*2.1. Experimental Method*

The experiment took tomato leaf mildew samples as the research object and collected test samples with different percentages of diseased spot areas. Using a hyperspectral imaging system and THz time-domain spectral measurement system, the near-infrared spectrum, power spectrum, and absorbance time-domain spectral information of samples with different grades of disease were collected. Algorithm optimization was used to remove interference, remove redundancy, and perform feature extraction. Finally, based on the extracted spectral feature data of different grades of disease, single-dimensional and multidimensional fusion tomato leaf mildew recognition models were established, respectively. The specific process is shown in Figure 1.

**Figure 1.** Flow chart of the experiment.
