Research On Plant Growth Status Diagnosis Method Based On Deep Learning

Plant growth process is a complex ecological process.Plant electrical signals are physiological signals that are ubiquitous in plants and have the function of transmitting information.Plant electrical signals can reflect the plant’s own physiological state in real time,mainly including physiological changes such as metabolism and material transportation.The electrical signals of plants in different growth states will show significantly different characteristics.Therefore,the growth state of plants can be monitored by monitoring the characteristics of plant electrical signals.This paper is based on a self-built plant electrical signal detection experimental platform,and uses aloe in different growth states as the experimental object.First,the BL-420 biological function experimental device is used to collect the electrical signals of aloe in different growth states,and the collected electrical signals of aloe are preprocessed,and the EMD empirical mode decomposition method is used to remove the non-stationary electrical signals of aloe.Perform smoothing processing,combined with wavelet threshold method,to remove the noise signal.Secondly,the short-time Fourier transform is used to obtain the energy spectra of aloe in different growth states,and the convolutional neural network is used to extract the features of the energy spectra.Finally,through convolutional neural network and CNN-LSTM neural network,starting from the growth and development mechanism of plants,explore the characteristic laws between electrical signals and plant growth states,and establish a model of the relationship between plant electrical signals in different growth states.For training and classification research of plant electrical signals.Classification performance of two network models: Convolutional neural network classification accuracy rate(ACC)of aloe electrical signals in different growth states is 0.9446,true positive(TPR)is 0.9327,false positive(FPR)is 0.0476,CNN-LSTM neural network The correct rate of model classification is 0.9722,the true positive is 0.9549,and the false positive is 0.0382.Experiments show that the CNIN-LSTM neural network model has a high accuracy in classifying the growth status of aloe through electrical signals.It can be used as an effective evaluation index for plant growth status detection,laying a foundation for realizing online diagnosis of plant growth status.