基于MeanShift算法的运动目标跟踪及在网球出界判罚中的应用

本科毕业设计论文基于MeanShift算法的运动目标跟踪及在网球出界判罚中的应用院、系:数理系学科专业:信息与计算科学学生:朱晶学号:051001119指导教师:薛鹏翔2020年6月I基于meanshift算法的运动目标跟踪及在网球出界判罚中的应用摘要本文的主要研究内容是基于meanshift算法的运动目标跟踪，并将这种跟踪应用于网球出界判罚。目标跟踪广泛应用于监控、视频编码以及军工领域.如何有效地在视频序列]1[中对感兴趣目标进行有效跟踪,是计算机视觉中一个极具挑战性的课题，其中最主要的环节就是选择合适的目标特征以及快速的搜索匹配算法，Mean-Shift作为一种高效的模式匹配算法,由于不需要进行穷尽搜索,已经成功地应用在对实时性要求较高的目标跟踪领域中。运动图像的检测不仅在工业和军事上具有极为重要的应用，而且在一些重要的场所，例如银行、公共交通等系统中也有广泛的应用。基本原理是将跟踪目标的色彩信息做为特征，将这些信息计算处理后投影到下一帧图像中，计算出这幅图像中的目标，用这幅图像作为新的源图，分析下一帧图像，重复这个过程就可以实现对目标的连续跟踪。本文主要工作如下：1、使用meanshift算法实现对运动物体的跟踪，根据颜色直方图的信息，将这些信息计算处理后投影到下一帧图像中，计算出这幅图像中的目标，用这幅图像作为新的源图，分析下一帧图像，重复这个过程就可以实现对目标的连续跟踪。2、传统的meanshift算法在目标快速运动时，容易受到背景区域的干扰，不能准确地对运动目标区域进行跟踪。本文提出一种基于背景优化的meanshift目标跟踪算法，通过引入混合直方图并对直方图重新量化来进行优化，以及减少背景像素在概率密度函数中的权重来降低背景区域对跟踪目标的影响，从而实现对运动目标的准确跟踪。3、将Mean-shift算法应用到RGB图像中,采用了以方向直方图建立目标模型的策略,提出了在RGB图像中以Mean-shift为核心的目标跟踪算法，实验结果表明,该算法具有不受光照条件影响的优点,在低对比度的情况下仍能实现稳定、实时的跟踪目标,复杂背景下的目标检测作为整个试验的关键技术发挥着十分重要的作用，并与帧间信息相结合来提取运动对象研究方案。本文分为六部分：第一部分首先阐述了课题的研究背景、意义及该领域的发展现状；第二部分介绍了meanshift经典算法；第三部分介绍Hough变换进行直线检测的思想及原理，第四部分使用meanshift算法II为核心的算法跟踪目标并得到了相关的实验结果，并给出了大量实验图片；第五部分提出了研究展望以及后续的主要内容。关键词：运动目标检测；meanshift算法；核函数；IIIBasedonthemeanshiftalgorithmtotrackmovingtargetsandtheapplicationofTennisoutpenaltyAbstractThemainresearchcontentisbasedonthemeanshiftalgorithmtotrackmovingtargets,andusingthisfollow-upoutofboundspenaltyappliestotennis.Trackmovingtargetsarewidelyusedinsurveillance,videocoding,aswellasthefieldofmilitaryindustry.Howtoeffectivetrackinginthetargetsofinterestvideosequences,thatisachallengingtaskincomputervision,whichisthemostimportantaspectofselectingsuitabletargetquicksearchfeatures,aswellasmatchingalgorithm,Mean-Shiftasanefficientpattern-matchingalgorithm,asaresultofwhichdoesnotrequireexhaustivesearch,hasbeensuccessfullyusedinreal-timerequirementsofthehigherareasoftargettracking.Detectionofmotionpicture,anotonlyinindustrialandmilitaryapplicationisextremelyimportant,butalsoinsomeimportantplacessuchasbanks,publictransportsystemsarealsowidelyused.Thebasicprinciplesofthegoalaretotrackthecharacteristicsofthecolorinformation.Usingtheinformationcalculationprocessingtothenextframeaftertheprojectionimage,thecalculatedobjectiveofthisimage,usethisimageasanewsourceofmaps,analysisofthenextframeimages,repeatthisprocessonthetargetcanbeachievedonacontinuoustrack.Inthispaper,themainworksareasfollows:Themainresearchcontentisbasedonthemeanshiftalgorithmtotrackmovingtargets,andusingthisfollow-upoutofboundspenaltyappliestotennis.Trackmovingtargetsarewidelyusedinsurveillance,videocoding,aswellasthefieldofmilitaryindustry.Howtoeffectivetrackinginthetargetsofinterestvideosequences,thatisachallengingtaskincomputervision,whichisthemostimportantaspectofselectingsuitabletargetquicksearchfeatures,aswellasmatchingalgorithm,Mean-Shiftasanefficientpattern-matchingalgorithm,asaresultofwhichdoesnotrequireexhaustivesearch,hasbeensuccessfullyusedinreal-timerequirementsofthehigherareasoftargettracking.Detectionofmotionpicture,anotonlyinindustrialandmilitaryapplicationisextremelyimportant,butalsoinsomeimportantplacessuchasbanks,publictransportsystemsarealsowidelyused.ThebasicprinciplesofthegoalaretotrackthecharacteristicsofthecolorIVinformation.Usingtheinformationcalculationprocessingtothenextframeaftertheprojectionimage,thecalculatedobjectiveofthisimage,usethisimageasanewsourceofmaps,analysisofthenextframeimages,repeatthisprocessonthetargetcanbeachievedonacontinuoustrack.Inthispaper,themainworksareasfollows:1.Meanshiftalgorithmusedfortrackingmovingobjects,accordingtothecolorhistograminformation,willbecalculatedafterthesemessagestothenextframeimageprojection,usethisimageasanewsourceofmaps,Analysisofthenextframeimages,repeatthisprocessonthetargetcanbeachievedonacontinuoustrack.2.Thetraditionalmeanshiftalgorithmforfastmotioninthetarget,theregionsusceptibletobackgroundinterference,cannotaccuratelytrackmovingtargetsintheregion.Thispaperpresentsanoptimizationbasedonthebackgroundofthemeanshiftobjecttrackingalgorithm,throughtheintroductionofmixed-histogramandhistogramforquantitativere-optimization,andthereductionofthebackgroundpixelsintheprobabilitydensityfunctioninthecontextofweighttolowerthetargetareasonthetrackofandthustoachieveaccuratetrackingofmovingtargets3.Mean-shiftalgorithmwillbeappliedtoRGBimagesusingthehistograminthedirectionofthetargetmodel-buildingstrategy,theRGBimagesinordertoMean-shiftatthecoreofthetargettrackingalgorithm,experimentalresultsshowthatthealgorithmhasnotaffectedbytheadvantagesoflight,inthecaseoflowcontrastcanstillachieveastable,real-timetrackingoftargets,thecomplexityofthecontextoftargetdetectionasthekeytechnologyofthepilotplaysanimportantrole,andthecombinationofinter-frameinformationmovingobjectextractionresearchprogramsThisarticleisdividedintosixparts:thefirstpartofthefirsttopicontheresearchbackground,andthesignificanceofdevelopmentsinthefieldofthestatusquo;meanshiftintroducedthesecondpartoftheclassicalalgorithm;thethirdsectiondescribesastraightlinehoughtransformtodetectthethoughtsandprinciples,PartIVmeanshiftalgorithmusedas