Binocular Vision Based Pose Detection For Cabins

Due to the huge size of large equipment such as rocket,aircraft,missile and ship,it is difficult to realize one-off manufacturing and assembly tasks,and the production mode of "partial assembly and docking" is often adopted.The production and initial assembly of components are finished during the partial assembly process;in the final process,the docking between the main sections is completed,and the position and attitude(position and attitude are simplified as pose)detection of each cabin has been a great concern.High precision pose detection method can ensure the quality of docking assembly.In this thesis,a real-time detection method based on machine binocular vision is proposed to obtain the poses of two cabins during the docking process.The visual pose detection system and adjustment platform are designed and tested.The main research work is as follows:(1)A binocular stereo vision based pose detection system is designed and developed.The measurement accuracy of the binocular stereo vision system is analyzed.According to the requirement of the docking accuracy of the cabins,the camera and lens used in the binocular vision system are selected and the system structure is designed.The software is designed based on the Windows system and Qt,and the parameters of the two cameras and binocular structure are calibrated.(2)A binocular vision based algorithm for cabin pose detection is proposed.Visual marks are pasted on the surface of the target cabin to provide artificial features.The marks are detected and their coordinates are extracted through the binocular stereo vision system,and then the pose matrices of marks are obtained.A device is designed to calibrate the relationship between the poses of marks and those of cabins.The pose matrices of the marks are transformed into the pose matrices of the target cabins,and the relative pose matrix between the two target cabins is obtained and solved with six degrees of freedom.(3)A 6-degree-of-freedom(6-DOF)pose adjustment platform is designed and developed.The platform can adjust three translational DOF and three rotational DOF of the cabin,and the D-H control model of the platform is established.The software of the platform is designed based on the Windows system and Qt.Stepper motors are used for servo,which are controlled via the CANopen network protocol.The platform includes three operation modes: location,speed and home.(4)The experiment of cabin pose detection is designed and the result is analyzed.Two kinds of visual markers are tested and compared.The pose detection system is verified using the 6-DOF adjustment platform,and the error is analyzed.Results show that the precision of 0.06 mm in three translations and 0.07° in three rotations can be achieved using the proposed binocular vision detection method.