Detección de fallas en los aisladores de la línea de transmisión eléctrica bloque MDC-PBHI aplicando visión artificial.

Abstract

In the present project has been made the fault detection in the insulators of the electric transmission line in block 46 (Mauro Dávalos Cordero "MDC") and block 47 (Paraíso-Biguno-Huachito-Intracampos "PBHI"), using artificial vision algorithms. In the first instance the feature extractor SURF (Speeded-Up Robust Features) has been selected because it is a good descriptor for the detection of objects in conjunction with the Adaboost classifier ("Adaptive Boosting") to detect the isolator in the scene and then classify it as good or bad. Secondly, photographs of another electric transmission line have been taken which form the Database "samples" with 200 images, of which 70% is used for the training of the classification methods and 30% for test and thus be able to locate faults in insulators such as breakage, fissure and dirt. The classification accuracy of each algorithm is detailed below: Adaboost (62.31%), Bayesian (75%), SVM (78.13%) and Neural Networks (81.25%).