Landslide monitoring using mobile device and cloud-based photogrammetry

Abstract

Landslides are one of the most commonly occurring natural disasters that can cause a serious threat to human life and society, in addition to significant economic loss. Investigation and monitoring of landslides are important tasks in geotechnical engineering in order to mitigate the hazards created by such phenomena. However, current geomatics approaches used for precise landslide monitoring are largely inappropriate for initial assessment by an engineer over small areas due to the labourintensive and costly methods often adopted. Therefore, the development of a costeffective landslide monitoring system for real-time on-site investigation is essential to aid initial geotechnical interpretation and assessment. In this research, close-range photogrammetric techniques using imagery from a mobile device camera (e.g. a modern smartphone) were investigated as a low-cost, non-contact monitoring approach to on-site landslide investigation. The developed system was implemented on a mobile platform with cloud computing technology to enable the potential for real-time processing. The system comprised the front-end service of a mobile application controlled by the operator and a back-end service employed for photogrammetric measurement and landslide monitoring analysis. In terms of the backend service, Structure-from-Motion (SfM) photogrammetry was implemented to provide fully-automated processing to offer user-friendliness to non-experts. This was integrated with developed functions that were used to enhance the processing performance and deliver appropriate photogrammetric results for assessing landslide deformations. In order to implement this system with a real-time response, the cloud-based system required data transfer using Internet services via a modern 4G/5G network. Furthermore, the relationship between the number of images and image size was investigated to optimize data processing. The potential of the developed system for monitoring landslides was investigated at two different real-world UK sites, comprising a natural earth-flow landslide and coastal cliff erosion. These investigations demonstrated that the cloud-based photogrammetric measurement system was capable of providing three-dimensional results to subdecimeter-level accuracy. The results of the initial assessments for on-site investigation could be effectively presented on the mobile device through visualisation and/or statistical quantification of the landslide changes at a local-scale.