一种基于GPU并行计算的无人机影像快速镶嵌方法

基金项目:
住房和城乡建设部科学技术项目"基于多源异构遥感影像的辅助城乡规划督察信息化平台建设及应用研究"

重庆市科学技术委员会社会事业与民生保障科技创新专项项目"基于倾斜摄影技术的重庆市城乡规划公众服务平台建设与应用"

详细信息

刊出日期: 2017-12-15

A method for rapid UAV images mosaicking based on GPU parallel computing

Publish Date: 15 December 2017

摘要: 提出了一种从匀光后无人机影像出发,以Voronoi图为镶嵌线网络,基于GPU并行计算的无人机影像快速镶嵌方法.首先,通过Wallis滤波处理影像间色彩不一致问题;然后,以测区影像位置自动生成Voronoi图镶嵌线网络;最后,基于GPU并行计算将无人机影像快速正射纠正并镶嵌.通过对230张空间分辨率为0.1 m的无人机影像进行快速纠正镶嵌,实验结果表明,该方法较传统方法效率有很大提升.
- 影像匀光 /
- 无人机影像 /
- 正射纠正 /
- 影像镶嵌 /
- GPU并行计算
Abstract: A method for rapid mosaicking UAV images is introduced, which is based on Voronoi network and GPU parallel computing. Firstly, the radiometric difference between the images is solved by Wallis dodging. Then the seamline network based on Voronoi diagram is created by the locations of all UAV images. At last, the UAV images are ortho-rectified and mosaicked rapidly based on the GPU parallel computing. Experimental results indicate that the method' s efficiency has increased by several times compared with the traditional methods by mosaicking 230 UAV images whose resolution is 0. 1 meter. The mere rectification of the valid area of every UAV image and the application of GPU parallel computing not only avoid the huge data redundancy but also greatly reduces the computation time.
- image dodging /
- UAV images /
- ortho-rectification /
- image mosaicking /
- GPU parallel computing

中国自然资源航空物探遥感中心	主办
地质出版社	出版