As a result of global warming, many natural catastrophes, including cyclones, floods, storms, and others, are becoming more intense. The effect of these disasters is only amplified by this increase in the intensity and density of the population, particularly around the coasts. It must be required to gather data about existing structures that have been relevant to natural hazard assessment and risk control to measure, reduce, & plan for the risk related to dangers in a location. The collection of architectural data at the regional or urban level appears to be a long and expensive undertaking. To facilitate regional hazard assessment, this study proposes a framework for advancing the dissemination of information and collection at the local level. In this system, several types of information were gathered from different sources and combined to create a semantic description of each structure in a metropolitan area. In particular, architectural information is extracted from road and satellite imagery in deep learning techniques. To handle the problem of high dimensionality, quantify uncertainties, and improve the data source, a new data mining technique was designed. Creating a structural inventory for cities using this methodology provides the necessary information to anticipate and simulate risks and disaster management.