In order to realize the accurate recognition of weld target by welding robot under complex working conditions, laser vision sensing technology is applied to robot weld recognition. Taking the laser vision sensor as the core hardware, we build the welding image acquisition system, use the PCI bus card to detect whether the sensor and external indicator are connected, obtain the laser welding image within the specified time through the PXR800 image acquisition card, and use the serial port communication program to complete the efficient image transmission. The bilateral filter and fuzzy operator are used to denoise and enhance the image, optimize the image corner and the center of the laser stripe, and use the equality constraint equation to modify the image coordinate value after Kalman filter to complete the task of high-precision weld recognition. Simulation results show that laser vision sensing can help the robot improve the accuracy of weld recognition and enhance the anti-interference ability, which plays a positive role in the popularization and application of welding robot.