Study On The Spectra Characteristic Of Different Texture Soil And The Discussion Of Automation Analyzed Instrument

Precision agriculture is one of the most important aspects for agriculture development in 21stcentury. It would promote the transformation from extensive agriculture to the knowledge andtechnology-based one. But the precondition of the action of the precision agriculture was theachievement of the automation information collection of the field. The soil was a limited resourcethat was essential to many lives on the earth and the soil science was performing more and moreimportant to the continual agriculture. The success of the 'Digital soil' might organize soil theinformation systemically to improve the utilization of the soil information which was necessary tothe information society. The settlement of many complex soil problems was depended on the correctrecognition of soil; however, from thousands ago the recognition of the soil was depend on theconfirmation of the experts and was difficult for popularization. In recent years, with thedevelopment of the spectroscopy technology, the in and abroad scholars applied this technology forthe nutrient information acquirement of the soil, but the contentment of N, OM, pH were famousand the P, K were unsatisfactory. The researches in the instruments of 'Digital soil' were weakness.In this paper, aiming at solving these problems mentioned above and making use of theapparatus in our laboratory, the experiments of researching the varieties and nutrient of the soil wereperformed. The primary contents were as following:Firstly, the spectra properties of four typical soils in the zhejiang province were analyzed.Chemometric technologies combined with the artificial neural network were used for dataprocessing, and at last the recognition rate was nearly 100%. Then we chose sensitive wavelengthsaccording to the standard error of the original spectra data of every wavelength, in the end 130wavelengths in the vis-NIR area and 170 ones in the NIR area were chosen for substitution of theoriginal 600 and 1103 wavelengths. With the sensitive wavelengths the recognition effect wassimilar to the original data set and it provided theoretic for the invention of the spectra soilinstrument.Secondly, the spectra properties of two soil nutrition of different varieties were analyzed. Inthis part different chemometric technologies such as SVM et al. were chosen for optimizing themodels established by the pre-processing NIR and MIR spectra data sets. In the end, the correlationcoefficient of available nitrogen (N) was 0.876 with the NIR spectra, and the correlation coefficient of available phosphorus (P) and available kalium (K) were 0.839 and 0.803 with the MIR spectra.So we considered that the infrared ray was suitable for the soil nutrition analyzed and the MIR wassuper to NIR in the soil available phosphorus and available kalium analysis. Based on these analyse,using the higher correlation coefficients the sensitive wavelengthes of the available nitrogen (N),available phosphorus (P) and available kalium (K) were chosen successfully.Thirdly, in this paper the automation analyzing instrument of the soil properties was discussedfrom the inside structure to the factors which could infect the instrument performance, and itprovided consulted value for the invention of the spectra soil instrument.