Research on Recognition Positioning and Grasping of Manipulator Based on Improved SURF Algorithm
Aiming at the problems of slow running speed and low matching accuracy when traditional SURF algorithm is used to identify objects in the working area of the manipulator,an improved SURF algo-rithm based recognition,positioning and grasping method for the manipulator is proposed.First,the tradi-tional SURF algorithm was used to extract the scale invariant feature points of the target.Then,rBRIEF de-scriptors and low variance filtering were used to describe and reduce the dimension of the extracted fea-tures.Then,2-NN feature points were cross-matched based on Hamming distance,and the wrong matching pairs were removed by RANSAC algorithm.Finally,three-dimensional coordinates of the target were ob-tained by binocular stereo vision 3D reconstruction,and each joint driving Angle obtained by inverse kine-matics of the manipulator was input into the control system to drive the steering gear to drive the end-effec-tor of the manipulator to grasp the target.The experimental results show that,compared with SIFT and tra-ditional SURF algorithms,the improved SURF algorithm can improve the recognition speed and matching accuracy of target feature points,and effectively improve the real-time and accurate grasp of the robot arm.
NETL
NSTL
万方数据
维普
