一种点云图像跨模态匹配约束的快速初始化方法

By constructing a cross-modal training set and a semantic mapping model, a hybrid feature map is generated, which solves the problem of mismatch in robot localization initialization in complex and repetitive texture environments, and achieves fast and accurate localization initialization.

CN122151105BActive Publication Date: 2026-07-17WUHAN HUANYU ZHIXING TECH CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
WUHAN HUANYU ZHIXING TECH CO LTD
Filing Date
2026-05-08
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing technologies struggle to achieve fast and accurate localization initialization in complex environments with repetitive textures, leading to system mismatches and navigation malfunctions.

Method used

By constructing a cross-modal training set, a cross-modal semantic mapping model is trained to generate a hybrid feature map. This map is then combined with real-time environmental images and point cloud data for feature matching and local rigid registration, resulting in an accurate six-DOF relocalization pose.

Benefits of technology

This technology enables rapid and accurate robot localization initialization in complex and repetitive texture environments, avoiding texture confusion and ensuring the accuracy and reliability of localization results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122151105B_ABST
    Figure CN122151105B_ABST
Patent Text Reader

Abstract

本发明公开了一种点云图像跨模态匹配约束的快速初始化方法，具体涉及移动机器人定位初始化领域，用于解决重复纹理场景下任意位置启动时图像特征易歧义、点云搜索范围大、粗略定位易失准且初始化过程难以兼顾速度与可靠性的问题；该方法通过对先验图像序列与先验点云数据执行投影对齐，构建跨模态训练集并训练跨模态语义映射模型，生成与视觉关键帧、先验位姿节点相绑定的混合特征地图，再依据实时环境图像提取当前几何感知视觉描述子，完成目标关键帧检索、匹配判别和粗略全局初始位姿解算，并结合当前实时点云数据实施受限搜索下的局部刚性配准，输出精确六自由度重定位位姿。
Need to check novelty before this filing date? Find Prior Art