Extracting Topic-Related Photos in Density-Based Spatiotemporal Analysis System for Enhancing Situation Awareness

Recently, people have begun to diligently post their situation, particularly during a crisis, on social media, therefore, the enhancement of situation awareness using social data is one of the most attractive research subjects. In this paper, we propose a novel density-based spatiotemporal system with a photo image classifier. The photo image classifier, which allows the system to enhance situation awareness during a crisis by showing accurate topic-related photos, is integrated using a support vector machine (SVM) based on the Bag-of-Features (BoF) model into the conventional density-based spatiotemporal system. To evaluate the proposed system, we used an actual data set related to a weather topic, "rain," in Japan. The experimental results indicate that the proposed system can extract photo images related to the weather topic "rain" with high accuracy and recall levels.