The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLII-3/W8
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-3/W8, 195–200, 2019
https://doi.org/10.5194/isprs-archives-XLII-3-W8-195-2019
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-3/W8, 195–200, 2019
https://doi.org/10.5194/isprs-archives-XLII-3-W8-195-2019

  21 Aug 2019

21 Aug 2019

USE OF MACHINE LEARNING TECHNIQUES FOR RAPID DETECTION, ASSESSMENT AND MAPPING THE IMPACT OF DISASTERS ON TRANSPORT INFRASTRUCTURE

P. M. Kikin1, A. A. Kolesnikov2, and A. M. Portnov3 P. M. Kikin et al.
  • 1Peter the Great St.Petersburg Polytechnic University (SPbPU), St.Petersburg, Russian Federation
  • 2Siberian State University of Geosystems and Technologies, Novosibirsk, Russian Federation
  • 3Moscow State University of Geodesy and Cartography,Moscow, Russian Federation

Keywords: Road Network, Machine Learning, Artificial Neural Networks, Unet, UAV, Rapid Mapping

Abstract. Road traffic infrastructure plays a key role in emergency management. It allows to evacuate people from the affected area in the shortest possible time, as well as to organize rapid emergency response. However, disasters often cause the destruction of roads, railways and pedestrian routes, which can significantly affect the evacuation plan and availability of facilities for emergency services, which increases the response time and thereby increases the losses. Therefore, it is very important to quickly provide emergency services with necessary post-disaster maps, created on the principles of rapid mapping. Change detection based on geospatial data before and after damage can make rapid and automatic assessment possible with reasonable accuracy and speed. This research proposes a new approach for detecting damage and detecting the state and availability of the road network based on the satellite imagery data, unmanned aerial vehicles (UAVs) and SAR using various methods of image analysis. We also provided an assessment of the resulting combined mathematical model based on neural networks and spatial analysis approaches.