一种无人机影像实时增量平差镶嵌算法

A Real-time Incremental Adjustment Mosaic Algorithm for UAV Images

当前无人机已成为空间信息获取的重要手段,可作为航空摄影测量系统的有益补充。但无人机数据量大、姿态稳定性差、排列不规则等缺点给快速处理带来较大难题。本文利用运动恢复结构SFM方法中的增量平差技术,借助初始POS实现无人机影像的快速实时处理,并利用GPU技术对平差后影像进行实时纠正和虚拟镶嵌展示。试验表明,增量平差和实时纠正镶嵌技术能大大缩短影像处理时间,可满足灾害应急响应时对影像进行实时处理、快速获取现场灾情信息的需求。

At present,UAV have become the important means of acquiring spatial information and can be used as a useful complement to aerial photogrammetry systems.However,the shortcomings such as large data volume,poor attitude stability,and irregular arrangement of UAV bring great difficulties to fast processing.In this paper,the incremental bundle adjustment technique in the SFM method of motion recovery structure is used to realize the real-time processing of UAV images with the initial POS,and the GPU technology is used to perform real-time correction and virtual mosaic display of the adjusted images.Tests show that for each additional image,the average processing time of the incremental bundle adjustment is less than the shooting interval,and the incremental bundle adjustment has the advantages of high stability,high automation,fast speed,etc.,and has great advantages in the emergency handling of emergencies.