Robot Recognition And Location System Research Based On Multi-vision

Nowadays,more and more robots take place of artificial labor in the industrial production.In order to make industrial robots more intelligent,add visual function to them is a trend.However,most of the current visual applications still remain in the field of monocular vision,which can not meet the three-dimensional space location requirements in industrial production,so three-dimensional visual technology has become a research hotspot recently.In this paper,we use multiple cameras vision system as well as parallel robots to research on camera calibration,target feature extraction and feature matching algorithm,target three-dimensional spatial location method and robot grab control.First of all,the multi-vision system used in this paper is calibrated,including two binocular cameras and a monocular camera.The concept of several coordinate used in the visual system and the relationship between them as well as the imaging model of the camera are analyzed.And then briefly explain several camera calibration methods and we use Zhang Zhengyou calibration method to complete the calibration of three cameras and the binocular cameras.The internal and external parameters of the cameras and the positional relationship between the camera image coordinate system and the world coordinate system are obtained,The calibration results are analyzed to verify that the calibration accuracy can meet the requirements of target location in this paper.Secondly,as for target recognition,we introduce the commonly used image features.Then the principle and advantages as well as disadvantages of several feature point algorithm are analyzed.The feature point extraction and matching method based on the SURF algorithm is used to recognize the target.In order to eliminate the false matching to make the recognition result more accurate,we use the RANSAC algorithm to remove the false matching feature points of the SURF algorithm and only reserve the right matching feature points to achieve the goal of accurate recognition.In the three-dimensional location of the target,this paper uses a two-step location method.Firstly,the location information of the feature points obtained by the SURF algorithm and the RANSAC algorithm as well as the binocular camera calibration results are used to get the initial loacation of target.And then use the monocular camera to take picture of the target object,extract the contours of it,locate the center of it,so as to get the more accurate location information of target object.Finally,the grabbing function of this paper is realized by using 3-HSS parallel robot.Parallel robot has a good quality of stiffness and the anti-solution of location is easy,so it's widely used in the industrial production line sorting and grabbing occasions.In this paper,the kinematics equation of the 3-HSS parallel robot is analyzed.Then we use the single chip to control the stepping motor to achieve the grab function of the target object.And the experiments of target recognition and grabbing in different situations are carried out,which prove that the methods we use in this paper are simple and effective.