Image processing and deep learning approach to evaluate earthquake resistance of urban buildings

Abstract

The capital of Bangladesh, Dhaka, is one of the most densely populated cities in the world, sits atop the world’s largest river delta at close to sea level, which can trigger a massive earthquake resulting in death of millions of people. To minimizing such casualties, marking risky buildings can be an efficient approach as these buildings have more chance to collapse. In this paper, a new approach has been introduced on spotting these buildings by taking visual view through images and calculating the risk factors using Image Processing and Deep Learning. By following FEMA154 method of calculating risk factors of C3 type URM INF buildings, Image processing has been applied on the image data to get results and SVM has been run on the manual data of the risk factors. We used neural network model VGG16 and altered it to a newer version for beam detecting via images. The buildings have been shown in the map marking red or green to let people know about the vulnerability of these buildings.