Fast and robust image processing for laser stripe-sensors in arc welding automation

For seam tracking in arc welding several sensors are used. Different tasks in welding automation lead to different sensors. Most sensors for welding automation don't have problems with the seam tracking, on the other hand there is the control of the welding process itself. For this job optical sensors are the best choice because they are able to provide the complete groove geometry. Combining arc welding and optical sensors causes natural distortions in the sensor signal. Different methods for reducing these kind of distortions are presented in the paper. Beginning with mechanical and optical shielding, the use of different lasers and the variation of mechanical parameters are discussed. The physical shielding strategies help reduce the distortions but cannot ensure a useable image of the welding groove. Therefore filter algorithms have to be applied to the sensor signals. High real time requirements need expensive multiprocessor systems to cope with the data rates. Therefore it is essential that the applied algorithms can be implemented on highly integrated logic ICs (FPGA). These programmable logic devices have been developed offering the same performance at a drastically lower price making laser stripe sensors attractive for cost sensitive applications in heavy industry. Image processing algorithms approved in other fields applied with the new FPGA hardware result in an extreme improvement in welding groove detection. Improvement in sensor use by welding automation is quantified using experimental data.