Among the fundamental problems of soybean cultivation, damage from selective and non-selective herbicides used in conventional agriculture continues to occur. In particular, damage from herbicides and nutrients during early growth is difficult to classify even through visual inspection by experts, so analysis is being conducted in a laboratory environment. As such, analysis in a laboratory environment has a delayed response time and requires a lot of manpower. To solve this problem, we developed a technology to detect herbicide damage using hyperspectral imaging. To diagnose damage, Daechan beans were treated with the most commonly used herbicides, foliar treatments, bentazone and glufosinate, and soil treatments, Linuron and Alachlor. Germination of the soil treatment agent began 10 days after soil treatment, and analysis was performed based on when the cotyledons were fully unfolded. Reneuron showed the highest classification accuracy when using 785nm and 890nm wavelengths, and Alachlor showed the highest classification accuracy when using 788nm and 944nm wavelengths. For Reneuron, the classification accuracy is 88.49% (1x), 89.51% (2x), 90.31% (4x), 99.99% (8x), and 100% (16x), and for Alachlor, the classification accuracy is 63.41% (1x). They were 82.82% (2x), 95.06% (4x), 96.62% (8x), and 96.71% (16x). The foliage treatment agent was applied during the third leaf stage of the crop growth season, and images were measured and analyzed for 10 days. Glufosinate showed the highest classification accuracy when using 1387nm and 1563nm wavelengths, and bentazone showed the highest classification accuracy when using 1408nm and 1493nm wavelengths. For glufosinate, the classification accuracy was measured at 77.97% (1x), 93.52% (2x), 93.70% (4x), 96.70% (8x), and 97.53% (16x), and for bentazone, the classification accuracy was 87.06% (1x), 94.04% (2x), 93.72% (4x), 93.02% (8x), and 93.64% (16x).