Thermal condition monitoring of electrical installations based on infrared image analysis

Infrared imaging is a commonly used tool for monitoring and assessing the thermal condition of electrical installations for ensuring a reliable power supply. However, the conventional evaluation approach is quite time consuming as the image is analyzed manually by a qualified personnel. Therefore, this paper proposed a fast thermal anomaly detection and classification based on qualitative infrared image analysis. First, regions of interest (ROIs) are semi-automatically selected by employing normalized cross correlation (NCC) for finding similar objects in the image. Statistical features are extracted from each detected region and classified using multilayer perceptron (MLP) neural network for determining the thermal condition of electrical equipment. The overall accuracy obtained by the proposed method is approximately 95%, which is highly encouraging.