摘要
设计并实现了一种适用于高分辨无人机序列图像的快速三维重建系统(FDroneMap)。通过两个方面的改进提高系统运行的效率:一方面对重建算法进行优化,使用了一种新的基于哈希表的图像匹配方法,并根据无人机图像的时空序列特性对匹配策略进行调整,加快图像匹配速度;另一方面对重建算法各个模块进行并行化设计,提高系统对硬件的性能使用率。对比实验表明,在处理高分辨率无人机序列图像时,FDroneMap能显著提升三维重建的效率,并且能保证重建精度。
A fast 3D reconstruction system called FDroueMap for sequence image of unmanned aerial vehicle（UAV） is designed and implemented. The efficiency of the system is improved through two aspects. One is optimizing reconstruction algorithm, using a new image matching method based on hash table and adjust image matching strategy according to the ＂sequential characteristic in time and space＂ of UAV to accelerate the speed of image matching. On the other hand, each module of the reconstruction algorithm is designed in parallel, which improves the performance of the system. The contrast experiments show that FDroneMap can significantly improve the efficiency of 3D reconstruction in the process of high resolution UAV images, and can guarantee the accuracy of the reconstruction.
出处
《电子技术应用》
北大核心
2017年第6期134-137,142,共5页
Application of Electronic Technique
基金
国家自然科学基金(41371384
41401465)
关键词
无人机
哈希匹配
时空序列
并行
unmanned aerial vehicle
matching based on hash table
sequential characteristic in time and space
parallel
