This paper proposes a real-time welding trajectory correction approach that integrates multi-simulation template matching for weld seam recognition with fuzzy adaptive PID control for motion correction. The recognition module also incorporates adaptive template sizing and a multi-template matching strategy to effectively handle weld seam variations and specular reflection noise in welding. Considering the geometric characteristics of line-structured light, a shape-based correlation measure is introduced to improve matching accuracy and robustness. A fuzzy adaptive PID controller is applied to ensure precise real-time correction during welding. Step response comparisons show that the fuzzy PID controller significantly reduces overshoot and rise time compared to a conventional PID controller. Additionally, the parameter adaptation behavior of ΔKp, ΔKi, ΔKd and converges to stable values. Experimental results on fillet welds under conditions of specular reflection and welding noise demonstrate high recognition accuracy, achieving a maximum error of 0.386 mm, an average error of 0.185 mm, and a root mean square error of 0.208 mm. Welding trajectory correction experiments demonstrate that the fuzzy PID controller enables faster convergence to the target path and achieves a lower steady-state error (0.3 mm) than conventional PID control (0.5 mm). These results validate the effectiveness of the proposed integrated approach in achieving robust weld seam recognition and accurate trajectory correction in robotic welding systems.