Irrigation Decision-making Methods Based On Multi-source Information Fusion

The intelligent irrigation management and the wisdom of irrigation decision-making are the core of the irrigation district information,It is of great significance to improve the utilization efficiency of water resources,to realize agricultural modernization,and to accelerate the implementation of the rural revitalization strategy.In order to achieve accurate irrigation,the foundation is to access the information quickly and accurately,and the core is to make irrigation decision-making timely and scientifically.Information fusion technologies were introduced in the farmland irrigation field,a study on irrigation decision methods based on multi-source information fusion was carried out by taking winter wheat as research object,.Through the field experiment from 2014 to 2017 and laboratory analysis,the change rule of the soil moisture,leaf water potential,stomatal conductance and canopy temperature was analyzed,the estimation model of soil moisture was built based on hyperspectral,the weight coefficient of soil moisture was determined in different soil layer at the different growing stages of winter wheat,the soil moisture fusion in point and regional scale was implemented,the basic probability assignment functions of different irrigation decision-making indicators were established,the fusion of multi-source irrigation decision-making information in the feature and decision layer was completed,finally the fusion problems of multi-source irrigation information were solved.The main results were as follows.(1)The change rule of the soil moisture,leaf water potential,stomatal conductance and canopy temperature was systematic analyzed.Using the method of Leave-one-out Cross Validation,the estimation model of soil moisture was constructed based on NVI,and it was verified that the MRE and RMSE of the model were 16.73% and 0.0478 respectively.The relationship between the meteorological factors and irrigation decision-making indicators was analyzed using the path analysis method.It was found that relative humidity had the most obvious effect on leaf water potential,and the saturation deficit had the greatest impact on canopy temperature and stomatal conductance.(2)The fusion of soil moisture was realized in the data layer.Through using the adaptive weighted average fusion algorithm,the weight coefficient of soil moisture was determined in different soil layers at the different growing stages of the winter wheat,and the soil moisture fusion in point scale was implemented.The fusion of soil moisture which from three different sources was realized using the bayesian maximum entropy method,the results showed that the BME3 which including the measured soil moisture,EC-5 and the gaussian probability soft data transformed by the soil moisture inversion,had the best fusion effect,and the fusion of the soil moisture at the regional scale was achieved.(3)The basic probability assignment functions of irrigation decision indicators were built,and the fusion of multi-source irrigation decision-making information in the feature and decision layer was completed.Based on the fuzzy reasoning theory,the attribute values of each irrigation decision factor were fuzzy.Then,by using the triangle membership function,the basic probability assignment functions of soil moisture,stomatal conductance,CWSI and leaf water potential were built,and the frame of discernment of multiple fusion irrigation factors was established,which providing the basis parameters for the decision layer.(4)D-S evidence theory was improved,and the fusion of multi-source irrigation information in decision layer was implemented.On the basis of considering the actual situation of the irrigation decision-making,when the conflict between multi-source irrigation factors occurred,the distance function was introduced,and the average distance replaced the conflict evidence,thus the fusion problem of the multi-source irrigation information and the combinatorial problem of the conflict factors were solved.The improved algorithm was verified using the measured data,and its uncertainty could reduce greatly compared with the classical D-S evidence theory.