Description

Computational vision-based flame detection has drawn vital attention within the past decade with camera surveillance systems turning into omnipresent. Whereas several discriminating options, like color, shape, texture, etc., have been utilized within the literature, this paper proposes a collection of motion options supported motion estimators. The key plan consists of exploiting the distinction between the turbulent, fast, fire motion, and also the structured, rigid motion of different objects. Since classical optical flow strategies don’t model the characteristics of fire motion (e.g., non-smoothness of motion, non-constancy of intensity), 2 optical flow strategies square measure specifically designed for the fireplace detection task: optimum mass transport models fire with dynamic texture, whereas a data-driven optical flow theme models saturated flames. Then, characteristic options associated with the flow magnitudes and directions square measure computed from the flow fields to discriminate between fireplace and non-fire motion. The planned features square measure tested on an oversized video info to demonstrate their sensible utility. Moreover, a unique analysis methodology is planned by fireplace simulations that leave a controlled environment to research parameter influences, like flame saturation, abstraction resolution, frame rate, and random noise.