When a fisheye camera is used for measurement, accurate intrinsic parameters are often required. We therefore study methods for obtaining intrinsic parameters easily and accurately. Specifically, we propose a method in which a fisheye camera is rotated on a horizontal plane and its intrinsic parameters are estimated from the shape of feature-point trajectories in three-dimensional space.

When a fisheye camera is used for measurement, accurate intrinsic parameters are often required. We therefore study methods for obtaining intrinsic parameters easily and accurately. Specifically, we propose a method in which a fisheye camera is rotated on a horizontal plane and its intrinsic parameters are estimated from the shape of feature-point trajectories in three-dimensional space.

When measurement is performed using a stereo camera, the error included in the result depends on the accuracy of the camera calibration performed in advance. In this research, we propose a correction method that uses the disparity of feature points obtained from objects at known distances. We calculate the disparity error relative to the true value and use it to create a disparity offset map that represents the disparity error for all pixels. By applying this map as a correction to the stereo matching process, the accuracy of the fisheye stereo camera can be improved.

As pipes age, the risk of road cave-ins and other accidents increases. To prevent such accidents, it is effective to develop a system that can perform precise inspection from images captured inside pipes. In this research, we propose a method that creates unfolded views from images captured inside pipes with a wide-angle camera and then connects those unfolded views based on the motion information of the camera.

We reconstruct a three-dimensional environment from a monocular fisheye camera using spatiotemporal images, which are created by arranging multiple images in chronological order. Because fisheye cameras have an ultra-wide field of view, they can measure a wide area, which leads to lower cost and makes them suitable for installation on vehicles. 

In this research, a fisheye camera is mounted on a mobile robot to measure the surrounding environment. Three-dimensional reconstruction is then performed while taking into account the special distortion characteristics of the fisheye lens.

One system that helps drivers understand the surroundings of a vehicle presents an image that appears to show the vehicle from directly above. This is created by transforming images obtained from multiple fisheye cameras mounted on the vehicle into a top-view format and combining them to generate a full-surround view. To do this accurately, the positions and orientations of the fisheye cameras, that is, their external parameters, must be known precisely. In this research, calibration markers with known shapes are placed around the vehicle, and the external parameters of multiple fisheye cameras are estimated from the shapes of the markers in the images.