SIFT-based algorithm for object matching and identification

Abstract

Object Recognition is the base and Guarantee of the robot moving object tracking system, the accuracy of the object recognition is directly effect the evaluation of the tracking system. an algorithm based on combining the SIFT and Kalman filter is proposed, using the SIFT to match the feature points, through Kalman filter algorithm to got the smallest effect of noise, and then simulate the algorithm, using sets of pictures with different types of noise of different intensity for object recognition, verify the high accuracy of the algorithm.