Slope surface deformation monitoring by close-range terrestrial photogrammetry

Landslide monitoring is an important means to prevent the landslide disaster. Among all elements of landslide monitoring, slope surface deformation is a piece of direct evidence to judge whether slope slips, which makes it indispensable in qualitative and quantitative analysis of slope stability. Current mainstream surface monitoring methods using GNSS are difficult to lay out densely on a large scale in a deformation region due to the high cost of equipment, leading to few surface points available for detection. With the rapid development of camera resolution and image processing, photogrammetry based on computer vision has great prospects in the application of slope real-time monitoring.

We introduce a low-cost landslide visual monitoring system using close-range terrestrial photogrammetry that deploys fixed cameras to capture the slope surface periodically and calculating the displacement of feature points from sequential slope images to generate the slope surface deformation network. A new machine learning framework is proposed to achieve image recognition, camera calibration and distance mapping altogether. We conduct indoor landslide experiments which verify the high precision, accuracy, and stability of our system.