The collision accidents between construction equipment and workers have increased over the past five years, and the main reason for collision accidents is the poor visibility of workers. The purpose of this thesis is to estimate the location of construction equipment and workers on the construction site and to determine whether workers approach construction equipment, to improve visibility. The YOLO-v5 model, a deep-learning-based object detection model, is applied to CCTV videos installed at construction sites to detect the bounding box and type surrounding each object for all objects in each CCTV video frame. The surface of the workplace is flat enough to be regarded as a plane, and a homography matrix is obtained by using the surface coordinates of four or more marking points installed at the construction site and the location and size of the surface of the construction equipment are estimated. It is achieved to increase safety by estimating the ground positions of objects set as workers to determine whether to enter the danger area and to determine the danger of an accident.