Detection Of Paddy Water Characteristic Parameters Based On Spectral Technology

Detection of paddy water quality,especially dissolved organic matter,is one of the basic needs of intelligent agriculture and digital agriculture.In this paper,the use of three-dimensional fluorescence spectroscopy to detect paddy water quality is a fast and efficient detection method.This paper discusses the theory of fluorescence spectroscopy,including the theory of spectrum and prediction model,and provides a theoretical basis for water quality detection.The spectral characteristics of dissolved organic matter in paddy water were studied.Analyze the water quality evolution process during the rice growth cycle,and plan the water quality sampling time and method.Field sampling of paddy water in Taonan research area was carried out to carry out the determination experiment of three-dimensional fluorescence spectrum.Fluorescence characteristic parameters were extracted to analyze the source and humification degree of dissolved organic matter in paddy water.The integral value of fluorescence region was calculated to characterize the cumulative fluorescence response of dissolved organic matter in paddy water.Aiming at the problem of fluorescence peak overlap in the fluorescence response,the parallel factor analysis model was used to identify the fluorescent components.In order to find the sensitive wavelength of fluorescence spectrum and water quality characteristic parameters,the correlation coefficient between fluorescence spectrum related parameters and water quality characteristic parameters was analyzed.The results showed that the dissolved organic matter in paddy water was dominated by endogenous input,supplemented by exogenous input,and the degree of humification was low.There were three fluorescent components in paddy water,and the stability of the components was ranked as terrestrial humus,fulvic acid terrestrial microbial humus,and protein-like substance(tryptophan),which was consistent with physicochemical analysis.The correlation coefficient between bimodal fluorescence intensity ratio and water quality characteristic parameters is stronger than that between single peak and corresponding indexes.The prediction of basic water quality indicators based on fluorescence spectroscopy was carried out.Draw the maximum correlation coefficient curve of water quality characteristic parameters at different excitation wavelengths,and select the optimal excitation wavelength;the correlation coefficient matrix of dual-band fluorescence intensity ratio and water quality characteristic parameters was further constructed,and the wavelength ratio with the largest correlation coefficient was selected to establish a linear regression model.Three machine learning algorithm models were established.The fluorescence emission spectrum of the optimal excitation wavelength was used as the input layer,and the water quality characteristic parameters were used as the output layer.The prediction results were obtained and compared with the traditional water quality detection results to evaluate the accuracy of the model.The results showed that the determination coefficients of the Bayesian optimized convolutional neural network-gated recurrent unit algorithm in the chemical oxygen demand and total phosphorus indexes were greater than0.91,and the combined prediction deviation ratio was greater than 2.5.The prediction model had the best accuracy,which proved that the three-dimensional fluorescence spectrum could be used for paddy water quality assessment.The fluorescence spectrum research method in this paper has the characteristics of fast,convenient,non-destructive and multi-parameter analysis,which is conducive to the treatment and rational utilization of paddy water,and has certain guiding significance for rice growth and cultivation.