This article proposes a contrastive learning-based scene classification method to extract features under occluison circumstance. Novel occlusion features analysis methods are proposed also.