Cloud detection is a critical step in the application of infrared high-spectral radiance observations, directly impacting the effectiveness of satellite data utilization. McNally proposed a method in 2003 based on observed and simulated brightness temperature differences for channel cloud detection, widely applied in satellite data quality control for numerical weather forecasting. Building upon McNally's method, this study utilizes Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) cloud classification data products to quantitatively assess the cloud detection performance of FY-3D High Spectral Infrared Atmospheric Sounder (HIRAS), using precision and recall as validation metrics. This enhances the quality and assimilated data volume of FY-3D HIRAS products. Results show: (1) The precision of FY-3D HIRAS channel cloud detection is 97.19%, with a recall of 93.74%, and the Root Mean Square (RMS) error of O-B (observed brightness temperature minus background brightness temperature) caused by false clear-sky channels (cloud channels detected as clear-sky) is 0.984 K, generally within observational error variance in numerical forecasting. This confirms the method does not compromise data quality and can effectively apply to numerical weather forecasting. (2) According to CALIPSO's analysis of different cloud types, stratocumulus (St), stratocumulus (Sc), and fractured cumulus (Cu fra) exhibit high precision but lower recall. Altocumulus (Ac), altostratus (As), and deep convective clouds (DC) demonstrate high precision and recall. Cirrus (Ci) shows lower precision but higher recall.