Abstract: Existing active vision-based weld seam tracking techniques have a significant decrease in tracking accuracy and stability when dealing with narrow gap welds(gap width ≤0.3 mm) due to the difficulty of laser stripe to produce significant deformation. For the reason, a narrow gap weld tracking algorithm was proposed based on the gray scale difference enhancement method. In terms of hardware, by increasing the LED light source with the same filer central wavelength and irradiating it to the surface of the welded workpiece, the narrow gap weld region is unable to reflect the light, so that the gray value of the region is significantly lower than that of the two sides of the region. In terms of algorithms, firstly, the exposure pre-processing technology was used to further enhance the gray difference between the narrow weld region and the surrounding area; secondly, a weld feature point extraction method based on the gray difference enhancement method was proposed to realize the narrow gap weld tracking algorithm; and then, a gray difference enhancement method based on the weld feature point extraction method was proposed, achieving precise detection of narrow-gap weld feature points. Finally, a weld tracking algorithm based on the target point guidance was proposed by combining the FIFO weld sequence stepping strategy with the front distance scanning. The results show that the tracking algorithm has an average Euclidean distance error of 0.203 mm in narrow gap weld tracking, which is highly accurate and meets the requirements of industrial applications.