CN114118157A - A method and system for diagnosing light information based on plant electrical signals - Google Patents

A method and system for diagnosing light information based on plant electrical signals Download PDF

Info

Publication number
CN114118157A
CN114118157A CN202111441798.XA CN202111441798A CN114118157A CN 114118157 A CN114118157 A CN 114118157A CN 202111441798 A CN202111441798 A CN 202111441798A CN 114118157 A CN114118157 A CN 114118157A
Authority
CN
China
Prior art keywords
plant
imf
vectors
light intensity
electrical signals
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111441798.XA
Other languages
Chinese (zh)
Inventor
田立国
李猛
王岳松
刘金奇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University of Technology and Education China Vocational Training Instructor Training Center
Original Assignee
Tianjin University of Technology and Education China Vocational Training Instructor Training Center
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin University of Technology and Education China Vocational Training Instructor Training Center filed Critical Tianjin University of Technology and Education China Vocational Training Instructor Training Center
Priority to CN202111441798.XA priority Critical patent/CN114118157A/en
Publication of CN114118157A publication Critical patent/CN114118157A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Signal Processing (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Cultivation Of Plants (AREA)

Abstract

本发明公开了一种基于植物电信号的光照信息诊断方法及系统。该方法包括：采集不同光照程度下样本植物电信号，构建训练集；通过所述训练集对混合神经网络模型进行训练，得到植物电信号与光照强度之间的关系模型；基于当前植物的电信号数据，根据所述关系模型确定当前植物的光照强度；判断当前植物的光照强度是否在最佳光照强度预设范围内；若否，则对当前植物的光照强度进行调节。本发明得到了植物电信号与光照强度之间的耦合规律，通过当前植物电信号信息，能够立即得到植物所处环境中的光照强度，通过调整周围环境光照大小，使植物处于最优生长状态。

Figure 202111441798

The invention discloses a method and system for diagnosing illumination information based on plant electrical signals. The method includes: collecting sample plant electrical signals under different illumination levels, and constructing a training set; training a hybrid neural network model through the training set to obtain a relationship model between plant electrical signals and light intensity; based on current plant electrical signals According to the relationship model, determine the light intensity of the current plant; determine whether the light intensity of the current plant is within the preset range of the optimal light intensity; if not, adjust the light intensity of the current plant. The invention obtains the coupling law between the plant electrical signal and the light intensity, and through the current plant electrical signal information, the light intensity in the environment where the plant is located can be obtained immediately, and the plant is in an optimal growth state by adjusting the light level of the surrounding environment.

Figure 202111441798

Description

Illumination information diagnosis method and system based on plant electric signals
Technical Field
The invention relates to the technical field of information agriculture, in particular to an illumination information diagnosis method and system based on plant electric signals.
Background
To judge the growth state of the plant, firstly, an evaluation method and an index system of the growth state of the plant are established, and the quality of the growth state of the plant is accurately judged. The traditional morphological index and physiological and biochemical index are complex to measure and have certain hysteresis. In the prior art, a computer vision is also used for collecting physiological shape images of plants and analyzing and judging the growth vigor of the plants. However, this method is limited to pest control and detection, and has low image recognition accuracy and certain hysteresis in information processing.
Disclosure of Invention
The invention aims to provide a method and a system for diagnosing illumination information based on plant electric signals, which adjust the illumination intensity of the current environment according to the coupling relation between the plant electric signals and the illumination intensity in environmental factors so as to enable plants to be in the optimal growth state.
In order to achieve the purpose, the invention provides the following scheme:
a method for diagnosing illumination information based on plant electric signals comprises the following steps:
collecting sample plant electric signals under different illumination degrees, and constructing a training set;
training the mixed neural network model through the training set to obtain a relation model between plant electric signals and illumination intensity;
determining the illumination intensity of the current plant according to the relation model based on the electric signal data of the current plant;
judging whether the illumination intensity of the current plant is within the preset range of the optimal illumination intensity;
if not, adjusting the illumination intensity of the current plant.
Optionally, the acquiring electrical signal data of the sample plant under different illumination degrees, and constructing a training set specifically includes:
dividing the sample plant electrical signals according to the time length, wherein the length of each section of the sample plant electrical signals is L;
carrying out empirical mode decomposition on each divided section of sample plant electric signal to obtain n IMF vectors;
selecting the IMF vectors with high first m relation numbers from the n IMF vectors;
splicing the m IMF vectors into a plurality of signal number sequences, and converting all the signal number sequences into a two-dimensional matrix of a time domain and a frequency domain by using short-time Fourier transform; the two-dimensional matrix constitutes a training set.
Optionally, the method further comprises: and eliminating interference noise in the m IMF vectors by adopting an improved wavelet threshold method.
Optionally, the removing, by using an improved wavelet threshold method, the interference noise in the m IMF vectors specifically includes:
judging whether the wavelet coefficients of the m IMF vectors are larger than a coefficient threshold value or not;
if yes, reserving;
if not, deleting.
Optionally, the hybrid neural network model comprises a convolutional neural network and a long-short term memory cycling neural network.
The invention also provides an illumination information diagnosis system based on the plant electric signals, which comprises:
the training set construction module is used for collecting the plant electrical signals of the samples under different illumination degrees and constructing a training set;
the training module is used for training the mixed neural network model through the training set to obtain a relation model between the plant electric signals and the illumination intensity;
the current plant illumination intensity determining module is used for determining the illumination intensity of the current plant according to the relation model based on the electric signal data of the current plant;
the judging module is used for judging whether the illumination intensity of the current plant is within the preset range of the optimal illumination intensity;
and the adjusting module is used for adjusting the illumination intensity of the current plant when the illumination intensity of the current plant is not within the preset range of the optimal illumination intensity.
Optionally, the training set constructing module specifically includes:
the dividing unit is used for dividing the sample plant electric signals according to time length, and the length of each section of the sample plant electric signals is L;
the empirical mode decomposition unit is used for carrying out empirical mode decomposition on each divided section of the plant electric signal to obtain n IMF vectors;
a selection unit, configured to select the IMF vectors with the first m high correlation numbers from the n IMF vectors;
the training set constructing unit is used for splicing the m IMF vectors into a plurality of signal number sequences and converting all the signal number sequences into a two-dimensional matrix of a time domain and a frequency domain by using short-time Fourier transform; the two-dimensional matrix constitutes a training set.
Optionally, the training set constructing module further includes: and the elimination unit is used for eliminating the interference noise in the m IMF vectors by adopting an improved wavelet threshold method.
Optionally, the eliminating unit specifically includes:
a judging subunit, configured to judge whether the wavelet coefficients of the m IMF vectors are greater than a coefficient threshold;
a retention subunit configured to retain the wavelet coefficients of the m IMF vectors when the wavelet coefficients are greater than a coefficient threshold;
a deletion subunit, configured to delete when the wavelet coefficients of the m IMF vectors are less than or equal to a coefficient threshold.
Optionally, the hybrid neural network model comprises a convolutional neural network and a long-short term memory cycling neural network.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the method collects plant leaf electrical signals under different illumination degrees, utilizes empirical mode decomposition and an improved wavelet threshold value method to perform noise elimination processing on the electrical signals, and converts the electrical signal array of a one-dimensional time domain into a matrix of a two-dimensional time-frequency domain through short-time Fourier transform. Inputting the matrix data into a mixed neural network, carrying out deep learning training on plant electric signals through CNN and LSTM to obtain a relation model of the plant electric signals and the illumination intensity, then determining the illumination intensity of the current plant based on the relation model, and enabling the plant to be in the optimal growth state by adjusting the illumination intensity of the surrounding environment.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
Fig. 1 is a flowchart of a method for diagnosing illumination information based on plant electrical signals according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a method and a system for diagnosing illumination information based on plant electric signals, which adjust the illumination intensity of the current environment according to the coupling relation between the plant electric signals and the illumination intensity in environmental factors so as to enable plants to be in the optimal growth state.
Similar to bioelectric signals of human beings and animals, electric signal transmission also exists in plants, and the transmission speed of the electric signal is faster than that of other general signals. The plant electric signal reflects the physiological growth state information of the plant, and the plant electric signal and the environmental factor have a specific coupling relation, so that when the growth state of the plant or the surrounding environment changes, the electric signal in the plant body changes accordingly. On the basis, according to the information transmitted by the plant electric signals, the invention analyzes and judges the current illumination intensity state of the plant through a trained model, and adjusts the environmental factors in the surrounding growth environment, so that the plant is always in the optimal growth state, thereby improving the crop yield.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
As shown in fig. 1, the method for diagnosing illumination information based on plant electrical signals according to the present invention includes the following steps:
step 101: and collecting the plant electric signals of the samples under different illumination degrees to construct a training set.
Step 102: and training the mixed neural network model through the training set to obtain a relation model between the plant electric signals and the illumination intensity.
Step 103: and determining the illumination intensity of the current plant according to the relation model based on the electrical signal data of the current plant.
Step 104: and judging whether the illumination intensity of the current plant is within the preset range of the optimal illumination intensity.
Step 105: and when the illumination intensity of the current plant is not within the preset range of the optimal illumination intensity, adjusting the illumination intensity of the current plant.
Under the condition that other environmental factors are kept unchanged, the plant electric signal real-time data under different illumination degrees are collected. The plant electric signal acquisition electrode has a positive electrode, a negative electrode and a reference electrode in common, and the positive electrode and the negative electrode are separated by 10cm to collect effective electric signals. Designing a Butterworth filter with a low-pass cut-off frequency of 100Hz and a high-pass cut-off frequency of 0.05Hz and an adaptive power frequency trap with a non-recursive system FIR structure, and simply preprocessing the plant electric signals under different illumination degrees.
Wherein, step 101 specifically includes:
step 1011: dividing the simply preprocessed plant electric signals at equal intervals according to the time length t to obtain a plant electric signal data array with the length of each segment being L;
step 1012: carrying out empirical mode decomposition on each section of plant electric signals, carrying out stabilization treatment on non-stable plant electric signals to obtain n IMF vectors, selecting the IMF vectors with high m phase relation numbers, and eliminating interference noise in the signals by an improved wavelet threshold method;
step 1013: splicing the m IMF vectors into a signal number sequence with the length of L x m, marking as I, and converting all the signal number sequences I into a two-dimensional matrix of a time domain and a frequency domain by using short-time Fourier transform (STFT); the two-dimensional matrix constitutes a training set, a test set and a verification set.
Wherein, step 1012 specifically includes:
finding out all maximum value points in the plant electric signal data sequence with the length of each segment being L, and forming an upper envelope line e of the discrete signal sequence through spline interpolation function cubic fitting+(t); finding out all minimum value points in the plant electric signal data sequence in the same way, and fitting to form a lower envelope line e of the discrete signal sequence-(t)；
Calculating the mean m (t) of the upper envelope and the lower envelope of the discrete signal sequence, and then: m (t) ═ e+(t)-e-(t)；
Subtracting the mean value m (t) of the upper envelope line and the lower envelope line from the original data sequence to obtain a new data sequence h (t);
repeating the steps until h (t) has no extreme value, and stopping;
a plurality of intrinsic mode functions EMD expression as
Figure BDA0003383613320000051
Wherein x (t) is the original signal sequence, Ci(t) is IMF of order i, rn(t) is the residual component after n decompositions.
The process of eliminating the interference noise in the signal by the improved wavelet threshold method specifically comprises the following steps: determining the wavelet base type according to the noise elimination requirement, and selecting the wavelet decomposition layer number, the wavelet step length and the support distance. A threshold lambda is selected and the wavelet coefficients of the signal are compared to the threshold lambda. Regarding the part with the absolute value less than lambda as noise, and setting the wavelet coefficient to zero; the part with absolute value greater than λ is regarded as valid information in the signal, and its new value minus λ is retained.
And performing wavelet inverse transformation on the electric signal subjected to the denoising treatment, and reconstructing the electric signal according to a wavelet scale. The recombined signal data can ensure the effectiveness in the plant electric signals after the wavelet threshold function is improved.
The step 101 further includes splicing the m IMF vectors into a signal number sequence I to ensure the integrity of the electrical signal information, converting the signal number sequence into a matrix by using short-time Fourier transform, converting the one-dimensional plant electrical signal data into a two-dimensional energy spectrum, and inputting the two-dimensional energy spectrum into the hybrid neural network.
The process of short-time fourier transformation is to divide a longer time signal into shorter segments of the same length, and to compute the fourier transform, i.e. the fourier spectrum, on each shorter segment. Selecting a window function which can be kept stable (is stable) in a short time, and continuously translating the window function to enable f (t), g (t) to be stable signals in each limited time range, so as to obtain energy spectrograms of the data at different moments.
The formula for the short-time fourier transform is as follows:
Figure BDA0003383613320000061
wherein Z (t) is the source signal of plant electricity, and g (t) is a window function.
The hybrid neural network model in step 102 includes a convolutional neural network and a long-short term memory cycle neural network. The input of the hybrid neural network model is an energy spectrogram of plant electric signal data under different illumination degrees, and the plant electric signal data set data is divided, wherein the training set is 70%, the testing set is 15%, and the verifying set is 15%.
The hierarchical structure of the Convolutional Neural Network (CNN) includes a data input layer, a convolutional calculation layer, a ReLU excitation layer, a pooling layer, and a full connection layer. The CNN processes input data by adopting local connection and weight sharing on the basis of a multilayer perceptron (MLP), and reduces the overfitting risk on the basis of reducing the complexity of a model.
The convolutional neural network introduces a residual error network structure, residual error learning is carried out by using the convolutional layer, the model structure can be deeper through the residual error structure, and the precision of model training is gradually improved.
The formula for residual learning is:
F(x)＝H(x)-x
where F (x) is the pre-summation network map and H (x) is the input-to-summation network map.
The convolutional neural network model uses a 3-time convolution process, and each module comprises convolution, batch normalization, a max pooling layer and a RELU layer.
And the convolutional neural network is used for extracting the characteristics of the energy graph, obtaining the parameters of a characteristic extraction layer through autonomous training and continuously updating the weight parameters.
Training errors and generalization errors occur in the training process of the model. The training error is the error expressed in the training set, and the generalization error is the error expressed in the testing set. The learned model parameters of the training data set may cause the model to perform better in the training set than in the test set.
The formula of the adjusting model is as follows:
Figure BDA0003383613320000071
in the above formula, ωkIs the weight parameter of the model, and b is the bias.
Training the convolution neural network electric signal data, and obtaining a one-dimensional signal through feature extraction.
And the one-dimensional signal is used as the input of the long-term and short-term memory neural network, the one-dimensional signal is subjected to learning optimization, and a relation model between the plant electric signal and the illumination intensity is obtained through classification training.
The specific embodiment is as follows:
the illumination intensity is set to five grades from 5000lx to 8000lx, and ten aloe leaf electrical signals with identical growth states are collected under each illumination degree, and are grouped into one group in ten minutes. 500 sets of samples were collected for each illumination level, for a total of 2500 sets of samples.
The aloe collecting electrode comprises a positive electrode, a negative electrode and a reference electrode, the reference electrode is grounded, the positive electrode and the negative electrode are inserted into the neck of the aloe, and the distance between the positive electrode and the negative electrode is 10 cm.
The cut-off frequency of the low-pass filter is 100Hz, the cut-off frequency of the high-pass filter is 0.05Hz, and the power frequency wave trap adopts a non-recursive system FIR structure to simply preprocess the aloe electric signal.
The method comprises the steps of decomposing an aloe electric signal into 8 stable IMF components according to an EMD principle, selecting 4 IMF components according to a correlation principle in mathematics, and performing improved wavelet threshold function denoising on the aloe electric signal.
The modified wavelet threshold function model was chosen as db4 wavelet with a wavelet scale of 3. Obtaining a threshold matrix of each IMF component through MATLAB software, completely zeroing the coefficients smaller than the improved wavelet threshold in the IMF components, and completely reserving the coefficients larger than the improved wavelet threshold.
Splicing the 4 IMF vectors into a signal number sequence with the length of L x 4, marking as I, converting all the signal number sequences I into a two-dimensional matrix of a time domain and a frequency domain by using short-time Fourier transform (STFT), and using the two-dimensional matrix as the input of the convolutional neural network.
Let the label category of five samples be 1, 2, 3, 4, 5, and convert into a matrix in the form of 5 rows and 5 columns of vectors, with the index of label position in each row being 1 and the other positions being 0. The class probability is 1 and the other class probabilities are 0. The plant electric signal data set data is divided, wherein the training set is 70%, the testing set is 15%, and the verifying set is 15%.
The hierarchical structure design of the convolutional neural network comprises a data input layer, a convolutional calculation layer, a ReLU excitation layer, a pooling layer and a full connection layer. The CNN adopts a local connection and weight sharing mode on the basis of a multilayer perceptron, introduces a residual error network structure, uses a convolution layer for residual error learning, uses a convolution process for 3 times, and comprises a convolution layer, a batch normalization layer, a maximum pooling layer and a RELU layer.
The convolutional neural network is set to a convolution kernel size of 5x5, a move step size of 2, a pad of 0, and a sigmod activation function is used. The pooling layers are all maximally pooled, the size of the pooled convolution kernel is 2 x 2, and the step length is 2. The fully-connected layer uses flatten to transform each sample into a two-dimensional vector representation, the vector length being the product of channel, height and width. The fully connected layer has three layers, and the output of each layer of network is 120, 84 and 10 respectively.
The parameters of the long-short term memory neural network model are set as follows: a forgetting gate 64, an updating gate 80 and an output gate 2, and the output result of the complete connection layer of the convolutional neural network model is input into the LSTM network model.
The testing accuracy of the CNN-LSTM hybrid network model according to the classification result is 0.9683, and the testing false alarm rate is as low as 0.0357.
According to the experimental results, the CNN-LSTM combined neural network model is adopted to deeply learn the coupling relation between the plant electric signals and the illumination intensity, and finally the illumination intensity of the surrounding environment can be reflected through the electric signal intensity, so that the illumination intensity can be adjusted to enable the plant to be in the optimal illumination environment.
By the method, the plant leaf electric signals under different illumination degrees are collected, the electric signals are subjected to noise elimination processing by using empirical mode decomposition and an improved wavelet threshold method, and the electric signal array of a one-dimensional time domain is converted into a matrix of a two-dimensional time-frequency domain by short-time Fourier transform. Inputting the matrix data into a neural network, carrying out deep learning training on the plant electric signals through CNN and LSTM to obtain a relation model of the plant electric signals and the illumination intensity, and continuously adjusting the illumination intensity of the surrounding environment through the plant electric signals. The method solves the problems of delayed acquisition of the existing plant growth state evaluation information, long period and the like, and has simple culture and measurement methods and reliable evaluation effect.
The invention also provides an illumination information diagnosis system based on the plant electric signals, which comprises:
the training set construction module is used for collecting the plant electrical signals of the samples under different illumination degrees and constructing a training set;
the training module is used for training the mixed neural network model through the training set to obtain a relation model between the plant electric signals and the illumination intensity;
the current plant illumination intensity determining module is used for determining the illumination intensity of the current plant according to the relation model based on the electric signal data of the current plant;
the judging module is used for judging whether the illumination intensity of the current plant is within the preset range of the optimal illumination intensity;
and the adjusting module is used for adjusting the illumination intensity of the current plant when the illumination intensity of the current plant is not within the preset range of the optimal illumination intensity.
The training set construction module specifically comprises:
the dividing unit is used for dividing the sample plant electric signals according to time length, and the length of each section of the sample plant electric signals is L;
the empirical mode decomposition unit is used for carrying out empirical mode decomposition on each divided section of the plant electric signal to obtain n IMF vectors;
a selection unit, configured to select the IMF vectors with the first m high correlation numbers from the n IMF vectors;
the training set constructing unit is used for splicing the m IMF vectors into a plurality of signal number sequences and converting all the signal number sequences into a two-dimensional matrix of a time domain and a frequency domain by using short-time Fourier transform; the two-dimensional matrix constitutes a training set.
Wherein the training set constructing module further comprises: and the elimination unit is used for eliminating the interference noise in the m IMF vectors by adopting an improved wavelet threshold method.
Optionally, the eliminating unit specifically includes:
a judging subunit, configured to judge whether the wavelet coefficients of the m IMF vectors are greater than a coefficient threshold;
a retention subunit configured to retain the wavelet coefficients of the m IMF vectors when the wavelet coefficients are greater than a coefficient threshold;
a deletion subunit, configured to delete when the wavelet coefficients of the m IMF vectors are less than or equal to a coefficient threshold.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims (10)

1.一种基于植物电信号的光照信息诊断方法，其特征在于，包括：1. a method for diagnosing illumination information based on plant electrical signals, is characterized in that, comprising: 采集不同光照程度下样本植物电信号，构建训练集；Collect the electrical signals of sample plants under different light levels to construct a training set; 通过所述训练集对混合神经网络模型进行训练，得到植物电信号与光照强度之间的关系模型；The hybrid neural network model is trained through the training set to obtain a relationship model between the plant electrical signal and the light intensity; 基于当前植物的电信号数据，根据所述关系模型确定当前植物的光照强度；Based on the electrical signal data of the current plant, determine the light intensity of the current plant according to the relationship model; 判断当前植物的光照强度是否在最佳光照强度预设范围内；Determine whether the light intensity of the current plant is within the preset range of optimal light intensity; 若否，则对当前植物的光照强度进行调节。If not, adjust the light intensity of the current plant. 2.根据权利要求1所述的基于植物电信号的光照信息诊断方法，其特征在于，所述采集不同光照程度下样本植物的电信号数据，构建训练集，具体包括：2. The method for diagnosing illumination information based on plant electrical signals according to claim 1, wherein the collection of electrical signal data of sample plants under different illumination levels to construct a training set specifically includes: 按时间长度对样本植物电信号进行划分，每段样本植物电信号长度为L；Divide the electrical signal of the sample plant according to the time length, and the electrical signal length of each sample plant is L; 对划分后的每段样本植物电信号进行经验模态分解，得到n个IMF向量；Perform empirical mode decomposition on each segment of the divided sample plant electrical signals to obtain n IMF vectors; 从n个IMF向量中选择前m个相关系数高的IMF向量；Select the first m IMF vectors with high correlation coefficients from the n IMF vectors; 将m个IMF向量拼接成多个信号数列，并利用短时傅里叶变换将所有所述信号数列转换成时域和频域的二维矩阵；所述二维矩阵构成训练集。The m IMF vectors are spliced into multiple signal sequences, and short-time Fourier transform is used to convert all the signal sequences into two-dimensional matrices in time domain and frequency domain; the two-dimensional matrices constitute a training set. 3.根据权利要求2所述的基于植物电信号的光照信息诊断方法，其特征在于，还包括：采用改进小波阈值法消除m个IMF向量中的干扰噪声。3 . The method for diagnosing illumination information based on plant electrical signals according to claim 2 , further comprising: using an improved wavelet threshold method to eliminate interference noise in the m IMF vectors. 4 . 4.根据权利要求3所述的基于植物电信号的光照信息诊断方法，其特征在于，所述采用改进小波阈值法消除m个IMF向量中的干扰噪声，具体包括：4. the illumination information diagnosis method based on plant electrical signal according to claim 3, is characterized in that, described adopting improved wavelet threshold method to eliminate the interference noise in m IMF vectors, specifically comprises: 判断m个IMF向量的小波系数是否大于系数阈值；Determine whether the wavelet coefficients of m IMF vectors are greater than the coefficient threshold; 若是，则保留；If so, keep it; 若否，则删除。If not, delete it. 5.根据权利要求1所述的基于植物电信号的光照信息诊断方法，其特征在于，所述混合神经网络模型包括卷积神经网络和长短期记忆循环神经网络。5 . The method for diagnosing illumination information based on plant electrical signals according to claim 1 , wherein the hybrid neural network model comprises a convolutional neural network and a long short-term memory recurrent neural network. 6 . 6.一种基于植物电信号的光照信息诊断系统，其特征在于，包括：6. An illumination information diagnosis system based on plant electrical signal, is characterized in that, comprising: 训练集构建模块，用于采集不同光照程度下样本植物电信号，构建训练集；The training set building module is used to collect electrical signals of sample plants under different light levels to construct a training set; 训练模块，用于通过所述训练集对混合神经网络模型进行训练，得到植物电信号与光照强度之间的关系模型；A training module for training the hybrid neural network model through the training set to obtain a relationship model between the plant electrical signal and the light intensity; 当前植物光照强度确定模块，用于基于当前植物的电信号数据，根据所述关系模型确定当前植物的光照强度；The current plant light intensity determination module is used to determine the light intensity of the current plant according to the relationship model based on the electrical signal data of the current plant; 判断模块，用于判断当前植物的光照强度是否在最佳光照强度预设范围内；The judgment module is used to judge whether the light intensity of the current plant is within the preset range of the optimal light intensity; 调节模块，用于当当前植物的光照强度时候不在最佳光照强度预设范围内时，对当前植物的光照强度进行调节。The adjustment module is used to adjust the light intensity of the current plant when the light intensity of the current plant is not within the preset range of the optimal light intensity. 7.根据权利要求6所述的基于植物电信号的光照信息诊断系统，其特征在于，所述训练集构建模块具体包括：7. The illumination information diagnosis system based on plant electrical signals according to claim 6, wherein the training set building module specifically comprises: 划分单元，用于按时间长度对样本植物电信号进行划分，每段样本植物电信号长度为L；a dividing unit, used to divide the electrical signal of the sample plant according to the time length, and the length of the electrical signal of each sample plant is L; 经验模态分解单元，用于对划分后的每段样本植物电信号进行经验模态分解，得到n个IMF向量；The empirical mode decomposition unit is used to perform empirical mode decomposition on each segment of the divided sample plant electrical signals to obtain n IMF vectors; 选择单元，用于从n个IMF向量中选择前m个相关系数高的IMF向量；The selection unit is used to select the first m IMF vectors with high correlation coefficients from the n IMF vectors; 训练集构建单元，用于将m个IMF向量拼接成多个信号数列，并利用短时傅里叶变换将所有所述信号数列转换成时域和频域的二维矩阵；所述二维矩阵构成训练集。A training set construction unit, used for splicing m IMF vectors into multiple signal sequence, and using short-time Fourier transform to convert all the signal sequence into two-dimensional matrix in time domain and frequency domain; the two-dimensional matrix form the training set. 8.根据权利要求1所述的基于植物电信号的光照信息诊断方法，其特征在于，所述训练集构建模块还包括：消除单元，用于采用改进小波阈值法消除m个IMF向量中的干扰噪声。8. The method for diagnosing illumination information based on plant electrical signals according to claim 1, wherein the training set building module further comprises: a eliminating unit for eliminating interference in m IMF vectors by using an improved wavelet threshold method noise. 9.根据权利要求8所述的基于植物电信号的光照信息诊断方法，其特征在于，所述消除单元具体包括：9. The method for diagnosing illumination information based on plant electrical signals according to claim 8, wherein the elimination unit specifically comprises: 判断子单元，用于判断m个IMF向量的小波系数是否大于系数阈值；a judgment subunit, used to judge whether the wavelet coefficients of the m IMF vectors are greater than the coefficient threshold; 保留子单元，用于当m个IMF向量的小波系数大于系数阈值时，保留；The reserved subunit is used to retain when the wavelet coefficients of the m IMF vectors are greater than the coefficient threshold; 删除子单元，用于当m个IMF向量的小波系数小于或等于系数阈值时，删除。The deletion subunit is used to delete when the wavelet coefficients of the m IMF vectors are less than or equal to the coefficient threshold. 10.根据权利要求6所述的基于植物电信号的光照信息诊断方法，其特征在于，所述混合神经网络模型包括卷积神经网络和长短期记忆循环神经网络。10 . The method for diagnosing illumination information based on plant electrical signals according to claim 6 , wherein the hybrid neural network model comprises a convolutional neural network and a long short-term memory recurrent neural network. 11 .
CN202111441798.XA 2021-11-30 2021-11-30 A method and system for diagnosing light information based on plant electrical signals Pending CN114118157A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111441798.XA CN114118157A (en) 2021-11-30 2021-11-30 A method and system for diagnosing light information based on plant electrical signals

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111441798.XA CN114118157A (en) 2021-11-30 2021-11-30 A method and system for diagnosing light information based on plant electrical signals

Publications (1)

Publication Number Publication Date
CN114118157A true CN114118157A (en) 2022-03-01

Family

ID=80368063

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111441798.XA Pending CN114118157A (en) 2021-11-30 2021-11-30 A method and system for diagnosing light information based on plant electrical signals

Country Status (1)

Country Link
CN (1) CN114118157A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI830545B (en) * 2022-12-19 2024-01-21 財團法人工業技術研究院 Method and systemfor intelligent control light supplement for plants

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN204101942U (en) * 2014-09-28 2015-01-14 天津职业技术师范大学 Plant growth environment multiple information detects and display device
CN105511526A (en) * 2014-09-25 2016-04-20 天津职业技术师范大学 Plant electric signal detection and contrast experimental apparatus
CN108197648A (en) * 2017-12-28 2018-06-22 华中科技大学 A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models
US20200302338A1 (en) * 2019-03-18 2020-09-24 Vivent Sárl Apparatus and Method for Assessing a Characteristic of a Plant
CN111741558A (en) * 2020-07-23 2020-10-02 浙江工业大学之江学院 LED plant lighting driving device based on neural network
CN113392771A (en) * 2021-06-16 2021-09-14 天津职业技术师范大学(中国职业培训指导教师进修中心) Plant growth state diagnosis method, system and device based on deep learning

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105511526A (en) * 2014-09-25 2016-04-20 天津职业技术师范大学 Plant electric signal detection and contrast experimental apparatus
CN204101942U (en) * 2014-09-28 2015-01-14 天津职业技术师范大学 Plant growth environment multiple information detects and display device
CN108197648A (en) * 2017-12-28 2018-06-22 华中科技大学 A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models
US20200302338A1 (en) * 2019-03-18 2020-09-24 Vivent Sárl Apparatus and Method for Assessing a Characteristic of a Plant
CN111741558A (en) * 2020-07-23 2020-10-02 浙江工业大学之江学院 LED plant lighting driving device based on neural network
CN113392771A (en) * 2021-06-16 2021-09-14 天津职业技术师范大学(中国职业培训指导教师进修中心) Plant growth state diagnosis method, system and device based on deep learning

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
刘闯: "基于深度学习的植物生长状态诊断方法研究", 中国优秀硕士学位论文全文数据库信息辑, no. 6, 15 June 2021 (2021-06-15), pages 2 - 5 *
王忠义, 陈端生, 黄岚: "利用人工神经网络建立植物电信号与环境因子关系", 农业工程学报, no. 03, 30 May 2001 (2001-05-30), pages 1 - 4 *
陆静霞;於海明;陈士进;凌威龙;丁为民;: "基于植物电信号的环境因子预测模型", 农业机械学报, no. 11, 25 November 2013 (2013-11-25) *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI830545B (en) * 2022-12-19 2024-01-21 財團法人工業技術研究院 Method and systemfor intelligent control light supplement for plants

Similar Documents

Publication Publication Date Title
CN102512158B (en) Electrocardiosignal quality evaluation method and device based on high-dimensional fuzzy recognition
CN112418013A (en) Complex working condition bearing fault diagnosis method based on meta-learning under small sample
CN109948647A (en) A method and system for ECG classification based on deep residual network
CN112426160A (en) Electrocardiosignal type identification method and device
CN116626753B (en) Microseism event identification method and system based on multi-modal neural network
CN118557162B (en) Sphygmomanometer calibration method and system
CN116755547A (en) Surface EMG signal gesture recognition system based on lightweight convolutional neural network
CN110443276A (en) Time series classification method based on depth convolutional network Yu the map analysis of gray scale recurrence
CN114469124A (en) Method for identifying abnormal electrocardiosignals in motion process
CN109682892B (en) A Signal Denoising Method Based on Time-Frequency Analysis
CN118626945A (en) A new interpretable bearing fault diagnosis method based on two-stage feature extraction
CN114611560A (en) SSVEP electroencephalogram signal classification method based on convolutional neural network
Liu et al. Classification of cetacean whistles based on convolutional neural network
CN114118157A (en) A method and system for diagnosing light information based on plant electrical signals
CN119689367A (en) Transformer fault diagnosis method based on multi-sensor data fusion and transfer learning
CN116628420A (en) A brain wave signal processing method based on LSTM neural network element learning
CN114065822B (en) Electromagnetic identification method and system for ocean tide fluctuation
CN113990303B (en) Environmental sound identification method based on multi-resolution cavity depth separable convolution network
CN117617888B (en) System and method for predicting myopic diopter
CN118173107B (en) Bird sound quality analysis method based on multi-mode depth feature level fusion
CN117727332B (en) Ecological population assessment method based on language spectrum feature analysis
CN119112183A (en) EEG emotion analysis method based on deep neural network
CN113392771A (en) Plant growth state diagnosis method, system and device based on deep learning
CN116584902B (en) Heart sound classification device based on feature optimization and visualization
CN111554319A (en) A multi-channel abnormal heart and lung sound recognition system and device based on low-rank tensor learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination