Automated single view 3D Texture Mapping and Defect Localisation of Thermography Measurements on large Components utilising an industrial robot and a laser system

Abstract

Autonomous measurement with NDT system like thermography on a large aircraft component (e.g. fuselage sections) is a major challenge for standard industry robot. Many factors may affect the absolute positioning accuracy of created robot TCP positions. 0.1° absolute robot inaccuracy (TCP) can cause for 1 meter working distance up to 3mm position error. Another major drawback of thermography system is that certified examiners evaluate every 2D thermography image per measuring field one by one. This unhandy and less intuitive method for large number of numerical data is inappropriate for industrial applications. There are plenty of existing methods and technologies like Stereo vision method, using 3D feature or multi-view method for 3D texture mapping. These methods require either a second camera or time consuming additional processes or several images from every single measuring field. This paper introduces development of a prototype robot based thermography measuring system and a new method for 3D evaluation along with visualization of thermography results, which minimizes drawbacks described above. Evaluation in 3D components coordinate, gives complex but complete 3D defect propagation knowledge. DLR-AQP department is working on two different methods to calculate 2D-3D camera projection parameters. Method one projects predefined 3D points per measuring field by using laser projector and applies 2D-3D point correspondence algorithm. The second Method uses robot position and applies kinematic-coordinate-relationships algorithm. For both methods single thermal image is captured by using thermography camera, which is also used for NDT. The first method is applicable in laboratories as well as in production with and without any manipulator. This method enables also contactless localization of detected defects. The second method requires robot. About accuracy, reproducibility and flexibility of both methods have been compared and discussed in this paper. The 3D results are to be imported directly into the CAD environment for further structure analysis.