Research on Weld Recognition Method Based on Mask R-CNN

In order to improve the efficiency of welding operations and the quality of welding products, reduce the work intensity of welders and improve the harsh working environment, through the welding seam tracking technology, the welding seam image collected by the camera is transferred to the computer for welding seam recognition and automatic welding seam image Recognition is the key technology for research on automated weld tracking. In this paper, the Mask R-CNN model is applied to the recognition of micro-gap welds. This method uses deep learning network for feature extraction, and realizes weld recognition and location. First, the Mask RCNN convolutional neural network is used to extract weld features from the weld image, and then the least square method and weld centerline extraction are used to obtain the width, angle and position of the weld to facilitate real-time tracking and positioning of the welding. The recognition accuracy rate is over 97%. The results show that, compared with traditional machine vision methods, the deep learning Mask R-CNN model has the ability to automatically learn from the training set, abandons the learning features manually set, and greatly improves the recognition efficiency and adaptability.