To realize high-quality robotic welding, an efficient and robust complex weld seam feature point extraction method based on a deep neural network (Shuffle-YOLO) is proposed for seam tracking and posture adjustment. The Shuffle-YOLO model can accurately extract the feature points of butt joints, lap joints, and irregular joints, and the model can also work well despite strong arc radiation and spatters. Based on the nearest neighbor algorithm and cubic B-spline curve-fitting algorithm, the position and posture models of the complex spatially curved weld seams are established. The robot welding posture adjustment and high-precision seam tracking of complex spatially curved weld seams are realized. Experiments show that the method proposed in this article can extract weld seam feature points quickly and robustly, which enables welding robots to accurately track the weld seams and adjust the welding torch postures simultaneously.

Seam feature point acquisition is the premise of the intelligent welding process such as initial point guiding and seam tracking. However, conventional seam feature point acquisition methods based on geometric feature have shortcomings of poor flexibility and robustness. In this paper, a seam feature point acquisition method based on efficient convolution operator (ECO) and particle filter (PF) is proposed, which could be applied to different weld types and could achieve fast and accurate seam feature point acquisition even under the interference of welding arc light and spatter noises. Firstly, a structured light vision sensor is developed to acquire welding image. Secondly, ECO algorithm is adopted to track the seam region and acquire seam feature point during gas metal arc welding (GMAW) process. Thirdly, the state and measurement equations of the weld seam position are established, and PF is applied to improve seam feature point acquisition accuracy. Finally, a welding experiment system is built and a series of seam feature point acquisition experiments of butt joint, lap joint and fillet joint are carried out to validate the performance of the proposed method. The experiment results demonstrate that the processing speed of the proposed method could reach up 35Hz, and the seam feature point acquisition errors are smaller than 0.15mm, which could meet the real-time and accuracy requirement for subsequent initial point guiding and seam tracking.

Start point guiding and seam tracking are important parts of the robot's intelligent welding process. However, some start point guiding methods have a large amount of calculation and low accuracy. In addition, due to the large curvature of the curved weld, the problem of visual advance in seam tracking should be solved. Therefore, an efficient and accurate start point guiding and seam tracking method for curve weld is proposed in this article. First, according to the characteristics of structural light of different welds, more efficient and accurate start point detection algorithms are proposed for curve v-groove welds and curve lap welds. The extraction of key characteristic points and the judgment of the relationship between them could achieve efficient weld start point detection, and the reasonable robot speed planning during the start point searching process could improve the accuracy of the start point guiding. Besides, a sliding data queue method with cubic B-spline fitting is proposed to overcome the visual advance problem and achieve accurate seam tracking of curve welds. The sliding data queue method can plan the data queue automatically, and improve the computational efficiency by reducing the amount of data to be processed. Cubic B-spline fitting can reduce the measurement error and eliminate the data fluctuation, so as to obtain a smooth welding trajectory. The experimental results show that the proposed method could achieve efficient and accurate start point guiding and seam tracking of curve v-groove weld and lap weld.

Spatial circular welding is an important application in industrial production. However, some feature points extraction methods for spatial circular welds have disadvantages such as large amounts of calculation and poor anti-interference ability. In addition, due to the welding thermal deformation, the spatial circular weld needs to be tracked. Therefore, a fast and robust seam tracking method for spatial circular weld based on laser visual sensors is proposed in this article. First of all, a self-updating template matching (SUTM) method is proposed to get weld feature points. Compared with the method based on geometric features, the proposed SUTM method has the advantages of rapidity and robustness. What is more, an algorithm for spatial circle center fitting based on random sampling consensus (RANSAC) is proposed, which overcomes the interference of outlier points and improves the robustness of circle center detection. At the same time, the tracking error model of the spatial circular weld is established and the welding deviation from the $x$ -, $y$ -, and $z$ -directions is obtained. By correcting the deviation in real-time, the fast and robust real-time seam tracking of the spatial circular weld is realized. Through welding test, the proposed method can obtain the position of the spatial fillet circular weld quickly and robustly and complete the high-precision seam tracking.

Robotic welding is an essential technology in industrial production. Various sensors are integrated into robotic welding systems to enhance welding efficiency and quality. Due to their advantages of non-contact measurement, high information capacity, and high accuracy, vision sensors play an increasingly important role in robotic autonomous welding. Therefore, the latest vision-based autonomous robotic arc welding technologies are comprehensively summarized, providing valuable reference and support for researchers. First, the research progress of visual sensing technology for welding robots is outlined, and its advantages and disadvantages are analyzed. Subsequently, the key technologies of robotic autonomous welding are introduced, covering weld type identification, initial point guidance, welding path generation, welding parameter planning, feature point extraction, seam tracking, welding posture adjustment, and welding quality control. Finally, the limitations existing in current research are summarized, and the future research directions of autonomous robotic welding are prospected.