The Visible Light Indoor Positioning System (VL-IPS) has been a popular research area recently. In VL-IPS, many localization methods have been proposed by leveraging the Received Signal Strength (RSS) based trilateration. However, the traditional RSS based trilateration localization (RSS-TL) method is very sensitive to the lighting environment and would results in a big localization error due to the presence of non-line-of-sight (NLOS) light signal. In light of this, we propose a novel NLOS-aware localization algorithm, namely Enhanced Fingerprinting-aided RSS-TL (EFP-RSS-TL). It permits to improve the localization accuracy for the corner regions of a room by eliminating the NLOS impact while keeping the same high accuracy for the room center. This is achieved by leveraging a RSS fingerprint database which records the line-of-sight (LOS) light power ratios beforehand. For validation purpose, we built a real VL-IPS platform and implemented the proposed algorithm. Experimental results show that the proposed NLOS-aware EFP-RSS-TL algorithm enables to reduce significantly the average positioning error (by up to 79%) compared to its counterparts. Besides, our proposal cuts the database size by 50% and is more robust to environment changes. In a room of 4.7 m x 2.7 m, the achieved average positioning error is around 6 cm when it is vacant and it is no more than 14.5 cm when it is occupied by several people.