Research On Robot Visual Servo Control Based On Linear Structured Light

With the development of machine vision and computer technology, robot visualservo control technology is studied increasingly by researchers and has been applied inindustrial environment. Meanwhile, the vision detection technology based on structurelight is a kind of three-dimensional measurement technique actively controlled andcommonly used in machine vision. Recently this technology has attracted a greatamount of attention in industrial environment for its many advantages, such as highaccuracy of measurement, easy to extract the information from laser stripe image,strong real-time ability and so on, which has a wide range of application in the fieldsof3-D auto-measurement, auto-processing, robot vision and so on. In this paper, thevision detection technology based on linear structure light is used to complete thethree-dimensional measurement for the object and achieve the simple servo controlconnecting with the robot. The main work includes the following several aspects:(1) The composition of visual measuring system and its mathematical model areintroduced and analyzed, and the visual measuring system is researched and designedintegrally in this thesis. By using three-dimensional motion mechanism as a drive toget the calibration points, the camera and the linear structured light are both calibratedby using the same set of calibration points according to the radial alignment constraint(RAC) approach. So the interior and exterior parameters of camera and linearstructured light sensor can be also figured out with that method.(2) Laser stripe image processing method is studied in the thesis, including imageacquisition and filtering, threshold method for image segmentation, the center line ofthe laser-stripe extraction, feature point extraction. Analyzing the factors that affect theaccuracy of center line of laser-stripe extraction, this paper proposes a new laser-stripeextraction method. After getting the center line, the least square method is used to fitthe center line. After getting two fitting line and calculating the intersection point,feature points can be obtained.(3) After studying the traditional robot hand-eye calibration algorithm, a fasthand-eye calibration algorithm is put forward in this paper. Just taking a photo forcalibration plate by CCD, we can obtain the relationship matrix between end-effectercoordinate and camera coordinate, and this method has many advantages such as less amount of calculation, fast and easy to operate, high efficiency, which can basicallymeet the accuracy requirement of the operating system according to the experiment.(4) The experiment stage. Firstly we finish the experiment of3-D measurementfor the target object according to vision measurement system in this thesis. Then wecan obtain3-D points cloud and size of the target object. Finally the simple servocontrol for robot can be accomplished according to the visual servo control systembased on the location with combining the robot.In this paper, the software development platform is based on Visual Studio2010,combining image processing library of OpenCV and PCL (Point Cloud Library) andusing C++to complete the task of each part of the experiment. The result ofexperiment shows that the positioning accuracy of robot vision system based onstructured light can basically meet the requirement of experiment.