Abstract: An infrared visual sensing method for Gas Metal Arc Additive Manufacturing (GMA-AM) weld pool was proposed. In order to evaluate the image quality of the molten pool objectively, the evaluation parameter φ was defined by four kinds of features containing gray level, texture, shape and spectrum. The greater the value of φ, the better the image quality. The image quality evaluation parameter φ of the infrared weld pool was much greater than that of the near-infrared narrow-band filter weld pool. Compared with 800, 850 and 930 nm, the 990 nm infrared filter could filter most of the continuous spectrum and characteristic spectrum, which obtained much more information, more contrast clearer edge and more details. The results proved that the GMA-AM weld pool image by 990 nm infrared had the best image quality.