ESTUDO DE PREVENÇÃO DE POSSÍVEIS INCÊNDIO UTILIZANDO ANÁLISE DE IMAGENS E INTELIGÊNCIA ARTIFICIAL

Abstract

Recently, cases of fires and incidents involving or not victims and which have grown alarmingly with the increasing increase in these occurrences, studies on the subject in question were developed in order to minimize the occurrences. From studies, research and varied experiments using mathematical calculations and the use of computational tools with diverse resources such as neural networks, Fuzzy Logic and Gaussian methods being processed with resources in the Matlab tollbox together in its applications to assist in the detection and tracking of fire in environments that have video monitoring. Study that determines improvements in the process of analysis of these images where the structure and the qualification process in the resources of color channels that is represented with level of adjunct signal with new algorithms, with the protocol of RGB standard and those of technological advances that consists of growth involvement in the computational treatment process. The video was obtained using a digital camera in HD resolution. Flame detection was performed with a threshold determination to separate the background from the image and the flames, which were diagnosed using the Gaussian mix method, among others, specifying the probabilities and decreasing the noise in the detection of image alteration applied to the tracking parameter. flames and foreground detection. It demonstrated effectiveness in the purpose that the study is oriented. From unpredictable perspectives, this algorithm will always use the components to which it has access, requiring discernment in the absence of external clues.