Urban environments, such as streets, roads and buildings, always require maintenance and management to better use. In this sense, this paper presents an approach to recognition pole utility in street images mapped by Google Street View. Features such as color, texture and shape were investigated in order to find the best set of features to represent the objects of interest. The classification was performed using an Multilayer Perceptron trained with the Levenberg-Marquardt algorithm. Initial results show a higher accuracy by using RGB mode and texture properties as features to represent objects.
Text in Portuguese, Original Title: Reconhecimento De Postes Da Rede Eletrica A Partir De Imagens Do Google Street View.