**3. Results**

#### *3.1. Analysis of Canopy Spectral Response of Corn during Growth Periods*

The original reflectance spectrum of the corn canopy is shown in Figure 3a. The figure demonstrates serious noise-point information in the spectral curve. The noise point of the spectral curve was significantly reduced after the S-G filtering (Figure 3b). The scattering effect of the sample reflection spectrum was significantly improved after the SNV correction (Figure 3c). The average spectral curves of three growth stages are shown in Figure 3d.

In general, 400–500 and 611–710 nm were two low-reflectance regions in the visible light band due to the strong absorption of blue and red light by leaf pigments. The absorption valleys appeared near 400 and 680 nm. Approximately 520–610 nm was a high-reflection area due to the strong reflection by the leaf pigments of green light. The reflection peak appeared near 550 nm. In the near-infrared region, the reflectance sharply increased from 711 nm to 760 nm due to the large cavity of the reflective surface in the spongy tissue structure of the mesophyll, thereby showing a "rapid climb" trend. The 761–1000 nm region was a strong reflection area, and the curve was close to horizontal, thereby showing a "high reflection platform". A weak absorption valley appeared around 970 nm due to the absorption of water.

Figure 3d demonstrates that the different growth periods varied in the four spectral ranges of 325–400, 401–700, 761–970, and 971–1075 nm. The spectral reflectance increased with the growth period in the ranges of 325–400 and 761–970 nm. The reflectance decreased with the growth period in the ranges of 401–700 and 971–1075 nm.

**Figure 3.** Corn canopy reflectance spectral curve. (**a**) Original canopy reflectance spectra; (**b**) canopy reflectance spectral after the S-G filtering; (**c**) canopy reflectance spectra after the S-G filtering and SNV; (**d**) average canopy spectra of three growth stages.

#### *3.2. Statistical Analysis and Sample-Set Division*

The trend of the average chlorophyll content with the growth period is shown in Figure 4, which demonstrated an increase from G1 to G3. From G1 to G3, the variation ranges of the chlorophyll content between samples gradually concentrated. The SPXY algorithm was used to divide the sample set according to the ratio of 2:1. The division result is shown in Table 1. One hundred and forty-four samples were included in the modeling set to establish a chlorophyll content detection model, and 72 samples in the verification set to test the performance of the detection model. The range of chlorophyll content of the samples in the modeling set was larger than that in the verification set. Thus, the sample set obtained by the SPXY algorithm was reasonable and was used for subsequent modeling.

**Figure 4.** Statistical box line graph of chlorophyll content of the corn growth stages.


**Table 1.** Statistical results of the calibration set and validation set (%).
