Computer Vision Based Measurement of Wildfire Smoke Dynamics (CROSBI ID 215985)
Prilog u časopisu | izvorni znanstveni rad | međunarodna recenzija
Podaci o odgovornosti
Bugarić, Marin ; Jakovčević, Toni ; Stipaničev, Darko
engleski
Computer Vision Based Measurement of Wildfire Smoke Dynamics
This article presents a novel method for measurement of wildfire smoke dynamics based on computer vision and augmented reality techniques. The aspect of smoke dynamics is an important feature in video smoke detection that could distinguish smoke from visually similar phenomena. However, most of the existing smoke detection systems are not capable of measuring the real-world size of the detected smoke regions. Using computer vision and GIS-based augmented reality, we measure the real dimensions of smoke plumes, and observe the change in size over time. The measurements are performed on offline video data with known camera parameters and location. The observed data is analyzed in order to create a classifier that could be used to eliminate certain categories of false alarms induced by phenomena with different dynamics than smoke. We carried out an offline evaluation where we measured the improvement in the detection process achieved using the proposed smoke dynamics characteristics. The results show a significant increase in algorithm performance, especially in terms of reducing false alarms rate. From this it follows that the proposed method for measurement of smoke dynamics could be used to improve existing smoke detection algorithms, or taken into account when designing new ones.
Image motion analysis; Computer vision; Computer aided analysis; Virtual reality; Pattern analysis
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
nije evidentirano
Podaci o izdanju
15 (1)
2015.
55-62
objavljeno
1582-7445
10.4316/AECE.2015.01008