With the deep integration of traditional Chinese medicine acupuncture and intelligent robot technology, acupoint recognition and tracking have become the key links to achieve precise and automated operation. In order to deal with the problems of weak acupoint features, frequent occlusion, and changeable postures under complex backgrounds, this paper studies and constructs an acupoint positioning and tracking algorithm based on template fusion and structure perception. It introduces multi-scale interpolation template matching and gradient direction response enhancement mechanism, and combines dynamic histogram update and state prediction to achieve high-precision and high-stability acupoint recognition. In 1920×1080 resolution images, the system achieves an average pixel error of 2.14 px, a physical error of 1.62 mm, a real-time frame rate of 18.6 fps, and a minimum tracking loss rate of 2.1%. Under different interference scenarios, the maximum offset is always less than 2.5 mm. The satisfaction rating of practitioners shows that the accuracy dimension score is 4.63/5. In summary, the system has good practicality and robustness, and can provide a feasible technical path for the promotion of acupuncture robots in real scenarios.