Square groove detection based on forstner with canny edge operator using laser vision sensor

• Main Author: Mohammed Naji, Osamah Abdullah Ahmed
• Format: Thesis
• Language: English; English
• Published: 2023
• Subjects: T Technology (General); TS Manufactures
• Online Access: http://eprints.utem.edu.my/id/eprint/29056/; https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=123844

Weld seam recognition is critical for providing infurma tio n for automated welding contro 1, promoting 1he advancement of welding sensing technology, and improving welding manufucturing automation. The obtained location infurmation of1he weld joint is used in 1he welding industry to direct 1he welding torch along 1he best welding direction, allowing 1he wid1h and dep1h variations of1he weld joint to be compensated during welding. This research addresses 1he feature extraction algorithm of 1he square weld groove to make weld seam recognition more accurate. The acquisition of 1he weld image is a complicated process, resulting in a large amotmt of noise. Specific me1hods must be used to process images. The objectives of 1his research are to develop a detection algorithm 1hat can extract 1he feature points of1he square-groove; and 1he second objective is to evaluate 1he detection algorithm and its ability to extract 1he image features of 1he square-groove in temis of 1he accuracy. In 1his work, the central line of the laser stripe is extracted based on Canny edge detection with Haralicks fucet model Based on 1he central line, 1he Forstner algorithm is used to recognize 1he comer points of 1he square weld groove. Following 1he establishment of a test platform, a series of detection tests for various sizes of1he square groove is established. The Forstner algorithm was compared agailist 1he Harris algorithm to evaluate which one was more accurate in extracting 1he square groove's features points. The obtained result showed 1hat 1he Forstner algorithm's maximum thickness and width measurement errors are 3 .193 % and 4.00%, respectively. Therefore, 1he acquired detection results are sufficiently accurate, demonstrating 1he rationale of1he suggested visual sensor's physical design and 1he validity of 1he proposed detection algorithms.