Development of a Rockfall Database with Fixed Photogrammetry: Comparison of the Multi-epoch and Multi-imagery Method to Classical Structure-from-Motion Workflows and Evaluation of Rockfall Seasonality
摘要
Rockfall can cause significant hazards to human health and infrastructure or interrupt transportation. Remote monitoring methods, such as fixed location photogrammetric monitoring systems, allow long-term high-frequency monitoring, the precise determination of source location, timing and volume of rockfalls, and creation of high-temporal-resolution rockfall databases. In this study, a section of rock slope on Interstate 70 west, west of Idaho Springs, has been monitored at a daily frequency since March 2018 using a 5-camera photogrammetric system. The purpose of this study was to develop an optimal method to construct photogrammetric models for rockfall database generation and to analyze the resultant rockfall trends over a unique 5.5-year monitoring period. The Multi-Epoch & Multi-Imagery (MEMI) method, which collectively performs camera calibrations and constructs co-aligned models from different collection dates, was evaluated against classical SfM construction methods where models are constructed independently for each collection date. The MEMI method was found to be the optimal method for rockfall identification as it minimized distortion, provided more consistent change results with a consistently lower or equivalent limit of detection, and improved identification of rockfall clusters, with minimal cost to real model accuracy. Given these results, the MEMI workflow is the recommended approach for long-term SfM monitoring and rockfall database generation. Magnitude-cumulative frequency curves developed for the database were found to be significantly affected by infrequent, large outlier volume events. Inspection of the database found that rockfall at the slope follows an approximately seasonal cycle, with the majority of rockfall occurring in spring and decreasing over the remaining seasons with little rockfall occurring in winter. This trend is consistent with those found in studies of similar slopes.
Highlights A 5.5-year rockfall database was constructed using high-temporal-resolution photogrammetry allowing for highly accurate identification of rockfall occurrence. A comparison was performed between classical SfM methods of producing point clouds and the MEMI method for accuracy and ability to identify rockfalls. The MEMI method was shown to perform best in model comparison with little cost to real accuracy, and solved differential distortion issues present in some classically built point clouds with fixed camera calibrations The rockfall database was analyzed to characterize trends in rockfall present at the site over a ~ 5.5-year period.