资源描述:
《基于视频的雾天驾驶场景及其能见度识别算法研究》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、JournalofImageandSignalProcessing图像与信号处理,2015,4,67-77PublishedOnlineJuly2015inHans.http://www.hanspub.org/journal/jisphttp://dx.doi.org/10.12677/jisp.2015.43008ResearchofFogDrivingScenariosandVisibilityRecognitionAlgorithmBasedonVideoWuxueZhu,ChunlinSongCollegeofElectronicsandInformationEngin
2、eering,TongjiUniversity,ShanghaiEmail:zhuwuxue1115@163.comstththReceived:Jul.1,2015;accepted:Jul.14,2015;published:Jul.20,2015Copyright©2015byauthorsandHansPublishersInc.ThisworkislicensedundertheCreativeCommonsAttributionInternationalLicense(CCBY).http://creativecommons.org/licenses/by/4.0/A
3、bstractObtainingreal-time,comprehensiveandaccurateroadtrafficinformationistheimportantpre-conditionandbasicguaranteetopreventtrafficaccidents,andalsoisthekeytorealizetheurbantrafficintelligent.Forrecognitionoffogdrivingscenariosandvisibility,thetraditionalalgorithmhastheproblemsofhighcomplexi
4、ty,poorrobustness,andmoreusinginfixedscene;itisdifficulttoapplytomobiledrivingscenarios.Thispaperproposedafogandvisibilityestimationalgorithmbasedonmonocularvision.Thealgorithm,basedonthelawofKoschmieder,compressesHoughtransformationvotespaceandreducescalculationamountandcomplexitybylimitingp
5、olarangleandradius.Customregionalgrowthsolvestheproblemofpooraccuracyinthemobilescenarios’roadsegmentation.Theweightedaverageofluminancemethodwhichisusedinestimationofin-flectionpointcaneffectivelyremoveinterferenceandensureaccuracy.Thesimulationresultsshowthatthealgorithmcanrealizetherecogni
6、tionoffogandvisibilityinmobilescenarioswithhighaccuracy,real-timeperformanceandrobustness.KeywordsFogDetection,Visibility,Koschmieder,RegionSegmentation,HoughTransformation基于视频的雾天驾驶场景及其能见度识别算法研究朱舞雪,宋春林67基于视频的雾天驾驶场景及其能见度识别算法研究同济大学电子与信息工程学院,上海Email:zhuwuxue1115@163.com收稿日期:2015年7月1日;录用日期:2015年7
7、月14日;发布日期:2015年7月20日摘要获取实时、全面、准确的道路交通场景信息是预防交通事故的重要前提和基本保障,也是实现城市交通智能化的关键。针对雾天驾驶场景及其能见度的识别,传统算法存在复杂度高、鲁棒性差的问题,且多为固定场景下的识别,很难应用于移动的驾驶场景下。本文提出了一种基于单目视觉的雾天识别和能见度估计算法。该算法以柯什米德定律为基础,通过限定Hough变换的极角、半径,压缩投票空间,减少了计算量和复杂度。自定义的区域增长条件,较好的解决了移动场景下的道路分割准确