Field trials for a rapid deployment lidarbased high temporal-frequency change detection monitoring system

Abstract

Remote sensing technologies are now prevalent in geohazard monitoring applications. Laser scanners (LiDAR), photogrammetry, and satellite-based topographical measurements can be used to map various geomorphologies, natural terrains, and man-made infrastructures. Change detection analysis with topographical data sets captured at a certain frequency can denote failure precursors, and active landslide events, and analyze events a-posteriori. Change detection in time is often conducted at high resolution and low temporal frequency. Dense data sets considerably reduce the inherent uncertainty arising from superficial roughness and statistical spread between two series of surveys. Long-term, high-resolution monitoring thus allows for precise measurements and the detection of relatively low change detection thresholds. Such an application is not readily applicable to the detection of rapidly progressing events. When statistical outliers occur such as intense rains, or earthquakes, landslides may be triggered and not captured within the time frame required to act and mitigate. The present study investigates low-resolution, high-temporal frequency measurements as an option for capturing rapidly evolving events. In this work, a low-cost fixed laser scanner was connected to a small Raspberry Pi to collect point cloud frames at hourly intervals. The montage consisted of the lidar-computer combination mounted on a tripod anchored to the ground. This paper describes the montage, material required to assemble, and installation protocol. Special considerations were given for rapid deployment with limited labor. To evaluate the monitoring setup, two experiments were done. One was that an artificial pile of humid less than 4 mm material was assembled in a local quarry. The pile was partially excavated at the toe to achieve a progressive failure as the material dried up. Monitoring of the pile was performed for approximately 3 weeks with hourly readings. A high-resolution scan was taken at the onset of the test for reference. The second experiment consisted of analyzing a floodplain in Grizzly Creek, Yukon for 10 months during the winter season, with six readings a day. Snow developed across the months and melted near the end of the period. A high-resolution scan was taken at the onset of both tests for reference. The results obtained are analyzed and discussed in the thesis. A discussion is provided on uncertainty associated with topographical measurements collected during this experiment.

Date	16 Apr 2025
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Jonathan D. Aubertin (Supervisor)