Application of non-contact video quantitative measurement method in reservoir bank landslide monitoring

The geological structure of the reservoir bank landslide is complex and intricate. After deformation and damage, it causes river blockage, surges, and loss of people’s lives and property, posing a huge threat. At present, in academia and engineering applications, a large number of techniques such as inclinometers, rain gauges, and surface GNSS deformation monitoring are still used for monitoring landslides on reservoir banks. This type of monitoring method has the problem of “point to surface” which can easily lead to missed detection and reporting in some areas, some disaster points are close to water and steep, difficult to reach, and equipment installation is difficult. This work designs and implements a non-contact video quantitative monitoring system for surface deformation of geological disaster. By constructing a deep learning neural network, deformation area recognition and displacement quantitative calculation are achieved; By obtaining continuous images for a long time, draw the surface displacement-time curve, and output the surface deformation data and landform changes of the disaster. Meanwhile, this work explores the impact of different lighting conditions on the recognition results of target areas. This work can provide non-contact monitoring methods and dynamic warning support for large-scale monitoring of geological disasters.