Gaussian mixture model and color separation models for the Arc-flashover detection

Abstract

This paper proposes a real-time, early warning technique for arc-flashover detection in different industrial applications. In the proposed scheme the camera system installed in the industrial areas monitors the indoor environment and issues an alarm as a part of the early warning mechanism. The ultimate goal of this technique is to provide an alarm at an early stage before the explosion becomes uncontrollable. The methods also provide additional information about the location and intensity of the explosion/fire. The proposed algorithm uses RGB, YCbCr, and HSI color space for image analysis. A Gaussian Mixture Model (GMM) based foreground detection mechanism has been used to detect the moving objects and sudden changes in the video frames. Finally, appropriate thresholding parameters are used to detect the flashover. The performance of the system is tested for the live video feed. The method can detect the flashover within 0.102 seconds of its inception on an average, which seems very promising. © 2020 IEEE.