Q-rough sets, flicker modeling and unsupervised fire threat quantification from videos

Chakraborty, Debarati B.; Detani, Vinay; Parshv Jigneshkumar, Shah

Abstract

This article defines new methods for unsupervised fire region segmentation and fire threat detection from the RGB video stream. Here, an unsupervised approach has been developed to determine the threat associated with fire. With this method, the threat of fire can be quantified, and accordingly, an alarm can be generated in automated surveillance systems of indoor as well as outdoors. To overcome the requirement of any manual intervention/ labeled data, or to deal with the incomplete knowledge-base due to the absence of any prior information, rough approximations have been used here to approximate the fire region. A new concept, namely, Q- rough set, has been proposed here, where the inexactness of the rough approximation is being minimized by employing a Q-agents of reinforcement learning. Utility maximization of Q-learning has been used to minimize ambiguities in the rough approximations. The newly developed Q-rough set approximation has been applied then for fire region segmentation from video frames. Another challenge associated with fire threat measures is not getting misled by other fire-like regions, even when the true fire is not present in the scene. Here, a new method of classifying true fire flame to fire-like regions is developed by incorporating the unique fluctuating feature of fire flame and modeling it with a flicker identification model. The threat index of true fire flame over the input video stream has been defined in sync with the relative growth in the fire segments on the recent frames. All theories and indices defined here have been experimentally validated with different types of fire and non-fire videos, through demonstrations and comparisons, as superior to the state-of-the-art methods.

Citation

Chakraborty, D. B., Detani, V., & Parshv Jigneshkumar, S. (2022). Q-rough sets, flicker modeling and unsupervised fire threat quantification from videos. Displays, 72, Article 102140. https://doi.org/10.1016/j.displa.2021.102140

• Journal Article Type: Article
• Acceptance Date: Dec 16, 2021
• Online Publication Date: Jan 4, 2022
• Publication Date: Apr 1, 2022
• Deposit Date: Mar 13, 2024
• Journal: Displays
• Print ISSN: 0141-9382
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 72
• Article Number: 102140
• DOI: https://doi.org/10.1016/j.displa.2021.102140
• Public URL: https://hull-repository.worktribe.com/output/4588810