Study On The Effect Of Spectral Transformation On The Accuracy For Diagnosing Soil Nitrogen

The transformations of hyperspectral reflectance which are important pre-treatment and procedure can eliminate the effect of background and noise and extract the spectral feature. In order to study the effect of spectral transformation on the accuracy for diagnosing soil nitrogen and determine the optimal spectral transformation to rapidly diagnose the soil nitrogen, the soil nitrogen and soil spectrum were measured and soil spectral reflectance were transformed based on the sixteen mathematical algorithms. Moreover, the correlation relationships analysis between the transformed spectrum and soil nitrogen were analyzed, and the performance for the models of soil nitrogen based on the transformed spectrum was evaluated to determine the optimal spectral transformation and diagnose the soil nitrogen. The result showed that:(1) The correlation relationship between the soil nitrogen and transformed spectrum (T5, T6, T7, T8, T9, T10, T11, T12 and T15) increased obviously, and the correlation coefficient could reach to the highest for the spectral transformations of T8, T11 and T12.(2) Considering both the accuracy and complexity of the soil nitrogen model, we found that the spectral transformations of T2, T4, T6, T7, T8, T10, T11, T12, T13, T14 and T15 could improve the model accuracy and optimal the monitor model. Moreover, the monitor model based on the transformed spectrum of T8 and T11 had a higher accuracy in monitor soil nitrogen and fewer factor numbers at the same time.(3) The result also was found that the higher correlative relationship was determined between the soil nitrogen and transformed spectrum, such as T5, T6, T7, T8, T10, T11, T12 and T15, the fewer optimal factor variables were selected to enter the monitor model of soil nitrogen.Overall, the spectral transformation of T8 was recommended to process the original spectral reflectance and the transformation of T6, T7, T11 and T12 could also be considered. The study will support some theory technique and reference in processing the hyperspectral data.