Object Posture Recognition And Robot Pickup With Guidence Based On Monocular Vision

As the wide application of robot technology in industrial field, more and more industrial robots begin to replace manual work, such as product sorting and loading or unloading material, etc. The more diversification of demand, the higher of requirement for robot in industrial field. So, it is one of difficulties to accurately identify and grab the right target in different shapes and sizes of targets in the field of automation production. Faced with industrial manufacturing and assembly line, various vision-guided industrial robots grasp real time positioning technology has been applied. Automatically grasp the target objects through guided industrial robots, can effectively solve the identification and location of the target objects to the problem, can realize manufacturing and assembly full automation, and improve production efficiency.A method of monocular vision technology was put forward, which can obtain three-dimensional position information of small or medium size rule target efficiently. It can overcome the disadvantage of the structured light technology and binocular stereo vision technology for camera and target position solidification, computation complexity and the poor robustness etc. Establishing the geometry model of any possible point in space, by the target shape size information as constraint conditions, make sure that each point can mapping with the pixel coordinates one-on-one. Then reconstructing the three-dimensional pose of target through calculating the feature points of the plane in the world coordinate system, according to the calibration of the machine vision system.In the process of image processing, aiming at the characteristics of the target and the background, a series of image processing algorithm was compared by NI VISION visual library. In terms of communication, a variety of the data communication process of industrial robots was designed which was suitable for machine interaction between the host computers and the slave computers, by detailed study of all kinds of industrial robot controller communication interface and communication protocol.Aimed at monocular vision guided industrial robots designed system processes to grasp the targets, and established hardware and software platform. Choose the known size rectangular and cylindrical targets to do posture recognition and target grasping experiment, the result show that the proposed monocular vision method can guide industrial robots grasp quickly and effectively, and workpiece posture recognition error was little, which can meet the expected requirement when the industrial robots operate the workpiece in industrial production.