A Digitized Fuel Load Surveying Methodology Using Machine Vision

Abstract

Fuel load is a crucial parameter for evaluating design fires for buildings. However, the availability of fuel load data is currently hindered by the lack of an efficient on-site data collection method. This research develops a new methodology for fuel load surveys that can facilitate the collection, storage, and analysis of fuel load data for a variety of building occupancies. The new survey method harnesses recent developments in mobile electronic devices, cloud storage, and machine vision to efficiently complete fuel load surveys in buildings. A four-step method is developed comprising digital inventory, data organization, item matching through computer vision, and fuel load estimation. The method is completed through an interactive electronic surveying form accessible on any mobile device with an internet connection, such as a tablet. The application allows taking digital measurements and pictures of the room and content, which are stored, and later searched through retail search engines using image recognition. Automatic matching of the picture with an online catalogue item gives access to further information about this item which are then used to evaluate the item fuel load, using a table of materials calorific values coded in the application. A wide adoption of this method could provide benefits by progressively populating such digital fuel load database. The results can then be used to provide design guidelines for fuel load density in codes and standards, for application in performance-based design of structures in fire.