Computer vision–based sensors for the tilt monitoring of an underground structure in a landslide area

Abstract

This study explores an innovative instrument to monitor potential tilting for an underground structure in a landslide area using computer vision technology. The instrument was composited of the Raspberry Pi, digital cameras, chessboards, and OpenCV programming. Image recognition technology was employed for detection of an inclined angle relevant to pixel changes of the center point on a chessboard in a computer vision–based tiltmeter. For laboratory testing, the resolution and accuracy of the tiltmeter were 0.01 and 0.03°, respectively. For field testing, five computer vision tiltmeters were installed in a drainage tunnel that was constructed under the ground in 80-m depth within a landslide area to monitor potential tilting of the tunnel from January to October in 2018. Maximum change in inclined angles of the five computer vision tiltmeters was 0.04° in the period. Overall, the computer vision tiltmeter for the inclined monitoring of the drainage tunnel is more cost-effective, real-time, and IoT-based than traditional monitoring devices in the landslide area.