摘要： Abstract Input: A query image Source: A point cloud reconstruction of a large scene (有一百多万3D点) Result：pose 关键：an efficient and effective search method 阅读全文