LC$^2$: 基于 LiDAR 和相机循环约束的跨模态地点识别（英文版）

英文标题：(LC)$^2$: LiDAR-Camera Loop Constraints For Cross-Modal Place Recognition中文摘要：本文提出一种名为 (LC)$^2$ 的新型交叉匹配方法，通过将 LiDAR 测量值表达为范围图像进行匹配，然后从视差和范围图像中提取本地化描述符来实现无先前点云地图的 LiDAR 本地化，并在姿势图中将最佳匹配作为循环因素，从而通过公共数据集证明了基于 LiDAR 的导航系统可以从图像数据库中优化，反之亦然。英文摘要：Localization has been a challenging task for autonomous navigation. A loopdetection algorithm must overcome environmental changes for the placerecognition and re-localization of robots. Therefore, deep learning has beenextensively studied for the consisten