Research Of Soil Parameter Retrieval With Fuzzy Logic Via UWB Signals

Soil parameter retrieval method becomes the key technology of the soil parameter acquisition and surveillance in precision agriculture.Though the retrieval techniques have been deeply researched for decades,the expensive,imprecise and inefficient methods restrict their widely applications in agriculture.To solve this problem,we concentrate on the retrieval methods of soil moisture and propose a soil database-construction technique in this paper.The database is constructed by the features extracted from soil signals with known soil moistures.The unknown soil moisture of the soil signals will be recognized according to the comparisons and classifications in these data.Then we propose a new soil signal collection method with ultra wide band(UWB)radar.The high resolution,low cost of UWB radar makes the signal collections fast and convenient.What's more,this paper applies the fuzzy logic system(FLS)to further simplify the soil moisture retrieval algorithms.FLSs are utilized to describe the uncertainty and extract features in the soil signals.Then,the machine learning algorithms are proposed to classify the extracted features in FLS.On the other hand,due to the joint time-frequency analysis(JTFA)is adapted at describing the frequency variation of the soil signals,we construct a database with the time-frequency distribution patterns and utilize the deep learning algorithms to classify the patterns corresponding to different soil moistures.Finally,three soil algorithms,soil moisture retrieval method with FLS forecasting templates via UWB radar,soil moisture retrieval with FLS feature extraction and ML classifications via UWB radar,soil moisture retrieval with JTFA and deep learning classifications via UWB radar are proposed according to the verifications with mounts of measured soil data.The main contents are as followed.A flexible,fast and efficient soil signal collection method with UWB radar sensor is proposed.A new soil moisture retrieval algorithm,soil moisture retrieval method with FLS forecasting templates via UWB radar is proposed.The algorithm utilizes the FLS to forecast the soil signals with known soil moistures and save the forecast signals as forecasting templates into database.Comparing with the templates corresponding to different soil moistures,the unknown signals can be classified and recognized with the right soil moistures.Utilizing this algorithm,The system generate the forecasting templates utilizing type-1 fuzzy logic system(T1FLS)and non-singleton fuzzy logic system(NT1FLS),respectively.Then the system compares the soil signals with forecasting templates due to root mean square error(RMSE)and mean absolute deviation,respectively.Generally,four soil moisture recognizing systems are constructed.According to the classify of soil moisture with measured signals,we verify both the four soil moisture retrieval system can precisely classify the soil signals and recognize their moistures.Whats' more,The NT1FLS+MAD system expresses a better performance when the noise interferes.We propose the soil moisture retrieval algorithm with FLS feature extraction and machine learning classifications via UWB radar.The algorithm extracts soil signals' features with FLS forecasting system and constructs the soil feature database.Then the soil signals' features are classified applying machine learning algorithm.The unknown soil signals are classified and recognized with different soil moistures in the database utilizing the trained machine learning system.Due to the algorithm,we apply T1 FLS and adaptive networkbased fuzzy inference system(ANFIS)in FLS feature extraction system,respectively.Then Artificial Neural Network(ANN)and random forest(RF)are respectively applied in classification stage.Generally,four soil moisture recognition system(ANFIS+ANN,ANFIS+RF,T1FLS+ANN and T1FLS+RF)are constructed.We verify the systems with measured data,results show all the system can reach high correct recognition rates(CRR)on soil moisture.The ANFIS+RF recognition system can maintain a high CRR when the noise deteriorates.We propose the soil moisture retrieval algorithm with JTFA and deep learning classifications via UWB radar.The algorithm transforms the soil signals into time-frequency distribution patterns and utilizes the deep learning algorithm to classify the patterns with soil moistures.we totally construct four soil moisture retrieval systems.In time-frequency distribution,Wigner-Ville distribution and Choi-Williams distribution are respectively used.In classification,the deep learning algorithm,VGGNet and AlexNet is respectively applied to classify the patterns with different soil moistures.Simulations with measured data show the system can reach an excellent performance when the soil moistures are in large data quantity.