To address surface defects such as flying edges and grooves that commonly occur during friction stir welding, an embedded system for image processing of friction stir welds has been designed. First, GigE industrial cameras and line lasers were used to capture images of the weld seam. Approximate polygons of laser strips were fitted by contrast enhancement, iterative method threshold segmentation, edge fitting, etc. The dimensions of flying edges and grooves were obtained by methods such as Shi-Tomasi corner point inspection. Finally, the fuzzy PID control algorithm was used to regulate the amount of downward pressure, and experiments are conducted to verify the feasibility of the system. The results showed that the method can quickly and accurately detect and eliminate flying edges and grooves, and is highly adaptable to meet industrial requirements.