Visual location is one of the most important issues in robot visual navigation. The locating efficiency has a large impact on the performance of the visual navigation. A fast visual location method is proposed: BRISK is employed as the local invariant features, the scene model based on aggregate is described and the similarity of the scenes is measured by Sorensen coefficient to avoid complex computation for efficiency. The method has been verified by simulation and experiment with a humanoid robot in indoor environment. The effectiveness is 99% and the average measuring time is 003 s (33 fps). The result demonstrates that the method has good performance in terms of effectiveness, robustness and real time.