COMPARISON OF FEATURE EXTRACTION METHODS FOR INSULATOR DEFECT DETECTION

ABSTRACT

Power-line insulators plays key role in the generation, transmission, and distribution of power in the national grid. They support electrical conductors, while preventing electric current from flowing across them. Therefore, the failure of power-line insulators can cause great damage to the power grid. Hence, it is very important to monitor the state of power-line insulators before its failure results to a great damage in the power grid. Computer vision has lately been recognized as a means to solve this challenge safely, speedily and accurately instead of the manual method of monitoring that has always been very dangerous and unsafe to humans. This work compares three feature extraction methods; Local Binary Pattern (LBP), Grey Level Co-occurrence Matrix (GLCM) and Local Directional Pattern (LDP), used in the detection of defects in power-line insulators. The feature extraction methods are used to extract features from the power-line insulators and then fed into a classifier, K-Nearest Neighbors (KNN), which then classify the insulators as defective or non-defective insulators. The results of the three feature extraction methods are then compared to determine the method with improved accuracy. Experiments performed in this research work showed that LBP had a higher accuracy of 96.5% compared to GLCM with an accuracy of 80.4% and LDP with an accuracy of 70.6% when they were all used as a feature extraction methods with K–Nearest Neighbors as classifier.