Robot Visual Servoing Research Under Complicated Natural Scene

Intelligent robot has been widely used in many fields, such as aerospace and industry,playing an important role, especially in machine vision of robot combining with robot vision sensor. Obtained image information by visual sensor, and got the controller to make robot move by processing image information, so that tracking or locating task can be fulfilled. Robot target tracking in the complex natural scene maybe interfered by target occlusion and target offset of vision, resulting in reduced tracking accuracy. So that overcome the interference factor is the key to improve the tracking accuracy.Aiming at the problem that keypoints stability of the tracking module is not high enough an improved tracking-learning-detection algorithm is proposed. The tracking part used LOG-FAST replaced even selects feature points to track using the pyramid optical flow method.Before tracking the current image needed to Gauss filtering and Laplace transform, and then obtained the keypoints by FAST. In order to get more stable keypoints using before-back error to get a select that improved the keypoints stability and the tracking performance. And then improved tracking-learning-detection algorithm is used to robot visual servo extracting current image feature. In order to solve the servo failure caused by the field of view excursion, a target disappearing strategy is used to obtain to predict the image feature as the current image feature.Based on that, the visual controller is designed combine Jacobin matrix and sliding mode control. Sliding mode control method is an effective method to control complex nonlinear system to improve the system's rapidity. The sliding surface and control are designed by image feature error.MOTOMAN SV3XL six degrees freedom industrial robot is used in the tracking experiments. The tracking experiments under the interference of target offset the vision,occlusion and lights change can be verified the validity of improved tracking-learning-detection algorithm. What's more, design experiment compared the visual servoing controller based on PID and the visual servoing controller based on the sliding mode controller can reflect that the sliding mode control can improve the system's rapidity.