Real-time monitoring of a dam breach for landslides is a critical issue. The study employs the computer vision technology to detect slope displacement in an artificial earth dam. A computer vision-based displacement monitoring (CVDM) system is innovatively developed, which includes a microcomputer, HQ camera module and industrial camera, chessboards, and real-time warning system. Indoor experimental tests analyzed the standard deviation and errors of three-axis displacements with a 30-cm distance between a 20-cm square chessboard and industrial camera in the CVDM system. Average standard deviations and resolution of the three axes are 0.05 pixels and 0.01 cm in the 30-cm distance. The study set a chessboard on the artificial landslide dam of the Landao Creek in Nantou County, Taiwan, and conducted the dam breach test in the field. The distance between the CVDM instrument and a chessboard was 29.72 m. After the chessboard had subsided by 12.82 cm, it collapsed during the dam failure for the real-time monitoring. The CVDM system in the field can detect 8 frames of image recognition per second with 5G transmission to an IoT cloud platform. Finally, the CVDM system is a set of the AIoT, high resolution, cost efficiency, and energy savings for displacement monitoring for a dam.