Among various welding process information sensing methods, the visual method is recognized as providing the most information and the best results. As early as the early 1980s, many researchers at home and abroad began studying visual sensing methods, including passive visual sensing using arc light as the light source and active visual sensing with laser-assisted lighting. In passive visual methods, the arc itself monitors the position, avoiding advance detection errors caused by thermal deformation and directly obtaining information about the weld seam and molten pool, which is beneficial for adaptive control of welding quality.
However, direct observation is easily disturbed by the arc, and there are still no mature industrial applications reported. Therefore, active optical vision, especially structured light or scanning methods based on laser triangulation principles, has become the primary visual sensing method in welding industrial applications. The greatest feature of laser vision sensing is its ability to obtain precise geometric shapes and spatial positions of weld seam cross-sections, suitable for real-time weld seam tracking and adaptive process parameter control.
The basic principle of laser vision sensing is optical triangulation. A laser beam shines on the surface of the target object, forming a light spot. This spot produces an image point on a photosensitive detector through a lens on the camera. Since the relative position of the laser and camera is fixed, changes in the distance between the laser sensor and the target object cause corresponding changes in the image point position on the photosensitive detector. Thus, height changes can be calculated based on the triangular relationship, measuring height variations.
When the laser beam scans in a certain shape (scanning method) or projects linear or other geometric patterns (structured light method) on the target object's surface through optical devices, the area-array photosensitive detector can obtain laser stripe images representing the target cross-section. As the laser sensor scans along the object's surface, it obtains contour information of the scanned surface. This information can be used for weld seam search and positioning, weld seam tracking, adaptive welding parameter control, and weld seam formation detection.
Currently, laser vision sensors used in welding mainly include scanning and structured light forms. Scanning methods primarily involve linear and circular scanning, with circular scanning being more complex in image processing. Comparatively, scanning methods handle reflections better than structured light methods. Additionally, scanning sensors have a greater depth of field, but their precision, especially lateral resolution, is relatively lower due to the influence of scanning laser spots. Mechanical scanning affects scanning speed, making scanning laser sensors mostly suitable for weld seam tracking and adaptive control of thick workpieces. Laser vision sensors used in high-precision and high-speed tracking or detection are mostly structured light sensors.
