A Survey of Deep Learning Techniques Based On Drone Images For The Search And Rescue Of Victims From Collapsed Structures

The extent and frequency of the disasters occurring in the world have increased drastically and this trend is expected to continue. The major key player in disaster management is the collection of timely and accurate disaster damage status of the disaster affected regions. Even today disaster damage assessment is largely based on manual operations and data gathering. Automating the disaster assessment, mapping and response by combining information obtained from ground-based sources and aerial sources is a promising method. This work is a survey of the state of the art in automation of disaster damage assessment and utilization of data based on aerial sources for structural damage detection in collapsed buildings for extracting meaningful information to assist disaster rescuers. The various approaches for data processing from the data collected are studied for identifying effective methodologies. Deep learning methods are studied and key factors are to be discussed in the view point of drone images based collapsed building detection. The various approaches for feature learning, the challenges, bottlenecks based on various performance metrics are studied. The potential optimizations and solutions are extensively studied presented in this paper. Drone based deep learning processing is also provided prominence in the critical study for practical applications.