Abstract: Most existing approaches to global localization are based on extracting feature points and their descriptors from raw laser scans or occupancy grid maps, and can't extract features efficiently in some large indoor scenes with sparse environment features. For this problem, correlative scan matching (CSM) algorithm is applied to the global localization task. By calculating the contribution of points to pose solving, the CSM algorithm is improved in the point cloud downsampling part and in the angular step-size choosing part. Finally, comparative experiments on the algorithm before and after improvement are performed in both simulation environments and real environments. Results show that, compared with the original CSM algorithm, the localization success rate of the improved algorithm is increased by 1.5%, and the time consumption is reduced by 1.1 s, which proves that the improved algorithm in this paper is more suitable for large indoor scene than the original CSM algorithm.