Annotation of Image Data From Disaster Reconnaissance

Abstract eng:
Post–event reconnaissance teams have a critical mission: to collect scientific data to be used to learn from disasters. Despite the large volumes of images that have been collected from past events, only a small portion of these are accessible to the public and archived with certain basic information such as date, event, or location. Thus, the capacity to access and facilitate reuse of these images based on the true visual contents on images is limited. Currently, there is no established image annotation method and formal ontology for describing visual contents in such images. This impedes the use of images for generating new knowledge, and large volumes of images remain largely unused for scientific research. To address this problem, a structured annotation method is proposed for images originating from earthquake reconnaissance. images of buildings are focused in this study, with the intention of supporting researchers focused on structural design and performance. Herein, an earthquake image ontology (EIO) is developed for formalized and structured descriptions of images. EIO consists of a large terminology and associated relationships, enabling rich descriptions of images based on their contents and the types of queries of interest to researchers. It is adequately extensible and flexible to successfully deal with a broader set of images in the future as needed. An image annotation tool assists human annotators as they choose appropriate terminologies and their relations in EIO. This facilitates image annotation and conversion of data into a searchable form. The capability and usefulness of the method is demonstrated using real-world images and their descriptions, based on a comprehensive examination of many publicly available reports, articles, and data repositories.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket