Research On Soil Moisture And Nutrition Detection Technology Based On Diffuse Reflectance Spectroscopyresearch

Precision fertilizer is an important part of modern precision agriculture. It claim toscientific irrigation and fertilizer formulation, according to the profit and loss situation ofwater and nutrients in the soil, so as to achieve full utilization of resources, effectivelyimprove the efficiency and reduce pollution. Accurate determination of the soil componentcontent is the premise of precision fertilization. Traditional conventional soil analysis iscarried out in a chemistry lab, with long analysis cycle, high costs, and non-real-time.Compared with the traditional technology, spectroscopy research has the characteristics ofonline, quickly, without chemical treatment and can be achieved for the rapid detection of thesoil component content. This paper based on the mature UV-visible and near infraredspectroscopy, was researched on soil moisture, organic matter, available nitrogen, availablephosphorus which affect crop growth. And we improved the accuracy and reliability of theprediction model by pretreatment and external parameters orthogonal methods. The mainwork and results are as follows,(1) For the situation of moisture content gradient narrow and uneven distribution basedon the existing detection of soil moisture research with applicable spectral analysis, themethod of formula under laboratory conditions has been used to make sample moisturecontent evenly distributed between drought and saturation. In this research, we forecasted thesample moisture using a UV-visible and near-infrared spectra, optimized the spectralpretreatment methods, used PLS and artificial neural network modeling. Finally, obtain theoptimal prediction of R_p~2, RMSEP which were0.984and0.0001respectively.(2) Using near-infrared spectroscopy to detect soil nutrient content. In the research, weestablished the model which having a direct impact on crop growth, available nitrogen andavailable phosphorus prediction in the case of less previous research, and the model with soilorganic matter prediction. The pretreatment method and model calibration method had beenoptimized in the modeling process, and the optimal prediction are organic matter (R_p~2=0.910,RMSEP=0.047), available nitrogen (R_p~2=0.882， RMSEP=16.63), and availablephosphorus (R_p~2=0.812, RMSEP=5.248). (3) For the situation of the near-infrared model greatly influencing external parameters, amethod was presented. This solved the interference of external parameters when detecting soilcomposition by using the method the EPO algorithm, and optimized the modeling processwhich proved by soil moisture and temperature, these two main external parameters. Finally,the feasibility of comprehensive processing of multiple external parameters by using the EPOmethod has been verified.