Visual Loop detection is an important method for finding similar places using image descriptors. However, because the usual cameras have a limited field of view, the information of the image that can be used as descriptors is also limited. To solve this problem, we propose to stitch the images using several pinhole cameras to configure it as a panoramic image and use it for visual loop closure detection to perform more robust loop matching. In addition, we demonstrate a proposed algorithm using our own dataset that is difficult to match in a general visual loop detection algorithm.