Video Based Smoke Detection Algorithm Using Image Processing Techniques

Abstract

In this paper, we present an image based algorithm for detecting smoke from the videos of forest fire. Proposed algorithm consists of following steps for features extraction for detecting smoke. A live video is acquired and chopped with size of 120 frames. The first frame of video is set as reference frame and image difference is calculated for current frame using background extraction technique. Potential objects for smoke in a frame are extracted using the calculated difference image. All required parameters like grey pixel, area change and position change are calculated for every object identified. Final decision to declare segment/object as smoke or non-smoke is carried out after evaluating parameters extracted for an object. The proposed technique is tested with 5 different videos, and results have shown 99% accuracy in detecting the fire smoke with no false alarms.