A Multi-sensor IoT Node Integrated in an Innovative Decision Support System based on GIS Technology in Wildfire Context

Financiador

Organização

Publicações

Accurate Rural Road Network within an Integrated Frame-work of Tools for a Decision Support System in Wildfire Management

Publication . Lourenço, Miguel Alexandre Gonçalves; Oliveira, Luís; Oliveira, Henrique

This thesis proposes innovative tools to enhance wildfire management by integrating critical data and addressing limitations in current state-of-the-art systems. Existing Decision Support Systems often use fragmented tools to address different tasks that must be conducted in wildfire scenarios, limiting data integration (such as: rural road networks, several sensor data) and system performance. The research started by developing accurate rural road detection and extraction to overcome current state-of-the-art challenges, notably road occlusions and shadows cast by vegetation or man-made objects, as well as to better interpret roads made of different surface materials. To achieve this, two methods have been utilised over two custom aerial image datasets reflecting real-world scenarios: (i) the first with DeepLabV3+ for road detection, followed by the Zhang-Suen and Guo-Hall thinning algorithms for road extraction; (ii) the second using four U-Net-based architectures that were confronted, incorporating two post-processing techniques to improve rural road detection and extraction. A mobile application has been developed to aggregate data from several hardware sensors, namely atmospheric sensor data for real-time monitoring of firefighting assets and infrared sensor data for post-wildfire hotspot detection, to enhance decision-making in wildfire management. An additional module leverages smartphone sensors and elevation data to geolocate wildfire outbreaks. This application also incorporates a dynamic road network that utilises the proposed road detection and extraction methods and accounts for current road obstructions to optimise route calculations. In addition, a device was developed, integrating a module that provides real-time sensory data during wildfire events and another module that collects sensory data in the wildfire aftermath. Lastly, an interoperable web service has been developed to ensure data interoperability among all developed tools, strengthening situa- tional awareness and leading to well-informed decision-making.

2025Tese de doutoramento

Acesso aberto

Ver mais

Entidade financiadora

Fundação para a Ciência e a Tecnologia

Número da atribuição

2021.06054.BD