GNSS拒止环境下基于Puzzle描述符的无人机点云匹配自主定位

Autonomous positioning of UAV based on puzzle descriptor in denial environment

针对GNSS拒止环境下无人机基于点云匹配自主定位中存在的搜索时间长、误配率高等问题,提出了一种基于Puzzle描述符的二视点云全局配准算法。该算法采用改进的ScanContext点云描述符,对点云进行分割与编码,设计出新的Puzzle描述符。通过SIFT匹配和模板匹配确定预搜索区域,并利用ICP匹配实现无人机的精确定位。仿真结果表明,算法有效降低了搜索时间和误配率,且具备良好的噪声抑制能力,三维定位误差控制在±20 cm以内,单次匹配耗时小于1 s。挂飞实测数据进一步验证了算法的性能,无人机自主定位的平均误差为19.94 cm,单次匹配平均耗时0.59 s,充分满足了精度和实时性要求。

In order to address the issues of long search time and high mismatch rates in UAV autonomous positioning based on point cloud matching under GNSS-denied environments,this paper proposes a two-view global point cloud registration algorithm based on Puzzle descriptors.The algorithm employs an improved ScanContext point cloud descriptor for segmenting and encoding the point cloud,leading to the design of a novel Puzzle descriptor.SIFT matching and template matching are applied to determine the pre-search area,followed by precise localization using ICP matching.The algorithm effectively reduces search time and mismatch rates,while also offering strong noise suppression capabilities.Simulation results show that the 3D positioning error is controlled within±20 cm,and the time for a single matching process is less than 1 second.Real-world flight test data further validate the algorithm’s performance,with an average UAV positioning error of 19.94 cm and an average single matching time of 0.59 seconds,meeting the requirements for both accuracy and real-time UAV positioning.