Hand-eye calibration of a robotized thermography end-effector NDT in Aerospase 2014 Madrid, Spain Thomas Schmidt German Aerospace Center for Lightweight-Production-Technology (ZLP site Augsburg)
DLR. de • Chart 2 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Outline • Motivation and context • Challenges • Hand-eye calibration in detail • Results from experiments • Summary • Outlook
DLR. de • Chart 3 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Motivation and Context • Parts to be measured in aerospace are much larger than field of view Need for automation to enable measurement of larger parts tiling with a flexible manipulator including accurate positioning of camera during measurement to allow defect localization
DLR. de • Chart 4 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 End-effector prototype • Flexible set-up • Ability to carry up to three halogen lamps • Camera and lamps are adjustable
DLR. de • Chart 5 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Boundary Conditions • One measurement is roughly DIN A 4 size • Large structures can be measured by tiling from a number of views • Robot or work cell can be used as a manipulator for the camera • Robot delivers information about location and orientation in space in relation to a coordinate system of your choice • Measurement process had been automated via industrial bus technology
DLR. de • Chart 6 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Challenges on measurement system 1. Optical distortion of thermography camera Real object Camera image • Due to lens effects • To differentiate between barrel, pincushion and mustache distortion depending on focal length How to get a correct image?
DLR. de • Chart 7 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Challenges on measurement system 2. Camera is mounted on robot flange via end-effector Position and orientation of camera as a relation between flange and camera coordinate system is important Null. Frame (Flange) How to get a correct TCP? TCP (Tool Point) Rob. Root Center
DLR. de • Chart 8 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Approaches 1. Estimation from the optical properties of camera 2. Several opportunities 1. Modelling via CAD • Problem: Simulation inaccurate and doesn‘t represent reality 2. Pseudonumerical approach – Combination of lasertracker measurement and CAD • Assumption: lens projection is normal to lens geometry • Problem: this is never the case
DLR. de • Chart 9 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Solution Numerical hand-eye calibration using camera images and robot position data [1] http: //www. rob. cs. tu-bs. de/research/projects/calibration/
DLR. de • Chart 10 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Definition of hand-eye calibration 1. Finding the quantities internal to the camera that affect the imaging process • Position of image center in the image • Focal length • Different scaling factors for row pixels and column pixels • Skew factor • Lens distortion 2. Calibration is important when we need to interact with the world • Robot, hand-eye coordination • It‘s not temperature calibration!
DLR. de • Chart 11 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Additional Challenge • Images from thermography camera at room temperature don‘t show contrast and are blurry
DLR. de • Chart 12 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Hand-eye calibration process using robot position data • Camera is mounted on end-effector and takes at least 10 images from pattern under different positions and angles • Note position and orientation of robot flange at every image • Use both information for calculation [2] HEC Toolbox TCP {X; Y; Z; A; B; C} DECL E 6 POS XP 1={X 1547, Y 671, Z 2007, A 156, B -24, C 108, S 22, T 51, E 1 0. 0, E 2 0. 0, E 3 0. 0, E 4 0. 0, E 5 0. 0, E 6 0. 0} DECL FDAT FP 1={TOOL_NO 0, BASE_NO 0, IPO_FRAME #BASE, POINT 2[] " ", TQ_STATE FALSE} [2] http: //www. vision. ee. ethz. ch/software/calibration_toolbox//calibration_toolbox. php
DLR. de • Chart 13 Video Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014
DLR. de • Chart 14 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Hand-eye calibration
DLR. de • Chart 15 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Hand-eye calibration • Feature recognition finds pattern and coordinate system • Offset values successfully determined (T{X, Y, Z}; R{A, B, C})
DLR. de • Chart 16 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Hand-eye calibration Iterative approach • At least two or more steps required • Step 1: Calculate pre-TCP • Teach robot to 10 random positions with roughly same distance to checker board • Step 2: Calculate TCP • Perform offline programming with pre-TCP as tool data and constant distance to center of checker board • Repeat step 2 to achieve higher accuracy if necessary
DLR. de • Chart 17 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Reprojection error results
DLR. de • Chart 18 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Distortion Images Distortion acc. to Zhang‘s algorithm [3] Camera image calibrated to correct optical effects [3] http: //research. microsoft. com/en-us/um/people/zhang/Calib/
DLR. de • Chart 19 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Experimental results • Camera calibration (Camera matrix) Focal Length: Principal point: Skew: Distortion: Pixel error: fc = [ 1973. 20306 1969. 91104 ] ± [ 3. 55756 3. 31701 ] mm cc = [ 293. 88650 255. 12651 ] ± [ 5. 91518 4. 43811 ] px alpha_c = not been optimized yet kc = [ -0. 43104 0. 96768 -0. 00047 0. 00000 ] ± [ 0. 03405 1. 10513 0. 00040 0. 00035 0. 00000 ] err = [ 0. 14661 0. 10570 ] px
DLR. de • Chart 20 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Experimental results • Hand-eye calibration Rotations -0, 8715 0, 4902 0, 0111 0 -0, 2394 -0, 4056 -0, 8822 0 Transl. -0, 4279 -123, 0687 -0, 7715 -201, 0337 0, 4708 952, 8292 0 1
DLR. de • Chart 21 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Experimental results • Hand-eye calibration after 6 runs Average Translation Standard Deviation Average Rotation Standard Deviation [mm] [°] X -123, 0427 0, 281 A 150, 643 0, 012 Y -201, 0124 0, 163 B -0, 635 0, 086 Z 952, 6491 0, 507 C -61, 912 0, 046
DLR. de • Chart 22 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Summary • Successfully demonstrated hand-eye calibration using an infrared camera mounted on an industrial six axis robot using an end-effector • Numerical solutions for camera calibration and tool data (TCP) • Complete camera matrix for image undistortion available • Image location and measurement of defect size achieves much higher accuracy than before
DLR. de • Chart 23 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Outlook • • • Higher accuracy through optimized pattern material and resolution Higher accuracy through measured robot position data laser tracker Use different patterns sizes for different focal lengths Implement the distortion algorithm in thermography NDT software Streamline HEC toolbox more user friendly
DLR. de • Chart 24 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Thank you! Questions?
DLR. de • Chart 25 Hand-eye calibration of a robotized thermography end-effector > NDT in Aerospace 2014, Madrid, Spain > Thomas Schmidt > 13. 11. 2014 Acknowledgements • The establishment of ZLP Augsburg had been funded by the City of Augsburg, the Free State of Bavaria as well as Federal Ministry of Economics and Technology • Many thanks to the team at DLR-ZLP
Скачать презентацию Hand-eye calibration of a robotized thermography end-effector NDT
b3d46c069f45c06ef7a880fb2d799509.ppt