Paper Info
Title
Perimeter Detection of Burnt Rural Fire Regions
Abstract
Rural fires are dangerous, and even after a fire has been extinguished hidden hotspots can continue to burn under burnt foliage. To aid in the post fire investigation procedure this paper proposes a method to locate the perimeter of a burnt fire region on the ground from an aerial camera. This process involves a boundary detection application which applies texture boundary detection algorithms. The Variance Ridge Detector can detect texture-boundaries in real-time yet still results in a high quality output. The Variance Ridge Detector is able to run at 47 frames per second on 320 by 240 pixel images. This paper discusses the methods to optimise the boundary detector for the purpose of recognising and highlighting the perimeter of burnt fire regions. This implements a threshold filter to remove undesired texture boundaries and obtain a mask of the fire ground. The mask is merged with the target image to identify the perimeter of the burnt fire region. This method is performed in real time and reliably detects a burnt fire ground with an accuracy of 74%. These objective measurements show that the proposed method outperforms all investigated prior burnt fire ground detectors on either quality or speed. The development of this perimeter detector could lead to deployment onto embedded systems for more practical tasks. This could include autonomous robots via tracking of a perimeter.
Year
DOI
Venue
2014
10.1145/2683405.2683451
IVCNZ
Keywords
Field
DocType
algorithms,design,experimentation,edge and feature detection,pattern analysis,threshold filter,measurement,real-time boundary detectors,reliability,theory,burnt fire regions,performance,texture boundary
Computer vision,Fire investigation,Computer science,Hotspot (geology),Ridge,Perimeter,Pixel,Frame rate,Artificial intelligence,Robot,Detector
Conference
Citations 
PageRank 
References 
0
0.34
3
Authors
2
Name
Order
Citations
PageRank
Rikki Shimazaki100.34
Richard D. Green210.71