Motion Planning Based On Computer Vision For Industrial Manipulator

With the development of industrial technology and the requirements of production.The robot technology is not only confined to complete the simple tasks.It also need to own stronger adaptability, flexibility, stability, security and other features.There are two main targets in the study of mechanical arm.One target is the research of kinematics.There has not gotten a general algorithm to inverse movement problem of a mechanical arm.The work efficiency will be improved if we find a scientific algorithm to resolve this problem.The other target is the study of visual algorithm.The industrial robot needs interact with the environment and high precision recognition to the environment to improve their flexibility.A reasonable image processing method can greatly enhance the visual system’s ability to image matching recognition and achieve multiple targets.The main task of this paper is researched on the above two points :First we researched the kinematics of manipulator.The mathematical model of mechanical arm is established by using the DH method.Forward kinematics equation is established.Then the inverse movement of mechanical arm is solved by using the artificial bee colony algorithm.The algorithm can accelerate the convergence rate by optimization the threshold of BP algorithm neural network.the manipulator inverse kinematics problem was transformed into a class of nonlinear MIMO problem.Finally,we used MATLAB to dynamic simulated the manipulator joints.We calculated the displacement of each joint when the end effector reach the specified spatial coordinates.The results show that the simulation model of manipulator can reach the goal to complete the task within a reasonable margin of error.Then we researched the part of visual on the basis of the base coordinate system is established by the kinematics.First the single, binocular vision positioning principle are introduced respectively.According to the target point’s pixel coordinates in the image,finally the target space coordinates are got by using coordinate transformation method.And then the SIFT algorithm is studied in the application of image processing.Each relevant concepts of SIFT algorithm is introduced.Finally SIFT algorithm is combined with neural network experimentally. Take the scale parameter of he SIFT algorithm and the focal length of the camera as the input parameters.Take the distance to target as the output parameters.A neural network is establish tomeasure the targets distance.The results show that the SIFT-BP algorithm can accurately find the feature point which matched the image and measure target distance within a reasonable margin of error.