Soil classification based on near infrared reflectance spectroscopy and orthogonal signal correction-partial least square

Soils with different soil textures have different spectra because of their different moisture-holding capacity and particle size. This feature provides a scientific basis for soil textural classification by Near-infrared Spectroscopy (NIR) technology. In the paper, the spectra of three different soil textural samples (sand, loam and clay soils) were analyzed. Since texture information in spectra is less than chemical information, the orthogonal signal correction (OSC) was applied in this research because it can keep down concentration information by orthogonal processing, and the partial least square (PLS) classification model was obtained. The results showed that the correlation coefficient of validation model was 0.946, and the correct recognition ratio for the three kinds of predicted samples were 93%, 86.6% and 86.6% respectively. The research indicates that OSC can extract texture information from weak spectra, so as to realize soil texture classification.