2019/01/10 by Kashif Ahmad, Ahmad, Kashif, Konstantin Pogorelov +11
Computer Science · Engineering · Environmental Science · #Anomaly Detection Techniques and Applications #Automated Road and Building Extraction #FOS: Computer and information sciences #Flood Risk Assessment and Management #Information Retrieval (cs.IR)
paper · pdf · doi:10.48550/arxiv.1901.03298
openalex publication_date 2019/01/10 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
This paper addresses the problem of floods classification and floods\naftermath detection utilizing both social media and satellite imagery.\nAutomatic detection of disasters such as floods is still a very challenging\ntask. The focus lies on identifying passable routes or roads during floods. Two\nnovel solutions are presented, which were developed for two corresponding tasks\nat the MediaEval 2018 benchmarking challenge. The tasks are (i) identification\nof images providing evidence for road passability and (ii) differentiation and\ndetection of passable and non-passable roads in images from two complementary\nsources of information. For the first challenge, we mainly rely on object and\nscene-level features extracted through multiple deep models pre-trained on the\nImageNet and Places datasets. The object and scene-level features are then\ncombined using early, late and double fusion techniques. To identify whether or\nnot it is possible for a vehicle to pass a road in satellite images, we rely on\nConvolutional Neural Networks and a transfer learning-based classification\napproach. The evaluation of the proposed methods are carried out on the\nlarge-scale datasets provided for the benchmark competition. The results\ndemonstrate significant improvement in the performance over the recent\nstate-of-art approaches.\n