Image based assessment of windstorm damage / by Lakshminarasimhan V. Sampath.
Sampath, Lakshminarasimhan V
MetadataShow full item record
Short-term and long-term planning of emergency response and recovery has stressed the importance of analyzing damage caused to a locality by windstorms. The most commonly used method of damage assessment is time-consuming manual ground-truthing, which sometimes becomes infeasible due to accessibility and manpower problems. Image processing techniques offer a quicker, less expensive and easier alternative to this problem. This project uses aerial photographs taken before and after a tornado to provide rapid and reliable statistical descriptors for detection and estimation of damage extent to a locality. The pre-disaster and post-disaster images are registered to the best practical accuracy. Difference in perspective and presence of debris or rubble make it difficult to attain a pixel-to-pixel accuracy in registration. Two different stages of registration are implemented to register the images to the best practical accuracy. The pre-disaster image is then segmented, and the various features are identified and uniquely labeled. Damage extent to each of these features is then assessed using several statistical descriptors. Change impacted on an area due to a windstorm can be detected quickly and accurately using this method.