Abstract: Disclosed is a method for monitoring ground settlement based on computer vision. Before monitoring starts, the first image frame is captured. For one measuring point, the area of the top LED lamp is defined as a tracking template, its pixel center is the reference point for settlement calculation, and a monitoring area is defined by an estimated range. After monitoring starts, the best matched of the lamp template is searched for in the monitoring area of a second image frame. When the best matched area is obtained, its pixel center is obtained as the new lamp position, and it is selected as the new template; the pixel displacement between two adjacent image frames can be obtained by comparison. The total pixel displacement of multiple points during the monitoring period is calculated through the accumulated displacement, and the actual settlement is calculated through a pixel-physical ratio.

Abstract: Disclosed is a method for monitoring ground settlement based on computer vision. Before monitoring starts, the first image frame is captured. For one measuring point, the area of the top LED lamp is defined as a tracking template, its pixel center is the reference point for settlement calculation, and a monitoring area is defined by an estimated range. After monitoring starts, the best matched of the lamp template is searched for in the monitoring area of a second image frame. When the best matched area is obtained, its pixel center is obtained as the new lamp position, and it is selected as the new template; the pixel displacement between two adjacent image frames can be obtained by comparison. The total pixel displacement of multiple points during the monitoring period is calculated through the accumulated displacement, and the actual settlement is calculated through a pixel-physical ratio.