面向车载相机采集图像的智能汽车测试场景关键性量化模型.pdfVIP

面向车载相机采集图像的智能汽车测试场景

关键性量化模型*

［摘要］车载相机图像是构建智能汽车测试场景库的主要数据来源，但其中关键测试场景发生概率低，大部分

场景的测试价值小，若将其直接应用于智能汽车测试会浪费大量测试资源。本文提出一种面向车载相机采集图像

的智能汽车测试场景关键性量化模型。首先，基于实车相机参数对实车采集的图像进行处理，输出对行车安全具有

影响的参数；其次，基于风险场理论将参数整合，输出测试场景关键性量化结果；最后，对实车采集的图像进行测试

场景关键性量化验证，结果表明本文模型可以精确输出测试场景关键性的具体数值，进而对比不同场景的测试价

值，证明本文提出的模型可以有效筛选智能汽车关键测试场景。

关键词：智能汽车；测试场景；关键性量化模型；车载相机图像；风险场理论

ACriticalityAssessmentModelfortheIntelligentVehicleTestScenario

BasedontheOnboardCameraImages

［Abstract］Onboardcameraimagesarethemaindatasourcesforconstructingtheintelligentvehicletest

scenariolibrary，buttheprobabilityofcriticaltestscenariosoccurringintheactualcollectedonboardcameraimag‐

esisverylow，andmostofthescenarioshavelittletestvalue.Ifitisdirectlyappliedtotheintelligentvehicletest，

itwillwastealotoftestresources.Inthispaper，acriticalityassessmentmodelfortheintelligentvehicletestscenar‐

iobasedontheonboardcameraimagesisproposed.Firstly，theimagescollectedfromrealvehiclesareprocessed

basedonthecameraparameterstooutputparametersthathaveimpactondrivingsafety.Then，theparametersare

integratedusingtheriskfieldtheorytooutputthecriticalityassessmentresultsoftheintelligentvehicletestscenar‐

io.Finally，thecriticalityassessmentvalidationisconductedontheimagescollectedfromtheactualvehicle.There‐

sultsshowthattheproposedmodelcanaccuratelyoutputthespecificvaluesofthecriticalityofthetestscenariosin

ordertocomparethetestvaluesofdifferentscenarios，provingthatthemodelproposedinthispapercaneffectively

screentheintelligentvehiclecriticaltestscenarios.

Keywords：intelligentvehicle；testscenario；criticalityassessmentmodel；onb