Fisheye image unfolding based on spherical perspective projection model

This paper proposes a fisheye unfolding method based on spherical perspective projection model. First, an improved scan line approximation method is used to extract the effective area of the fisheye image, in order to determine the center and radius of the fisheye image, then introduces an improved latitude and longitude mapping method based on the spherical projection model, in order to obtain a global unfolding image. During the mapping from the world coordinate system to the spherical latitude and longitude coordinate system, the imaging plane is translated from the initial equator to the north pole to the plane where a certain latitude circle is located, which can effectively improve the image quality of the upper edge. Experimental results show that the method can obtain a rectangular image conforming to the observation habits of the human eye without loss of edge information, and it can also weaken the interference of the inherent distortion of fisheye lens on the behavior representation and recognition, thus meeting real-time requirements.