The current flexible production of small batches puts forward higher requirements for welding robots. Intelligent welding faces many challenges, such as welding target recognition, 3-D scene reconstruction, welding path generation, seam tracking and welding defect detection. However, it is difficult for a single sensor to accomplish all of the above tasks. To improve the intelligence level and autonomous welding ability of robots, a hybrid vision sensor was developed. The sensor integrates monocular vision, laser structured light vision and coded structured light vision simultaneously, which has a compact structure but is mighty in function. The hardware, electrical and software systems of the sensor are introduced in detail, and experiments are carried out on the hand-eye system composed of the sensor and the welding robot. Experiments show that the welding robot system integrated with hybrid vision sensors realizes multiple functions such as weld type identification, initial point detection, point cloud acquisition, welding path generation, feature point extraction and seam tracking.