基于自动驾驶典型场景的多模态感知数据集构建

打开文本图片集

中图分类号：U469.79；TP181 文献标志码：A DOI： 10.20104/j.cnki.1674-6546.20250039

【Abstract]Toaddressthescarcityof multi-source heterogeneous dataand insuficient scenario adaptability incurrent perceptionalgorithm trainingand testingofautonomous driving，atypical scenario-basedmultimodal perceptiondatasetis constructed.Itcontains 10 specific typicalscenariosegments，covering multimodalsensordata from LiDAR，cameras，and 4D millmeter-waveradar.Thedatesetprovidesannotationinformationforsixcategoriesof targetsandofersdetaileddescriptions of dataacqusitiondeviceconfigurations，includingsensorparameters，calibrationdata，andatimesynchronizationproesing scheme.Bydeliveringscenario-specificdrivingcontext，theconstructed dataset enhances perceptionaccuracy incomplex environments，thereby improving the safety and reliability of autonomous driving systems.

Key Words： Perception dataset， Typical scenarios，Multimodal，Autonomous driving

【引用格式】刘艺，王瑶，裴世康，等.基于自动驾驶典型场景的多模态感知数据集构建[J].汽车工程师，2025（8）：15-21. LIU Y，WANG Y，PEI S K，etal.Construction of a Multimodal Perception Dataset for Autonomous Driving in Typical Scenarios[J].Automotive Engineer，2025（8）： 15-21.

1前言

随着自动驾驶技术的快速发展和智能交通系统的广泛应用，环境感知能力成为衡量自动驾驶车辆安全性能的核心指标。（剩余8143字）