The paper conducted a systematic review of 97 scientific articles published between 2003 and 2023 to analyze current practices and identify future research directions. Key areas explored include data preprocessing techniques, hyperspectral data acquisition, data compression methods, segmentation, and the role of hardware accelerators like FPGAs and GPUs in speeding up data processing.

Hyperspectral imaging (HSI) and machine learning accurately predict gypsum tofu quality from soybean seeds. A new Yield and Texture Trade-off Theory was proposed. XGBoost outperformed other models, achieving 96-99% accuracy in classifying soybean seeds based on tofu quality.

Hyperspectral imaging combined with machine learning accurately differentiates Palmer amaranth from soybeans in field conditions. Quadratic discriminant analysis was the best model, achieving high accuracy, precision, recall, F1 score, and MCC. This research paves the way for real-time weed detection and autonomous weed management systems.

The review analyzed 60 research papers on weed detection and discussed the potential of deep learning techniques for improving weed detection.

Hyperspectral imaging (HSI) effectively differentiates soybeans from five weed species. A robotic platform captured HSI data, which was analyzed using partial least squares regression (PLSR). The best model achieved 86.2% accuracy in classification, identifying key wavelengths linked to plant chemicals. This research shows promise for automated weed control in soybean fields.

Hyperspectral imaging can effectively differentiate Palmer amaranth from soybeans. Chemometrics methods (PLS-DA, SIMCA) were used to analyze spectral data. Preliminary results show promise for using this technology to improve weed control in agriculture.