摘要
针对基于视频的无人机三维重建图像匹配计算量大,整体优化延迟高等问题,在视觉SLAM的基础上,该文提出一种融合无人机POS信息、并行密集匹配、点云匹配技术的实时无人机三维重建系统。首先,利用POS信息获取视频帧之间的重叠度,结合特征提取获取关键帧,减少关键帧连续提取导致计算量大且初始化不稳定的缺点。其次,结合POS信息提供的位姿获取地面点初始值优化初始三维立体模型。最后,采用点云融合技术对点云进行匹配从而得到精确的点云。在自建无人机实飞数据上进行实验验证可知,该文算法能够实时完成稠密点云的三维重建,重建后的三维模型局部纹理特征明显、精确度高,适合无人机测绘等领域的实时三维重建场景。
Aiming at the problems of large calculation amount and high overall optimization delay of video-based UAV 3D reconstruction image matching,a real-time UAV 3D reconstruction system based on visual SLAM is proposed,which integrates UAV POS information,parallel dense matching and point cloud matching technology.Firstly,POS information is used to obtain the overlap between video frames,and feature extraction is used to obtain key frames,which reduces the shortcomings of large amount of calculation and unstable initialization caused by continuous extraction of key frames.Secondly,the initial value of the ground point is obtained by combining the pose provided by POS information to optimize the initial 3D model.Finally,the point cloud fusion technology is used to match the point cloud to obtain the accurate point cloud.Experimental verification on the self-built UAV real-flight data shows that the algorithm in this paper can complete the three-dimensional reconstruction of dense point clouds in real time.The reconstructed three-dimensional model has obvious local texture features and high accuracy,which is suitable for real-time three-dimensional reconstruction scenes in the fields of UAV surveying and mapping.
出处
《科技创新与应用》
2024年第7期1-4,共4页
Technology Innovation and Application
基金
北京市科技新星计划(Z191100001119111、Z201100006820107)。