面向无人机识别的同平台LiDAR-可见光数据集构建OA

To meet the requirements of unmanned aerial vehicle recognition,multi-source information fusion-based unmanned aerial vehicle recognition technology provides an effective solution.Existing anti-unmanned aerial vehicle datasets are reviewed.Currently,publicly available unmanned aerial vehicle datasets are dominated by visible-light images and lack multi-source datasets with strict spatiotemporal registration and alignment,making it difficult to effectively support research on multi-source information fusion-based unmanned aerial vehicle recognition.The publicly available Multi-Modal Anti-Unmanned Aerial Vehicle Dataset(MMAUD)multimodal dataset integrates multiple modalities,including visible-light fisheye cameras and Light Detection And Ranging(LiDAR).However,its raw data suffer from several issues,such as uncalibrated image distortion,a large amount of invalid LiDAR data,and the lack of data annotations,which collectively limit its use in anti-unmanned aerial vehicle application research.Therefore,building upon the MMAUD multimodal dataset,a Co-platform LiDAR-visible Dataset for unmanned aerial vehicle Recognition is constructed by performing temporal-spatial alignment,geometric calibration,distortion correction,and unmanned aerial vehicle data annotation.This provides a reliable dataset foundation for research on multimodal anti-unmanned aerial vehicle perception and information fusion.