Abstract: For the existing problems in welding process of H section steel, such as mostly manual welding, high labor intensity, poor operation environment and high randomness of manual measurement, an automatic welding position detection technology based on vision and image processing was proposed. This technology was mainly based on line structured light measurement method to realize outline detection of section steel to determine welding position, and the position of weld was determined by the position of laser line, and then the purpose of automatic welding was achieved by combining with welding robot. Firstly, Zhang’s calibration algorithm was used to obtain the camera internal parameter and distortion coefficient. Then, the laser plane was calibrated by fitting the laser plane with the least square method and the whole measurement system was calibrated by combining the two methods. And then, based on OpenCV programming interface, image processing was realized. Finally, ABB robot tool coordinate system was use to evaluate the accuracy of the measurement method. For the large coordinate error problem of detection point in the depth direction, an error fitting method was put forward and it was applied to various types of H section steel measurement test. The test results showed that the measurement error of the line structured light visual measurement system calibrated by the above method was less than 1 mm, which achieved high calibration accuracy, significantly reduced the coordinate error, and met the requirements of welding.Highlights: A method of error fitting based on standard measuring block can effectively reduce coordinate error.