Robot robust object recognition based on fast SURF feature matching

Abstract

The local invariant features SURF (Speeded Up Robust Features) is introduced into the robot visual recognition field to solve scale changes, rotation, perspective changes, changes in illumination and other problems. A Speeded up SURF (SSURF) algorithm is proposed to meet the needs of robot visual identification. In SSURF algorithms, the main direction determination step of SURF algorithm is modified which make the search scope of the main direction becomes {-α, +α} (0 ≤ α ≤ 30°) from the original scope 360 According to compressed sensing ideas and interest points distribution histogram, the main scale search space is selected to improve the interest points searching step of SURF algorithm, so the interest points searching time-consuming is reduced. Matching the sample object and the scene using SSURF descriptor, and positioning the target position in the scene and giving ROI(region of interest). Experimental results in the autonomous mobile robot platform show that the proposed method significantly improves the speed of the robot to identify the target object, and proved robust to the scale changes, rotation, perspective changes, changes in illumination.