Research On Real-time Binocular Vision Positioning And Tracking Technology For Series Robot

Based on the binocular vision of the tandem robot vision servo system can directly access the target location information and more image features to achieve visual closed-loop control,is the typical structure of the visual servo.This article revolves around real-time,stability and accuracy.The subject is divided into four parts: the positioning of the target in the image,the positioning of the target in three dimensions,the visual servo control method,the target position prediction and the tracking,and the basic theories and methods involved are studied respectively.And through simulation and theory to evaluate the applicability of these methods to the subject,to explore the real-time binocular vision servo stabilization of the real-time method.(1)Use the target feature on the image to locate the target.Through the comparison of the principle and characteristics of the common point feature extraction algorithm,it is found that the Ransac optimized Orb algorithm has strong real-time and stability.(2)The mapping process from the image space to the three-dimensional space is analyzed,and the accuracy of the parallax is found to affect the accuracy of the three-dimensional positioning.Which leads to the ELAS stereo matching algorithm which is better than the traditional stereo matching algorithm.Aiming at the problem that the algorithm has high error matching rate,an improved ELAS stereo matching algorithm based on parallax continuity constraint is proposed,which reduces the error matching rate of the original algorithm by at least 15.63%.(3)Based on the eye-in-hand configuration of the 6-DOF series robot and the stationary point of the target,compared the PBVS,IBVS,HBVS visual servo method.Aiming at the shortcomings of pseudo-inverse construction method in existing IBVS,a hybrid Jacobian pseudo-inverse construction method is proposed,and the validity of the new construction method is verified by simulation.(4)Aiming at the problem of moving target trajectory prediction,the linear kalman filter under the known target motion state,the extended kalman filter of the multi-sensor fusion under the known target motion state,and the interactive multi-model kalman filter under the unknown target motion state are tracked to the target,Simulation verifies the effectiveness of these methods in image and 3D tracking.In the experiment,the tracking smoothing error is less than 5 pixels or 5mm.And finally obtains the effective method of real-time binocular vision servo stabilization of series robot.It uses the Ransac-Orb to extract the target point feature,using the improved ELAS to obtain the parallax,using the IBVS based on the hybrid Jacobian pseudo-inverse construct method as the visual servo control law,using the corresponding kalman filter to predict the trajectory of the moving target.