The optimal selection of medium-wave channels in the geostationary interferometric infrared sounder (GIIRS) of FY-4A is one of the key technologies for effective variational assimilation of the data, which can reduce the ill-posed assimilation and inversion caused by redundant information. In this paper, the general entropy reduction method is applied to GIIRS channel optimization. In addition, since the observations of infrared sounders are easily affected by clouds, cloud detection is needed to obtain field points or cloud parameter information in the clear sky during the variational assimilation of GIIRS bright temperature data. In this paper, the minimum residual method is used to conduct cloud detection of GIIRS data. This method can not only determine whether there are clouds, but also obtain the information of effective cloud amount and cloud top pressure at the field-of-view. Since the detection accuracy is affected by the combination of different channels, the influence based on the data of Typhoon "Lekima (2019)" is further discussed in this paper.