The article presents the concept of how to use convolutional networks as a method for processing digital images acquired in visible region of light for the needs of smoke detection in large open area. The meaning and consequences of massive blaze were underlined on the basis of statistical data concerning the forest fires. The proposal to overcome the difficulties in using traditional methods for detection of fire threat by image processing techniques was discussed. The idea, inner structure and properties of a convolutional neural network as a tool for automatic feature generation and image recognition were presented. The algorithms of data processing used in vision systems for fire detection were analyzed including the solutions implementing the networks. On the basis of the analysis the proposal to develop a neural network for smoke detection with the use of the strategy called transfer learning was presented. Using the image base of fires available on the web, the quantified assessment of the proposed approach was conducted. In the research the AlexNet framework was adopted to recognize smoke in images. The processing of the net was illustrated with examples of activations of selected layers when fed with images containing smoke. The 99% sensitivity reached by the proposed processing together with the 1% of false alarm rate seems to be very promising for the system of fire surveillance based on watchtowers or air vessels monitoring large open areas.

Date Published: 27 March 2019
PDF: 11 pages
Proc. SPIE 11055, XII Conference on Reconnaissance and Electronic Warfare Systems, 110550F (27 March 2019); doi: 10.1117/12.2524560

Published in SPIE Proceedings Vol. 11055:
XII Conference on Reconnaissance and Electronic Warfare Systems
Piotr Kaniewski, Editor(s)