Loop detection is an important and challenging problem in simultaneous localization and mapping (SLAM). At present, most researches have been focused on the visual position recognition, which is to identify overlapping positions through appearances. Therefore, a loop detection laser SLAM algorithm with global descriptor is proposed, which is used to improve the efficiency of loop detection, calculate the distance between two scanning contexts by the similarity score and detect loops effectively with a two-stage search algorithm. The algorithm is evaluated by testing on the public dataset KITTI, and the result shows that compared with the existing LeGO-LOAM algorithm, this kind of algorithm works better.