Author: Mirahmad Chabok
Chabok, Mirahmad, 2019 Seeing through the smoke: using airborne or ground-based consumer-grade DSLR cameras in fire monitoring, Flinders University, College of Science and Engineering
Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact copyright@flinders.edu.au with the details.
In this research, a simple and novel solution is designed to reduce the smoke and haze effect on imagery acquired using consumer grade Digital Single Lens Reflex (DSLR) cameras that, after modification, might be mounted on unmanned aerial vehicles, manned aircraft or be captured from the ground. The presence of substantial amounts of haze or gaseous smoke particles, caused for example by an active bushfire at the time of data capture, dramatically reduces image visibility and quality due to electromagnetic radiation being scattered by atmospheric particulates. Although most modern hyperspectral and thermal imaging sensors can capture a large number of narrow range bands of the shortwave and thermal infrared spectral range, and therefore have the ability to penetrate smoke and haze well, the resulting images do not contain sufficient spatial detail to enable important objects to be located or to assist in search and rescue.
First, this research investigates the limitations of existing methods for penetrating smoke, and image visibility enhancement techniques. For example, many recently developed algorithms exploit the differences between two or more images of the same scene that have different properties. These approaches require exact co-registration of the images captured using multiple sensors which is a non-trivial step in the processing workflow and therefore are not applicable to dynamic scenes in airborne applications where ground objects and smoke particles can move quickly relative to each other.
Secondly, it examines multispectral, thermal and visible wavelength images collected over bushfire smoke to simulate and evaluate the main hypothesis of this research, i.e., the feasibility of using DSLR camera sensors for smoke penetration. The results from this simulation phase confirmed the research hypothesis and helped form the future research design methodology and structure.
Thirdly, based on achieved feasibility confirmation and results from previous sections the research investigates and verifies the spectral sensitivity of unmodified and modified DSLR camera sensors by taking measurements in a controlled laboratory environment with a spectroradiometer. The results show that DSLR camera sensors, after modification, can collect enough information between 700 nm – 1100 nm (the nearinfrared) that the resulting image data will, theoretically, be less affected by smoke and haze based on Rayleigh scattering theory.
Fourthly, an optical filter was designed to preserve the semi-natural colour appearance of the captured photographs while collecting near infrared radiation simultaneously with visible blue and green light. A custom-designed filter was manufactured so that it would allow visible blue, visible green and infrared radiation between 950 nm and 1100 nm to pass. Its spectral responsivity was verified and confirmed using laboratory spectroradiometry. The custom-designed filter enabled the camera to collect these three bands simultaneously without any co-registration requirement or any major negative effects on the visual appearance of the captured imagery.
Finally, a smoke penetration algorithm (SPA) was developed to enable effective scene visibility enhancement by eliminating or minimizing the smoke effect from collected data. This modified camera and custom-designed filter were used to collect data on the ground and on an aircraft flown over areas covered by smoke and results processed using the SPA. The final processed results as an outcome of SPA, show significant improvements in visibility confirming the efficiency of the proposed method.
Keywords: smoke penetration, remote sensing, photogrammetry, optical filter, bushfire mapping
Subject: Engineering thesis
Thesis type: Doctor of Philosophy
Completed: 2019
School: College of Science and Engineering
Supervisor: Andrew Millington