Abstract
Timely clearing-up interventions are essential for effective recovery of flood-damaged housing, however, time-consuming door-to-door inspections for insurance purposes need to take place before major repairs can be done to adequately assess the losses caused by flooding. With the increased probability of flooding, there is a heightened need for rapid flood damage assessment methods. High resolution imagery captured by unmanned aerial vehicles (UAVs) offers an opportunity for accelerating the time needed for inspections, either through visual interpretation or automated image classification. In this study, object-oriented image segmentation coupled with tree-based classifiers was implemented on a 10 cm resolution RGB orthoimage, captured over the English town of Cockermouth a week after a flood triggered by storm Desmond, to automatically detect debris associated with damages predominantly to residential housing. Random forests algorithm achieved a good level of overall accuracy of 74%, with debris being correctly classified at the rate of 58%, and performing well for small debris (67%) and skips (64%). The method was successful at depicting brightly-colored debris, however, was prone to misclassifications with brightly-colored vehicles. Consequently, in the current stage, the methodology could be used to facilitate visual interpretation of UAV images. Methods to improve accuracy have been identified and discussed.
Highlights
The risk of occurrence of floods has risen globally, being driven by climatic, terrestrial and hydrological as well as socio-economic factors [1]
With the automated methods for flood damage detection from unmanned aerial vehicles (UAVs) imagery in their infancy, the overarching aim of this study is to develop a modelling framework for the automated detection of flood damage using UAV RGB imagery and machine learning (ML) methods coupled with object-oriented image analysis, allowing for rapid recognition of signs of flood damage
Our study provides a methodological framework for the application of treebased ML algorithms coupled with object-oriented image analysis in rapid post-event flood damage detection as well as insights into their capacity to distinguish between different types of debris
Summary
The risk of occurrence of floods has risen globally, being driven by climatic, terrestrial and hydrological as well as socio-economic factors [1]. In England, one in six properties is at risk of flooding from rivers and the sea [3], with the scales of economic costs to residential properties varying between 320–1500 million £ (2015 prices), depending on severity of flooding events [4]. Rapid assessment of damages for insurance purposes is vital for fast recovery, it may be delayed due to the necessity for door-to-door inspections to be carried out before cleaning up activities can take place. Such delays may cause further deterioration of the properties and increase overall insurance costs. They slow down the pace of recovery, contributing negatively to increasing urban resilience to flooding
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
