Fisheye cameras offer a wider field of view compared to traditional pinhole cameras. This paper presents a pose estimation method for fisheye cameras based on the EPnP algorithm. A unified projection model is used to analyze the imaging process, and it is shown to be applicable to fisheye cameras. A virtual pinhole camera is constructed, where the image captured by the fisheye camera is mapped onto the virtual camera’s image. The resulting virtual 2D reference points are then used as input to the EPnP algorithm for estimating the fisheye camera’s pose. The mapping error from the fisheye image to the virtual camera image is analyzed, revealing that the process does not amplify pixel error when extracting 2D reference points. It is recommended to select points near the image’s central focus as 2D reference points. Experimental results demonstrate that the proposed method is both reliable and efficient.