Penetration identification method based on two-way molten pool geometric and texture feature fusion

• Main Authors: YANG JIAJIA, WU TONGLI, CHEN YAODONG, CAO HUI, NI ZENGLEI, HAN PENG, WANG QINGBO, YANG WENGUANG, LI YINGJIE
• Format: Patent
• Language: Chinese; English
• Published: 19.03.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a penetration identification method based on two-way molten pool geometric and texture feature fusion, and relates to real-time sensing of a welding molten pool image, image information extraction and a penetration identification method based on molten pool image information. Clear and synchronous molten pool images are obtained from the rear upper portion and the side upper portion of a molten pool through a two-way synchronous vision sensing system, then image processing is conducted on the two-way molten pool images, geometrical characteristics of the rear upper molten pool images and texture characteristic information of the side upper molten pool images are extracted, and an RBF penetration recognition model is established based on the geometrical characteristics and the texture characteristic information. And the penetration state is pre-judged by utilizing the real-time molten pool image, and is fed back to the welding machine through the controller to realize penetration contro