Defect detection and preventive maintenance prioritization of distribution cubicles by infrared statistical image processing

Major defects in distribution cubicles create hot spots which can be detected by infra-red cameras. Preventive maintenance teams use infra-red images for finding defects in distribution cubicles, because these defects can lead to electricity distribution interruption, fire accident and damage of other part of cubicle. Infra-red image analysis, which is called Thermovision, is one of the main branches of machine vision. Although thermovision has been being used for finding defects in electrical components, related infrared images are inspected only by empirical methods and few researchers who work on infra-red image processing have not used control charts for this purpose. Variety of components and configurations, lack of training data, high serial dependency and complex behaviour of thermal conductivity are challenging matter which should be overcome when a proper method is developed. Image monitoring, which was used in current research, is based on finding change in images by spatial control charts. Spatial control chart is a special control chart which is used for finding abnormality in image. In this research combination of spatial control chart and robust regression was used for defect detection in infrared images and its ability for this purpose was evaluated.