Hyperspectral Imagery for Various Crop Growth Information Extraction

Aerial hyperspectral imagery has potential for agriculture applications. The objective of this study is to identify significant wavelength ranges (image bands or band combinations) from hyperspectral imagery for different field information extraction. The field information include corn nitrogen content, plant population, yield, and grain quality such as oil, protein, and extractable starch. All the images were processed using the GA-SPCA (Genetic Algorithm based Selective Principal Component Analysis) method. The GA-SPCA method can filter out significant image bands and reduce the image data dimension to only one principle component image through a cascade two-step dimension reduction process. It was found the most significant wavelength range for yield estimation was in the near infrared (NIR) region. For corn nitrogen content measurement, the wavelength range was located at the center of green, red, and with some scattered bands in the NIR region. The most significant wavelength range for protein content was in the green and NIR region. It is expected that information provided by this study will also helpful for future remote sensing imaging sensor design, which will provide more accurate, less expensive, and higher signal-to-noise (SNR) images with reduced image bands.