Research On The Method For Predicting The Moisture Content Of Living Trees Based On UHF RFID

In recent years,the utilization and protection of forest resources have become more and more important.How to effectively manage forest resources is the subject of joint research all over the world.The measurement of water content of living trees plays a vital role in the monitoring and management of living trees,and its measurement accuracy is of great significance to the development of forestry informatization industry.At present,many technical means have been used in the detection of wood moisture content,but the problems of high energy consumption,low applicability and low prediction accuracy in the detection process and results need to be further solved.Aiming at the development needs of water physiology research and forestry information technology of living trees,this thesis carried out research on water content of living trees based on passive ultra-high frequency radio frequency identification(UHF RFID),aiming at non-destructive testing and efficient identification of water content of living trees.The main ideas of this thesis are as follows:using low power RFID technology to acquire RF signal data,and collecting key environmental parameters in the measurement environment;Then,advanced machine learning technology is improved and integrated to establish the best moisture content prediction model of living wood,so as to realize non-destructive testing of moisture content of various kinds of wood.Finally,the online prediction model is verified by a large number of experiments.The results prove that the proposed scheme has the advantages of accuracy,non-destructive,practical and low power consumption,which can meet the detection requirements of the current forestry Internet of Things field.The specific research content of this thesis are as follows:(1)Summarized the current research status at home and abroad and the characteristics of passive ultra-high frequency radio frequency identification technology,designed an ultra-low cost moisture content detection scheme for living trees based on UHF RFID technology.An efficient data acquisition system was established in the campus forest station,and the host computer program was designed to conduct long-term data acquisition operations.The information such as signal strength index(RSSI)and phase in the backscatter signal was obtained,and other key environmental parameters were collected,Verified the feasibility of inverting the moisture content of standing trees using RSSI and phase.(2)Based on UHF RFID sensors and optimizing existing recognition mechanisms,an improved online sequence parallel extreme learning machine algorithm(OS-PELM)is proposed.OS-PELM optimizes the hidden layer network structure in the traditional online sequence extreme learning machine(OS-ELM),optimizes the initial parameters using the Pathfinder algorithm and adjusts the model parameters using an online neuron adjustment strategy.Based on the improved algorithm,a prediction model is established and applied in the water content prediction system.(3)Based on a large amount of data collected from laboratory and outdoor live trees,a moisture content prediction model based on the OS-PELM algorithm was used for living trees measurement and analysis.The experimental results showed that the OS-PELM algorithm had significantly improved performance compared to the traditional OS-ELM algorithm,and had higher accuracy on four different types of trees(metasequoia,poplar,pine,and beech).Through robustness experiments,it has been proven that the proposed non-destructive testing system for the moisture content of standing trees has advantages such as high accuracy,good stability,and strong robustness,which can provide practical and feasible solutions for the current forestry Internet of Things industry.For all collected samples,the MAE between the predicted and actual values of the proposed model is0.224,RMSE is 0.252,and R~2 coefficient is 0.9629.Comparing the prediction model based on OS-PELM algorithm with four other comparative algorithms(DE-OSELM,WE-OSELM,GSA-PELM,OLSPELM),the optimization strategy adopted has a significant advantage in prediction accuracy.In summary,the thesis provides strong design ideas and technical support for real-time and accurate prediction of the moisture content of living trees,and has extremely broad application prospects in the demonstration and promotion of precision forestry by the national forestry and grass administration.