超高分辨率地形场景纹理图像生成方法研究  

A method for generating ultra-high resolution terrain texture images

作　　者：于子豪 龚光红 李妮[1] YU Zihao;GONG Guanghong;LI Ni(School of Automation Science and Electrical Engineering,Beihang University,Beijing 100191,China)

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《中国体视学与图像分析》2021年第1期39-46,共8页Chinese Journal of Stereology and Image Analysis

基　　金：装备预研领域基金项目(No.61400010203)。

摘　　要：目的航空倾斜摄影图像的空间分辨率受限于无人机飞行高度和图像传感器精度,导致三维重建得到的三维模型的纹理分辨率低,清晰度差。可利用超分辨率重建算法还原图像的纹理细节并提高图像的空间分辨率,最终获得具有超高分辨率纹理的三维模型。方法利用单帧超分辨率重建算法LapSRN处理航空倾斜摄影图像,可以得到的高分辨率图像,并进一步应用于ContextCapture的三维重建流程中。结果可以获得具有超高分辨率纹理的地形场景三维模型。结论研究提出了将超分辨率重建算法应用于三维重建中的一种方法,实验表明,单帧超分辨率重建算法成功地提高了三维模型的纹理贴图分辨率。Objectives The spatial resolution of oblique aerial photographic images is limited by the flying height of drones and accuracy of imaging sensors,which will result in the low texture resolution of the 3D models obtained by 3D reconstruction algorithms.Therefore,super resolution algorithms shall be used to restore the details of texture images and improve the spatial resolution of image to obtain 3D models with ultra-high resolution texture images.Methods A single-frame super resolution algorithm LapSRN is used to process aerial oblique photographic images to obtain high resolution images which can be further processed by the 3D reconstruction pipeline of ContextCapture.Results A 3D model with ultra-high resolution texture images is obtained.Conclusions This paper proposes a method that applies a super resolution algorithm for 3D reconstruction.Experimental results demonstrate that the single-frame super resolution method successfully improved the texture resolution of 3D models.

关 键 词：超分辨率重建 航空倾斜摄影 三维重建 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]