Fault detection of insulator based on saliency and adaptive morphology

We study the problem of glass insulator fault detection from image in this work. It is a challenging task as no reliable electromagnetism cues are available. Since the characteristics of insulator fault are not clear and the positions of the insulator fault are uncertain. Previous efforts have been focusing on insulator classification and the insulator location. Recently, there is mounting evidence that saliency detection are setting new records for various vision applications. On the other hand, considering the particularly structure of insulator, fault detection can be naturally enhanced by morphology problem. Therefore, we in this paper present a saliency and adaptive morphological based insulator fault detection algorithm, aiming to jointly explore the capacity of saliency and morphology. Specifically, we propose an adaptive learning scheme which learns the saliency and morphology in a unified adaptive framework. According to the experiment, we can process most of the circumstances with 92 % accuracy in 0.5 second on average, which suits the Unmanned Aerial Vehicle (UAV) patrol device well.