At present

the conventional image recognition algorithm cannot be well applied to mobile platforms due to their large computational cost and high storage requirements.In this paper

a novel image recognition algorithm for smartphone is proposed:Firstly

BRISK detection algorithm and low-byte FREAK descriptor are used to extract image feature and represent image respectively

which solves the problems of long detection time and large memory footprint;Secondly

BRISK features are improved to accomplish the distinction between the similar structure features by adding the gravity information of smartphone to the image feature;Finally

multi-level index of descriptors have been established to achieve fast finding on similar descriptors which solve the matching problem.The experimental results show that the proposed algorithm can be effectively run on the resource-constrained general smartphone to achieve scene recognition in real-time.