In recent years, 3D models that imitate the real environment have been used in various fields, such as work support at nuclear power plants and disaster experiences using VR. One of the methods to create such 3D models is the 3D reconstruction method which uses images taken by a camera. In most cases, this method estimates the camera pose, including the 3D position and orientation of the camera at the time the image was taken by image processing. Then, a 3D model is created based on the estimated camera pose. However, since estimating the camera pose of an image with few features is difficult, regions with few features, such as a monochromatic flat wall, are missing from the 3D model created. Therefore, this study aims to develop a 3D reconstruction method without missing regions with few features by using the laws of the pendulum motion. In this method, the entire pendulum motion is estimated from a portion of the pendulum motion estimated by image processing. A 3D model is then created based on the estimated entire motion. The evaluation shows that the proposed method can reconstruct the environment which includes the area with few features.