Study On Real-time Nondestructive Acquisition Method Of Standing Tree Stem Water Content And Its Applications

Plant water physiology information is closely related to plant abiotic and biotic stresses.Better acquisition of plant water physiology information in forest regions is conducive to the healthy development of agroforestry.In order to realize the whole seasonal continuous monitoring of standing tree stem water content and overcome the adverse effects of the harsh environment on the monitoring equipment in forest area,this paper proposed a real-time nondestructive acquisition method of standing tree stem water content which was applied to the analysis of abiotic and biotic stresses in plants.The main research contents,methods and conclusions of this paper are as follows:(1)In order to meet the demand of nondestructive measurement of stem water content outside woody heartwood,this paper proposed a method for measuring stem water content of standing tree based on standing wave ratio.Firstly,the measurement circuit of the sensor was designed;the measurement frequency with 100MHz of the sensor was determined;the sensor that can be used for field measurement was assembled.Subsequently,the sensitive distance was determined(approximately 53mm in axial direction,20mm in radial direction);the calibration equations between sensor output and related factor(including dielectric constant,electric conductivity,temperature,stem diameter)were fitted;the feasibility of detecting freezing point of solution by the sensor was verified.Finally,the measurement error of the sensor was analyzed based on oven-drying method(average error=0.008cm3cm-3).(2)Because the forest region data acquisition system is limited by storage space and operation power,this paper proposed a forest region data acquisition method based on compressed sensing.Firstly,the performance of the compression algorithm based on DFT,DCT and learning dictionary was analyzed.It can be concluded that when the sparsity K is the same,the data compression ratios are sorted as learning dictionary,DFT,DCT in descending order;as the sparsity K increases,the data compression ratios of all three decrease.Subsequently,the compressed sensing algorithm was implemented in the hardware acquisition system based on DFT and K=16.Compared with the conventional acquisition system,the data compression ratio of this system reaches 4.24.The total power dissipation and transmission power dissipation decrease by 75.72%and 13.62%respectively.(3)In order to solve the problem of continuous missing of data in the process of collecting stem water content time series of standing tree in forest region,this paper proposed a method of filling time-series data based on deep learning.Firstly,two unidirectional RNN models and three bidirectional RNN models were built based on simple RNN model.Among them,the bidirectional segmented RNN model has the best performance with an average prediction error of 0.014cm3cm'3.Subsequently,five similar LSTM models were built based on simple LSTM model.Among them,the bidirectional decreasing weight LSTM model has the best performance with an average prediction error of 0.009cm3cm-3,which is very close to the sensor's average measurement error of 0.008cm3cm-3.The prediction accuracy can basically meet the requirements of subsequent experimental analysis.(4)In order to better apply standing tree stem water content to the analysis of abiotic and biotic stresses in plants,firstly,the characteristics of stem water content were interpreted based on the microenvironment and physiological parameters.And an estimation model between the microenvironment parameters and stem water content was established based on the oblique ellipse.Subsequently,the variation rule of stem water content was analyzed in Pachira macrocarpa under drought stress;the variation rule of stem water content was analyzed in poplar under cold stress;the variation rule of stem water content was analyzed in Lagerstroemia indica under disease and pest stress;the supervised and unsupervised models for early diagnosis of plant disease and pest level were established based on the characteristics of stem water content.This paper innovatively proposed a real-time nondestructive detection method of plant vital signs based on standing tree stem water content,which can provide an available nondestructive quantifiable indicator for evaluating drought,cold and disease stresses.