This paper provides high- precision image recognition service for mobile platforms. It analyzes and researches on SIFT and BRISK algorithm, and proposes a novel, efficient, lightweight adaptive clustering algorithm for image recognition which is suitable for Android. It designs a high-precision image recognition system based on the Android platform, using a variety of resources of Android to develop image recognition software. The results show that the system hardware is simple, low-priced, reliable, easy to be used and extended.