欢迎来到天天文库
浏览记录
ID:58312791
大小:1.06 MB
页数:9页
时间:2020-06-01
《地面剧烈旋转运动背景中运动目标检测.pdf》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、第35卷第9期仪器仪表学报V01.35No.92014年9月ChineseJourna1ofScientificInstrumentSep.2014地面剧烈旋转运动背景中运动目标检测:l:闫钧华一,储林臻,艾淑芳,李大雷,段贺(1.南京航空航天大学航天学院南京210016;2.光电控制技术重点实验室洛阳471009)摘要:针对地面背景剧烈旋转运动下运动目标难以检测的问题,提}=};基于彩色光流计算与运动补偿的运动目标检测算法。首先利用金字塔分层计算、I,ucas局部运动一致性,提出一种改进的彩色金字塔Lueas光流算
2、法,以获得背景剧烈旋转运动下的准确可靠的图像光流场;然后基于光流场估计背景运动透视模型参数,重构图像纯背景运动矢量场,对图像进行背景运动补偿;最后通过差分二值化图像与形态学处理检测出运动目标。实验结果表明算法能够有效地改善光流计算效果,光流场局部方向一致性平均误差为0.29。,局部矢量模值一致性平均误差为0.25pix,该算法能在地面剧烈旋转运动背景中准确地检测出运动目标。关键词:目标检测;彩色光流场;地面背景;旋转运动;运动补偿中图分类号:TP391.41TH7文献标识码:A国家标准学科分类代码:510.40Mov
3、ingobjectdetectioninviolentlyrotationalgroundbackgroundYanJunhua一,ChuLinzhen,AiShufang,LiDalei,DuanHe(.CollegeofAstronautics,NangUniversityofAeronautics&Astronautics,Nanjing210016,China;2.KeyLaboratoryofScienceandTechnologyonElectro—opticControl,Luoyang471009,C
4、hina)Abstract:Aimingatthedificultyinmovingobjectdetectionunderviolentlyrotationalgroundbackground,weproposeamovingobjectdetectionmethodbasedoncoloropticalflowfieldandmotioncompensation.First,animprovedcolorpyramidLucasopticalflowalgm‘ithmisproposed,whichisbased
5、onimagepyramidcalculationandLucaslocalmotioncoherence.Withthehelpoftheimprovedcoloropticalflowalgorithm,theaccurate,reliableopticalflowfieldisobtained,whichreflectsthemovementoftheimagepixelsunderviolentlyrotationalgroundbackground.Fromtheopticalflowfieldthepar
6、ametersoftheperspectivebackgroundmotionmodelareestimatedandthemotionvectorfieldforpurebackgroundinformationisreconstructed.Thenthebackgrounddifferenceoftwoflamescanberemovedbymotioncompensation.Finally,themovingobjectisidentifiedthroughtheframedifferencemethoda
7、ndmorphologicalprocessing.Experimentresultsshowthatthemethodcansignificantlyimprovethequalityofopticalflowfieldcalculation;theaverageerroroflocaldirectionconsistencyoftheopticalflowfieldis0.29degree;theaverageerroroflocalvectormodulusconsistencyoftheopticalflow
8、fieldis0.25pixe1.Theproposedalgorithmcanaccuratelydetectthemovingobjectfromvi—olentlyrotationalgroundbackground.Keywords:objectdeteetion;coloropticalflowfield;groundbackgrou
此文档下载收益归作者所有