|
The robot has been increasingly used in welding, painting, assembling, handling a variety of modern industrial production processes, as the robot is the most important sensory information, the visual mainly used for measurement of the target and the end of the robot position and orientation in order to posture of the robot end-bit control. Robot visual servo system, make the robot able to capture to the target, the visual system is not only to be fed back to the goal of the three-dimensional position information, and to calculate the three-dimensional position and orientation of the robot arm parameters for real-time tracking of the actual task, the feedback of the end of the operation the state of the robot controller path planning. In practical applications, due to the background, illumination, and other interfering factors, the visual control methods are often difficult to receive good results. Therefore, visual examination link robot visual servo system have been investigated, trying to establish binocular vision oriented series six degrees of freedom robotic arm to capture target detection system are analyzed and related issues. Firstly, according to the application target, the visual servo system overall framework to build a hardware platform for the analysis of the various sub-modules, and that this structure helps to guarantee and improve the real-time nature of the system. On the acquisition of the video signal, the Microsoft DirectShow technology, and the image data into a the OpenCV data format, for subsequent processing. Detection and matching of the target and the robot arm is the center of this study. In this paper, a small red ball as experimental subjects, analysis of three detection strategies based on the Hough transform, color and motion information, the results show sports information combined with the color feature detection algorithm is the best choice under the present experimental conditions. Due to the complexity of the mechanical arm of a statistical model classifier method - based on the the class the Haar characteristics of AdaBoost algorithm applied to a series robots detect six degrees of freedom, and achieved good results, the detection rate of 85% or more . Mathematical modeling 3D position and orientation of the robot, according to the the SIFT matching several pairs of feature points, the last series six degrees of freedom robot arm space is three-dimensional position and orientation parameters (t_x t_y t_z, θ φ, (?)), and the actual value difference is small, the motion information can be used as feedback to the robot controller. Finally, this paper work summary and outlook.
|