Avaliação do algorítimo SVM na detecção de comportamentos suspeitos em cenas de vídeo

Abstract

The parameters optimization in the SVM (Support Vector Machines) algorithm, allows to extract optimal results in the ML (Machine Learning) context. The Support Vector Machines take place between the ML algorithms that have received great interest since their introduction in the ML community. However, to achieve an optimal result it has to be studied along with it’s parameters optimization. This work performed experiments , using atual tecnics with the purpose of achieving optimal rates of suspect moviments recognition on vídeo scenes. Through this experiments have been proved that, value atribuition parameter selection method can provide similar results as the automatized method can provide. The studiedmethods was applied and also the empirical methods, each one over an people moviments on video scenes data base. The variations of the SVM are discussed, the each step results achieved and the works being done with the SVM algorithm recently in ML. The result was able to be compared with those achievedin previous works