Otomatize Guvenlik Sistemlerinde OlayAnalizi EventDetection inAutomated Surveillance Systems
B. Birant Orten, A. Aydın Alatan · 2006
AX) 9(XXk) (1) Eventrecognition isprobably theultimate purpose ofanautomated N surveillance system. Inthis paper, hidden Markovmodels (HMM) e,itligi kullanilmaktadir. Burada{x1, X2,, x}kulmesi birpikselin areutilized torecognize thenature ofanevent occurring inascene. zaman i,erisinde gozlemlenen degerlerinden olu*maktadir. Forthis purpose, object trajectories, whichareobtained through a E,itlikteki (p(.) fonksiyonu isep(x)'in hesaplanmasi sirasinda herbir successful track, arewritten asa sequence offlowvectors that ornekdegerin katkisi ileilgili bilgi veren pencere fonksiyonudur. contain instantaneous velocity andlocation information. These Pencere fonksiyonu Gaussolarak se,ildigi takdirde (1)dekie,itlik vectors areclustered byK-meansalgorithm toobtain aprototypea,agidaki ,ekilde olacaktir: representation. HMMs aretrained withsequences obtained from usual motion patterns andabnormality isdetected bymeasuring