Rockfall triggering mechanism analyzed from video using optical flow technique

This work presents an approach to identify the rockfall triggering mechanism from video employing Optical Flow Technique. The video was captured by phone camera on 3rd, October 2017 when the massive rockfall happened at a quarry in Le Locle Jura mountains, Switzerland. Time-series frames were extracted from the video and registered using SIFT (Scale-Invariant Feature Transform), kNN (k-nearest neighbor classification) and affine transformation algorithm, which efficiently eliminate the video jitters. After that, the transformation of pixels in the time-series image sequence and the correlation between adjacent frames are used to find the correspondence, so as to calculate the motion data of the object between adjacent frames by Optical Flow Technique. The instantaneous velocity of pixel movement of failure rock mass or debris on the video frames during rockfall dynamic behavior can be obtained. The basal failure surfaces and two main phases of the failure have been anlayzed for the rockfall triggering mechanism. The workflow proposed here can be applied in a slope disaster monitoring and early warning system to identify and track rockfall events effectively.