Single-frame lidar cross self-adaptive feature fusion-based descriptor extraction and recognition method and system, and medium

By using the RMT-PR method to preprocess LiDAR point cloud data with forward and reverse views and fuse cross-view features, the problems of insufficient view reversal and multimodal interaction in LiDAR position recognition are solved, thus improving the robustness and accuracy of position recognition.

CN122416319APending Publication Date: 2026-07-17HUNAN MECHANICAL & ELECTRICAL POLYTECHNIC
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HUNAN MECHANICAL & ELECTRICAL POLYTECHNIC
Filing Date
2026-05-21
Publication Date
2026-07-17

Smart Images

  • Figure CN122416319A_ABST
    Figure CN122416319A_ABST
Patent Text Reader

Abstract

本发明公开了一种基于单帧LiDAR交叉自适应特征融合的描述子提取、识别方法及系统、介质，该方法首先给定一帧点云数据，将其投影为距离图像和鸟瞰图像，并同时构造对应的反向视角表示，即显式构建了距离图、鸟瞰图及其反向视图表示；然后进行模态内跨视图交互，直接建模原视角与反向视角之间的对应关系；此外，还构建了一种混合式Mamba‑Transformer编码器以建模长程依赖关系；最后，对NetVLAD通道进行自适应特征融合，得到判别性更高的全局描述子；再将得到的描述子直接用于搭载激光雷达的移动物的位置识别技术方案上，大大的提升位置识别性能。通过实验验证了各模块设计有效提升了LiDAR位置识别的鲁棒性。
Need to check novelty before this filing date? Find Prior Art