Research And Application Of High-precision Position And Attitude Measurement Device In Intelligent Assembly

Since the 21 st century,the vigorous development of the national economy has promoted the continuous improvement of the level of science and technology.Intelligent manufacturing and assembly have become the mainstream development direction of modern manufacturing.The movement of traditional assembly robots is mostly under the control of human teaching or is programmed by programmers in advance.Less intelligent,less agile,and less capable of sensing production environment as it happens.Considering the above problems,the robot system is equipped with a visual measuring device,and the measuring device is combined with the robot to form a coordinated hand-eye system,which can enable the robot to automatically perform tasks such as positioning the target,grasping and assembling,which will greatly enhance the flexibility of the robot’s ability.Based on the actual application project of a research institute,this paper combines the position and attitude measurement device with the robot,builds a robot vision system to measure the position and attitude of the target part,and realizes the robot grasping and assembling the target part.The main tasks are as follows:(1)In order to solve the problem that the monocular camera has lost the depth information and cannot measure the position and attitude,a visual measurement system is built by using multiple laser displacement sensors and a monocular camera,and the surface position and attitude measurement of the target part is completed by combining the surface circular features of the target part.First,to find out the circular feature of the part surface,the Otsu and Canny algorithms are used to segment the image of the target part.Then,to fit the ellipse edge in the image,the RANSAC algorithm and the least squares method are combined to compute the ellipse equation.Afterwards,the solution of the spatial ambiguities in the camera frame is computed based on the radius of the object.Finally,the plane is adjusted according to the measurements of multiple laser displacement sensors,and the ambiguity solution is constrained by the normal vector of the target plane to obtain the correct solution.Experiments show that the project standard can be achieved in target part position and attitude measurement,and assembly tasks can be completed for multiple targets.(2)A commercial 3D scanner is used to obtain the point cloud of the target part to solve the problem that the monocular camera needs an auxiliary device to perform spatial 3D measurement,and a robot 3D vision system is built.The 3D-NDT registration algorithm and the ICP registration algorithm are studied,and the principle of point cloud registration is explained.Based on the experimental results of the aiming pieces,the ICP registration algorithm is used to realize the point-cloud registration of eight different pieces.The conversion relationship between the coordinate system,the end coordinate system,the jig coordinate system and the object coordinate system are investigated,the point clouds of the current target part are obtained and registered with the template point clouds,and the current target is successfully detected by the coordinate system conversion relationship.The robot attitude of the part to achieve target grasping.In the experiment,different parts were verified many times,and the point cloud registration accuracy met the project requirements,and the target grasping was completed.(3)A particular application requires that the inner parts be measured within a cylindrical cavity with a 170 mm diameter and 520 mm height.However,it is difficult to use the existing attitude measurement devices because of their large size or the need for an additional scanning motion mechanism.Therefore,a two-line laser-based monocular active vision attitude measurement method is proposed.It is suitable for small spaces.First,the center point of the line laser is extracted for light plane calibration.Next,the conversion matrix between the light plane and the image plane is calculated in accordance with the central projection relation.After that,the pixel coordinates of the center point of the line laser in the target plane are extracted and converted into camera coordinates.Finally,the RANSAC algorithm will be used to filter the anomalous points,and the least squares procedure will be used to fit the object plane formula to realize the target plane position measuring.Experimental results show that the measurement error of the device in roll and pitch angle reaches the measurement accuracy of commercial scanners,and finally the target plane attitude measurement in cylindrical cavity is completed,verifying the practicability of the method.