacts can be rectified to correct position in true orthophoto and it will not be occluded other objects. True orthophoto has both geometric accuracy of map and visual properties of image. Uniform scale and the absence of relief displacement in orthophotos make them an important component of GIS databases. The most siginificant difference between true orthophoto and DOM is occlusion detected and visibility analysis when differential rectification.Most of the research use the traditional photogrammetry method to process the images of UAV, which didn’t consider the small and abundant characteristics of UAV images. Meanwhile, there is no visibility analysis when ortho-rectified processing. The Z-Buffer methodology is the popular method on the identification of occluded portions, which suffer from several problems such as its sensitivity to the sampling interval of the DSM as it relates to the ground sampling distance of imaging sensor.This paper focus on cites computer vision 3D reconstruction method by Structure from Motion(SfM), and apply Scale-Invariant Feature Transform(SIFT) method to the sequence of UAV images to get a large number of matching points. Then derive photography station position and further estimate the camera orientation parameters(camera calibration). At last, restore the three-dimensional scene information based on stereo matching algorithms and then generate digital surface model. Then, this paper use the angle-based methodology for true orthophoto generation identifies occluded areas to circumvent the problems associated with Z-Buffer method. Moreover, this paper will also introduce angle-based methodology for occlusion detection and true orthophoto generation by checking the angle along a radial direction while moving away from nadir point.Through the experiment analysis, we can find that it’s feasible and effective to generate the DSM through computer vision 3D reconstruction method. However, for the area of low buildings and abundant plants, we need modify the DSM to meet the demands. Based on the modified DSM, using the angle-based true orthophoto generation method to replace Z-Buffer method. Checking the α angle along a radial direction while moving away from nadir point for the occlusion detection does not depend on the relative relationship between the DSM cell size and the GSD of imaging sensor. Meanwhile, spiral sweep is introduced for occlusion detection and true orthophoto. The experimental results have shown the improved performace of angle-based methodology when compared to the traditional Z-Buffer technique.