2019/06/20 by Christos Kyrkou, Theocharis Theocharides, Kyrkou, Christos +1 · 4 citations
Computer Science · Engineering · #Advanced Image and Video Retrieval Techniques #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Robotics (cs.RO) #Robotics and Sensor-Based Localization #UAV Applications and Optimization
paper · pdf · doi:10.48550/arxiv.1906.08716
openalex publication_date 2019/06/20 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
Unmanned Aerial Vehicles (UAVs), equipped with camera sensors can facilitate\nenhanced situational awareness for many emergency response and disaster\nmanagement applications since they are capable of operating in remote and\ndifficult to access areas. In addition, by utilizing an embedded platform and\ndeep learning UAVs can autonomously monitor a disaster stricken area, analyze\nthe image in real-time and alert in the presence of various calamities such as\ncollapsed buildings, flood, or fire in order to faster mitigate their effects\non the environment and on human population. To this end, this paper focuses on\nthe automated aerial scene classification of disaster events from on-board a\nUAV. Specifically, a dedicated Aerial Image Database for Emergency Response\n(AIDER) applications is introduced and a comparative analysis of existing\napproaches is performed. Through this analysis a lightweight convolutional\nneural network (CNN) architecture is developed, capable of running efficiently\non an embedded platform achieving ~3x higher performance compared to existing\nmodels with minimal memory requirements with less than 2% accuracy drop\ncompared to the state-of-the-art. These preliminary results provide a solid\nbasis for further experimentation towards real-time aerial image classification\nfor emergency response applications using UAVs.\n