Combined with the spaceborne lidar active observations， an Arctic summer cloud detection model is studied here based on the data from the FY-3D/MERSI-II （FengYun-3D/Medium Resolution Spectral Imager-II）. By using probability density function analysis method and introducing loss rate to optimize the correlating thresholds， an infrared cloud detection model for the Arctic summer is developed based on the confidence levels. The validation results reveal that the cloud detection results are highly consistent with the matched spaceborne lidar observations. The high confidence levels basically represent the cloudy pixels， while the low values correspond to the clear ones. The case study shows that the cloudy pixels is 100% consistent with the pixels of the confidence level higher than 0.8. When the confidence level is lower than 0.2， 10.15% of the cloudy pixels are still misjudged as clear pixels， which are primarily single-layer clouds with the cloud top heights between 4 and 6km. This may be caused probably by the cirrus clouds， which needs further study.