Color Space Model with Chan-Vese Algorithm for Fire Detection in Videos

Abstract

Detecting the breakout of fire rapidly is vital for the prevention of material damage and human casualties. The vision-based flame detection has drawn significant attention in the past decade with camera surveillance systems becoming quite common. Conventional fire detectors use physical sensors to detect fire. However, this may lead to false alarms. In order to prevent false alarms, a computer vision-based fire detection algorithm is developed. In this paper, a new method for identifying fire is proposed .Firstly the RGB image is converted to Lab color space and Optical flow estimation computes correspondence between pixels in the current and the previous frame of an image sequence to detect the moving pixels. Secondly Chan-Vese model is applied for segmentation. Segmentation is the process of partitioning a digital image into multiple segments; Chan-Vese model for active contours is a powerful and flexible method to segment images and here the fire region is segmented. And then a novel fire color model is developed in CIE Lab color space to identify fire pixels. Experimental results show the proposed approach can classify flame and non flame objects, and also has a high time effectiveness.