Research Of Feature Points Fast Match Base On Mobile Devices

At present, augmented reality technology become more and more popular. With therapid development of hardware technology, Many mobile devices such as smart phones,Tablet PC, Their computing and storage capacity are increased substantially, and peoplerely on them heavy in the real life. So applying augmented reality technology on thetraditional PC to a mobile device will be up to be very promising. Compared with the PC,though mobile devices exist some shortcomings, That are computing and storage capacityrelatively low, unable to meet the augmented reality technology requirements in real-time.Traditional augmented reality technology are generally based on the plane, which hassignificant limitations on non-plane rotation for the mobile devices and real-time tracking.With regard to some problems such the mobile device augmented reality poorreal-time and limited storage capacity, the paper apply BRIEF descriptors to replace thetraditional SIFT descriptor, making the mobile device storage space is used effectively.According to traditional algorithm can only implementation in the plane, non-planerotation and real-time tracking for the mobile devices under great constraints, we makeuse of mobile devices’ gravity sensor and orientation sensor to obtain the rotate gestureand use gravity sensors data to determine the feature points descriptors dominant direction,and make it replace the steps of the algorithm itself calculate descriptors direction toreduce the algorithm time overhead. Moreover, the article also apply gravity sensor toSIFT-FAST algorithm, and the match performance and time consuming are improved.The experimental results show that after treating BRIEF as feature points descriptorsand regarding gravity direction from gravity sensor as feature point descriptors’ dominantdirection,we design algorithm again, which not only greatly increased the flexibility ofmobile augmented reality, but also effective solution to small memory space, lowcomputing power and poor real-time on the mobile devices, more importantly, the featurepoints match performance is greatly improved. The similarity, SIFT-FAST algorithmimproved can ensure effective match performance on the premise of reducing timeoverhead.