**Background:** Nitrogen is the primary limiting nutrient for crop yield, and maize varieties differ substantially in nitrogen use efficiency (NUE). Spectral vegetation indices from UAV-mounted multispectral sensors offer a non-destructive method to monitor crop growth and nitrogen status. However, the relationship between canopy spectral indices and growth indicators across maize varieties with contrasting N efficiencies has been underexplored. This study aimed to evaluate how nitrogen supply affects vegetation indices and growth parameters in four N-efficiency maize types and to quantify correlations between spectral indices and agronomic traits.
**Methods:** The field trial was conducted in 2021 in Yuzhou, Henan, China (113°56′E, 34°09′N) on a wheat-maize rotation system with soil pH 7.5, organic matter 24.8 g/kg, alkaline nitrogen 98.34 mg/kg, available potassium 152.22 mg/kg, and available phosphorus 30.76 mg/kg. A split-plot design with three replicates was used: main plots received five nitrogen levels (0, 90, 180, 270, 360 kg/hm² as N0–N360), and sub-plots contained four maize varieties—ZD958 (low-N-efficient), XY335 (high-N-efficient), QL368 (double-high), and YD606 (double-inefficient). Planting density was 67,500 plants/hm² with plot size 6.0 m × 3.6 m. All phosphorus and potassium (150 kg/hm² each) were applied at jointing. Sowing was June 15, 2021; harvest September 29, 2021. At the trumpet, filling, and maturity stages, three representative plants per plot were sampled, separated into organs, dried at 80°C to constant weight, and weighed. Nitrogen content was determined via the Kjeldahl method using a continuous flow analyzer (AA3, SEAL-Analytical, Germany). Yield was measured from the middle two rows of each plot, threshed, and adjusted to 14% moisture. A DJI Matrice 600 Pro UAV with a MicaSense Rededge-Altum multispectral camera (five bands: red, green, blue, red edge, near-infrared; 3.45 μm pixel; 2064×1544 resolution; 48×36.8° FOV) collected canopy imagery at the same three growth stages under clear, windless conditions between 10:00–14:00 h, with 80% heading and side overlap. Images were processed in Pix4D and ENVI 5.3 to extract NDVI, GNDVI, GOSAVI, and RVI. Statistical analyses used SPSS 19.0 with Duncan’s test for significance.
**Key Results:** Yield increased with nitrogen application then stabilized. Under N0, N90, N270, and N360, ZD958 yield was 83.15%, 15.89%, 11.34%, and 16.83% higher than YD606, respectively; XY335 was 28.55%, 0.19%, 1.33%, and 7.40% higher; QL368 was 46.70%, 19.29%, 27.26%, and 34.26% higher. Under medium and high nitrogen, QL368 had the highest yield. Nitrogen fertilizer and cultivar effects on yield were both significant (F = 61.84* and 19.12*, respectively). Leaf nitrogen content at the filling stage showed significant cultivar differences (F = 9.56**), with YD606 consistently lowest. Dry matter accumulation at filling and maturity was significantly affected by cultivar (F = 5.75** and 8.45**, respectively) and nitrogen (F = 33.21** and 26.90**), with QL368 highest and YD606 lowest. All four vegetation indices (NDVI, GNDVI, GOSAVI, RVI) increased with nitrogen and were highest in QL368 under medium and high nitrogen. At the filling stage, correlations between vegetation indices and yield were strongly positive: NDVI r = 0.775–0.963, GNDVI r = 0.809–0.970, RVI r = 0.754–0.927, GOSAVI r = 0.809–0.960. Correlations with leaf nitrogen content at filling were also strong: NDVI r = 0.839–0.979, GNDVI r = 0.871–0.958, RVI r = 0.851–0.960, GOSAVI r = 0.873–0.957. Correlations with dry matter at filling: NDVI r = 0.772–0.942, GNDVI r = 0.774–0.927, RVI r = 0.787–0.878, GOSAVI r = 0.777–0.927. At maturity, correlations were often negative or non-significant.
**Clinical Implications:** The study demonstrates that UAV-based multispectral vegetation indices, particularly GNDVI and GOSAVI at the filling stage, can reliably predict yield, dry matter, and leaf nitrogen content across maize varieties with different nitrogen efficiencies. The optimal nitrogen application rate for maximizing yield was identified as 270–360 kg/hm². These findings support the use of spectral monitoring for precision nitrogen management and for screening high-N-efficient maize varieties under non-low nitrogen conditions. The approach could reduce reliance on destructive sampling and enable real-time crop status assessment in breeding programs and agricultural practice.