In recent years, demands of pervasive smart services and applications increase explosively. Device-free human detection through sensors or cameras has been widely adopted but with privacy issues as well as misdetection for motionless people. To resolve these defects, channel state information (CSI) captured from commercialized Wi-Fi devices is capable of providing plentiful signal features for accurate detection. The existing systems has inaccurate classification under a non-line-of-sight (NLoS) and stationery scenario of a person standing still at corner in a room. In this work, we have proposed a colorization and contrastive learning enhanced NLoS human presence detection (CRONOS) system. CRONOS is capable of generating dynamic recurrence plots (RPs) and coloring CSI ratios to distinguish mobile people and vacancy of a room, respectively. Furthermore, supervised contrastive learning is conceived to retrieve substantial representations, where consultation loss is formulated to differentiate the representative distances between dynamic and stationery cases. Furthermore, a self-switched static feature enhanced classifier (S3FEC) is proposed to determine the utilization of either RPs or coloring CSI ratio. Finally, comprehensive experimental results have revealed that our proposed CRONOS outperforms the existing systems applying machine learning, non-learning based methods as well as non-CSI based features in open literature, which achieves the highest presence detection accuracy and moderate computational complexity in vacancy, mobility, LoS and NLoS scenarios.