Extraction And Processing Of Soil Moisture Information Based On Near Infrared Spectroscopy

Soil moisture is not only an important parameter to reflect energy flow and material exchange in the soil-plant-atmosphere system, but also an importance factors for formation, development, utilization and protection of climate, water, biological and soil resources and environment, which is an important material source and environmental conditions of biological survival. Soil moisture is an important factor affecting agricultural operations such as tillage, crop planting, plastic sheeting cover, irrigation, fertilization effect of spraying etc, and important for agricultural water-saving technology and precision agriculture technology. Soil moisture is also a necessary parameter related to agriculture, soil, environmental, engineering for scientific research. Therefore, the monitoring of soil moisture, extracting information of soil moisture content and dynamic state timely and accurately is very important. At present, as the world's growing shortage of water resources, water has a very uneven spatial and temporal distribution of water environment deterioration. China's water resources and water resources per capita is less, and because the quality of the water pollution, the uneven spatial distribution, misallocation of time, its effective water resources even more scarce. In the case of scarcity of water resources, water waste is very large, especially agricultural water use as the majority of community's total water consumption which efficiency and productivity are all low. Order to ensure the rational exploitation,utilization, protection and sustainable development of water resources, continuously improve water resources utilization and productivity, protection of agricultural water-saving technology system implementation, urgently promote a suitable practice use, control large area, high monitoring speed, accurate, effective, continuous, automatic, non-destructive soil, do not pollute the environment, economic new soil moisture detection technology, which is a long-term goal at home and abroad, and also a hot research key at home and abroad. But at present, soil moisture has been detected in the application of technology can not reach this requirement. The topics of this research is carried out around this requirement of soil moisture based on near infrared spectroscopy study, its findings have important significance on agricultural production, water conservancy, ecological environment construction and some industrial production, scientific research.The key findings of this study are::(1) Confirmed the spectral characteristics of soil are the result of multi-factors affections, soil moisture content is the control factor of the soil spectral characteristics in some characteristic bands. First use of adding soil composition material (soil organic matter and related types of clay minerals) analog testing and other analog methods (international system of classification of soil texture classification of grain size particles, water stable aggregates, soil bulk density) to research the affection of soil organic matter, soil particle size, soil clay mineral type and content, water stable aggregates, soil bulk density on soil moisture spectrum. The results showed that with the tested soil clay minerals and organic matter content increase,the diameter of water stable aggregates increase,the diameter of the soil bulk density decrease, the decrease of soil reflectance; with soil particles decrease, the reflectivity increases, but to the clay (<0.002mm) their reflectivity decrease.(2) It is designed testing of soil bulk density and salt content concentration on the impact of soil reflectance with change in soil moisture, and the affection for soil reflectance by variable soil pH. The test results are showed that in the case of bulk density and salt concentration caused changes by soil water content, the soil reflectance is also changed, with the soil drier, bulk density increases, the reflectivity also increases; soil moisture high, that is, soil salinity concentration is low, as the salt concentration increased, the reflectivity decreases, but when soil dries out of salt crystallization, with the increase of salt,its spectral reflectance increases. The experimental study also showed that the modified red soil acidity to make it into a little acid from the acid or neutral range, the change will not cause soil acidity soil spectral reflectance changed significantly. These results provide scientific basis for building the model of inversion soil moisture content from soil reflectance which take accounts of the effects of soil composition and the nature and implementation of soil modifications for soil improvement. Changes in soil bulk density is also strongly influenced the changes of soil moisture patterns types, soil moisture exist conditions, soil water movement and soil water utilization.(3) Research on implementation the relationship model of variable bulk density soil moisture and soil bulk density, confirms three-line model and exponential model to describe the relationship between soil moisture and bulk density changes,and analyze the tested soil characteristics by use of three-line model.Firstly research on the accompanied varying relationship among variable bulk density soil water content-bulk density-spectral reflectance, building two types of model, one is the exponential model for soil volume moisture content and spectral reflectance with the bulk density changes; the other is the surface model for the soil water content,spectral reflectance and bulk density based on the mechanism analysis of dual function affection of soil water content and bulk density for spectral reflectance. The former is reflected the affection of bulk density change on the relationship between soil water content and spectral reflectance; the latter is reflected that in the case of spectral reflectance influenced by dual-function of soil water content and bulk density, soil moisture inversion models by reflectance with change bulk density and soil water,which provides a theoretical basis for mechanism analysis and model construction of soil moisture content inversion by variable bulk density reflectance.(4) In the case of extraction information of soil water, comparatively study of several spectral data preprocessing methods, which include the elimination of noisy band, filtering (moving average smoothing, median smoothing filter, SG convolution smoothing, wavelet transform de-noising method), the derivative method etc. The results showed that:the use of combination of wavelet de-noising method and first order derivative to preprocess spectral data can achieve better filtering results than other methods, and also to keep the details of the desired signal characteristics.(5) Chosen the features band spectral to reflect well of soil moisture characteristics in the 1100nm-2400nm spectral range, through selecting of 11 features bands to be determined, the final selection of 1400nm,1900nm,2050nm,2200nm 4 bands as a soil water features spectrum band。Among which, in the area with soil moisture decrease, the reflectivity increasing,occurs the 1400nm and 1900nm apparent absorption peak; when the soil moisture is low, a significant absorption peak in the 2200nm also exist.(6) Explored the impact of soil water content and spectral reflectance representation on these two relationship modeling, and choose better matching method. About water content,choice of the weight water content, invariable bulk density volumetric water content, variable bulk density volumetric water content; about spectral reflectance, choice of measured spectral reflectance, relative reflectance, reduced soil reflectance and normalized reduced soil reflectance.With the comparison, results show that variable bulk density volumetric water content is matched well with the relative reflectance and reduced soil reflectance, while variable bulk density volumetric water content is matched better with normalized reduced soil reflectance. In construction model for inversion soil reflectance to soil moisture, first proposed and the use of variable bulk density soil moisture, reduced soil reflectance and normalized reduced soil reflectance, and the experiments results indicated this representation is acquired good affection.(7) Using two modeling methods of BP neural network and support vector machines to research on model construction for the relationship between soil water content and reflectance, and to study the model adaptability for the relationship between soil water content and reflectance. In the case of soil moisture content inversion model construction from soil reflectance, using the wavelet transform to extract signal characteristics in both time- and frequency domain, and thus can more fully reflect the signal characteristics; to remove the other characters of soil effect on the soil reflectance inversion soil water content,the change in signal characteristics are normalized by relatively dry soil feature vector and constructed three feature vectors:the relative characteristics of the original signal value variation, the first order derivative signal corresponding characteristics, the relative first order derivative signal characteristics variation; using Genetic algorithm in BP neural network and support vector machine parameters effective optimization, and use three kinds of feature vectors in the two models modeling and simulation results show that using the relative characteristics change of the first order derivative signal feature vector as input signals, genetic algorithm optimized support vector machine parameters is better for model construction of soil weight water content and volumetric water content prediction based on NIRS in the case of small sampling group.