华中科技大学学报（自然科学版）2025，Vol.53Issue(5)：104-127,24.DOI:10.13245/j.hust.250099
面向自动驾驶的BEV感知算法研究进展
Research progress on BEV perception algorithms for autonomous driving
摘要
Abstract
To further promote the research of bird's eye view(BEV)perception algorithms and improve the performance,the advantages and latest work of BEV perception were discussed,and the relevant datasets and performance evaluation metrics of BEV perception were introduced.The BEV perception processing methods in recent years were summarized,and BEV perception was categorized into three types,which were BEV Cameras,BEV Lidars and BEV Fusion based on the input data of BEV sensor,and in-depth descriptions of the three types of algorithms and solutions were provided.Among them,BEV Camera was classified into explicit transformation and implicit transformation according to geometric transformation methods,and the explicit transformation could be classified into"2D-3D"and"3D-2D"methods,while the implicit transformation could be classified into Transformer-based and multilayer perceptron(MLP)-based neural network modeling methods.BEV lidar could be classified into voxel-based method and Pillar-based method according to the method of point cloud feature extraction.BEV fusion could be classified into camera-lidar fusion,temporal-spatial fusion and depth-height fusion according to the form of input data from multiple sensors.In addition,this review also identified six major directions for future research in this field,aiming to provide valuable insights for researchers engaged in BEV perception algorithms and related technologies,and to encourage a broader academic community to devote increased attention and exploration to the research and application of BEV perception.关键词
自动驾驶/鸟瞰图(BEV)/BEV相机/BEV激光雷达/BEV融合Key words
autonomous driving/bird's eye view(BEV)/BEV camera/BEV lidar/BEV fusion分类
信息技术与安全科学引用本文复制引用
时培成,董心龙,杨爱喜,齐恒..面向自动驾驶的BEV感知算法研究进展[J].华中科技大学学报（自然科学版）,2025,53(5):104-127,24.基金项目
中央引导地方科技专项-长三角科技创新共同体联合攻关计划资助项目(2023CSJGG1600) （2023CSJGG1600）
安徽省自然科学基金面上项目(2208085MF173) （2208085MF173）
芜湖市"赤铸之光"重大科技资助项目(2023zd01,2023zd03). （2023zd01,2023zd03）
