Pole and Crossarm Identification in Distribution Power Line Images

Society is getting more and more dependent of the energy industry. Therefore it is producers, distributors and consumers interest that the energy system is maintained accordingly. Nowadays, only 20% of a distribution grid is covered by preventive inspections. This work present an initial step in a system of preventively inspect a network using an Unmanned Aerial Vehicle that will acquire low altitude images of the distribution system network and an intelligent image analysis is used to look for irregularities in the network. Here, a method for identifying poles and cross arms using neural networks is presented. Also, measurements done in the image database allowed to define a somewhat non-conventional set of entries increasing the hit hate by more than 20%. The results justify, once more, the advantages of exploring the available knowledge around a problem.