Colloidal gold, quantum dots (QDs), and upconversion nanoparticles (UCNPs), which were respectively labeled with monoclonal antibody against norfloxacin (NOR), were employed as signal probes to establish three types of immunochromatographic strips, and the strips were evaluated according to their ability to detect NOR in milk samples. The results showed that the visual detection limit of NOR was 2.0 μg L−1 in the standard solution and 20.0 μg L−1 in milk samples when using colloidal gold-immunochromatographic strips (ICS). When QD-based fluorescence immunochromatographic strips (FICS) were used, the visual detection limit was 2.0 μg L−1 in the standard solution and 10.0 μg L−1 in milk samples. If UCNP-based FICS strips were employed, the visual limit of detection was 0.5 μg L−1 in the standard solution and 2.5 μg L−1 in milk samples. Besides, the results from the proposed methods showed high agreement with results yielded by using commercial ELISA kits, indicating the good accuracy of these strips. Compared with colloidal gold-ICS, QDs-FICS showed similar sensitivity in the standard solution and a two-fold higher sensitivity in milk samples. Meanwhile, UCNPs-FICS yielded the highest sensitivity both in standard solution and in milk samples, and they were also more stable in normal laboratory conditions (that is, they could be exposed to regular room light). Additionally, the two fluorescence nanoparticle-based test strips could be built with less antibody and coating antigens and can be cost-effective. In conclusion, the proposed QDs-FICS and UCNPs-FICS could be employed as cost-effective alternative methods for sensitive, rapid, and on-site detection of NOR in milk samples.