Reconstructed hyperspectral imaging for in-situ nutrient prediction in pine needles

Area of Science:

• Agricultural Science
• Remote Sensing
• Computer Vision

Background:

• Hyperspectral imaging (HSI) offers non-destructive plant nutrient analysis for forestry.
• High cost and complexity limit practical field application of HSI.

Purpose of the Study:

• To develop a cost-effective, deep learning-based method for in situ hyperspectral image reconstruction.
• To enable accurate prediction of tree needle nutrient content using reconstructed HSI data.

Main Methods:

• A deep learning model reconstructs hyperspectral images (400-1000 nm) from RGB inputs.
• Nutrient prediction utilizes reconstructed spectral data with CARS and PLSR.
• The model achieves a spatial resolution of 768×768.

Main Results:

• Accurate prediction of needle nitrogen (R²=0.8523), phosphorus (R²=0.7022), and potassium (R²=0.8087).
• Prediction accuracy comparable to traditional HSI methods.
• Successful reconstruction of hyperspectral data from RGB images.

Conclusions:

• The proposed method reduces HSI system cost and complexity.
• Enables efficient in situ nutrient detection for sustainable forestry.
• Offers a promising tool for precision agriculture and forest management.