Authors: Andries Heyns*, University of Alabama, Kevin Curtin, University of Alabama, Warren du Plessis, University of Pretoria, Mike Kosch, EnviroVision Solutions, Gavin Hough, Envirovision Solutions
Topics: Spatial Analysis & Modeling, Geographic Information Science and Systems, Hazards, Risks, and Disasters
Keywords: Fire, detection, surveillance, MCLP, NSGA-II, GRASS, Gurobi
Session Type: Paper
Start / End Time: 8:00 AM / 9:40 AM
Room: Roosevelt 4, Marriott, Exhibition Level
Presentation File: No File Uploaded
Systems of specialised tower-mounted cameras provide an effective means of early forest fire detection. The towers from these systems are sited within vast feasible site placement regions with the aim of maximising system visibility of vulnerable areas. Historically, tower sites have been identified by foresters and locals with intimate knowledge of the terrain and without the aid of computational optimisation tools. However, when moving into vast new territories and without the aid of local knowledge this process becomes cumbersome and daunting. Common terrain features at existing tower sites may, however, provide a guideline for semi-automated candidate site identification. Terrain feature classes - determined using the GRASS geomorphon terrain classification function - at 165 towers from systems in South Africa, Canada and the USA are analysed along with sites proposed by a heuristic optimisation approach. Ridges and peaks are identified as the prevailing common site characteristics and are then exploited to minimise the feasible site search area of an area in South Africa currently monitored by the ForestWatch detection system. Using the reduced search area, the system site selection problem is solved anew and compared to previous optimisation efforts. Furthermore, the number of candidate sites may now be reduced to such an extent that exact optimisation procedures may be used to find optimal system layouts. The results suggest that the approach will result in improved system detection and some components have already been used to determine new tower sites in South Africa.