Theimmensecrimeratesresultingfromusingpistolshaveledgovernmentstoseeksolutionstodeal
withsuchterroristincidents.Theseincidentshaveanegativeimpactonpublicsecurityandcause
panicamongcitizens.Fromthispoint,facingapandemicofweaponviolencehasbecomeanimportant
researchtopic.Onewaytoreducethiskindofviolenceistopreventitviaremotedetectionandtogive
anappropriateresponseinashorttime.Videosurveillanceistheprocessofmonitoringthebehavior
ofpeopleandobjects.Surveillancesystemscanbeemployedinsecurityapplicationsaslegalevidence.
Moreover,itisusedwidelyinsuspiciousactivitydetectionapplications.Intelligentvideosurveillance
systems(IVSSs)aretheuseofautomaticvideoanalyticstoenhancetheeffectivenessoftraditional
surveillancesystems.WiththerapiddevelopmentinDeepLearning(DL),itisnowwidelyusedto
addresstheproblemsexistingintraditionaldetectiontechniques.Inthisarticle,anapproachtodetect
pistolsandgunsinvideosurveillancesystemsisproposed.Thepresentedapproachdoesnotneed
anyinvasivetoolsintheweapondetectionprocess.ItusesDLintheclassificationandthedetection
processes.TheproposedapproachenhancestheobtainedresultsbyapplyingTransferLearning(TL).
ItemploystwodifferentDLtechniques:AlexNetandGoogLeNet.Experimentalresultsverifythe
adaptabilityofdetectingdifferenttypesofpistolsandguns.Theexperimentswereconductedona
benchmarkgundatabasecalledInternetMovieFirearmsDatabase(IMFDB).Theresultsobtained
suggestthattheproposedapproachispromisingandoutperformsitscounterparts. |
