CN114936699A - Method for predicting death rate of seedlings in flower cultivation process based on deep learning - Google Patents

Method for predicting death rate of seedlings in flower cultivation process based on deep learning Download PDF

Info

Publication number
CN114936699A
CN114936699A CN202210597524.8A CN202210597524A CN114936699A CN 114936699 A CN114936699 A CN 114936699A CN 202210597524 A CN202210597524 A CN 202210597524A CN 114936699 A CN114936699 A CN 114936699A
Authority
CN
China
Prior art keywords
mortality
prediction model
seedlings
data
mortality prediction
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
CN202210597524.8A
Other languages
Chinese (zh)
Inventor
刘静
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.)
Guangdong Vocational and Technical College
Original Assignee
Guangdong Vocational and Technical College
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 Guangdong Vocational and Technical College filed Critical Guangdong Vocational and Technical College
Priority to CN202210597524.8A priority Critical patent/CN114936699A/en
Publication of CN114936699A publication Critical patent/CN114936699A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/21—Design, administration or maintenance of databases
    • G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • 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
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic 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
    • G06N3/048—Activation functions
    • 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
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Business, Economics & Management (AREA)
  • Probability & Statistics with Applications (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Databases & Information Systems (AREA)
  • Human Resources & Organizations (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)

Abstract

The invention discloses a seedling death rate prediction method in a flower cultivation process based on deep learning, which comprises the following steps: acquiring characteristic data of seedlings in the whole growth cycle and corresponding mortality of the seedlings to generate a death characteristic data set of the seedlings; establishing a first mortality prediction model and a second mortality prediction model according to the death characteristic data set; acquiring environmental data and growth data of seedlings in the current growth cycle to generate a current characteristic data set of the seedlings; according to the current characteristic data set, the first mortality prediction model predicts and outputs a first predicted mortality, and the second mortality prediction model predicts and outputs a second predicted mortality; and respectively endowing the first weight and the second weight with the first predicted mortality and the second predicted mortality, and accumulating the first weight and the second weight to obtain the comprehensive mortality of the seedlings in the next growth cycle. The method can predict the death rate of the seedlings in advance, and cultivators can take rescue measures to the seedlings according to the comprehensive death rate, so that the survival rate of the flower seedlings is improved.

Description

Method for predicting death rate of seedlings in flower cultivation process based on deep learning
Technical Field
The invention belongs to the technical field of artificial intelligence, and particularly relates to a seedling mortality prediction method in a flower cultivation process based on deep learning.
Background
Flowers are herbaceous plants with ornamental value, usually enjoying the sun and being cold-resistant. Flowers have both broad and narrow meanings, and the flowers in the narrow sense refer to herbaceous plants with ornamental value (such as impatiens balsamina, chrysanthemum and cockscomb), and the flowers in the broad sense comprise herbaceous or woody ground cover plants, flower shrubs, flowering trees and bonsais (such as lilyturf roots and sedum roots) besides the herbaceous plants with ornamental value. China is the country with the largest flower cultivation area in the world, has a wide consumption market, and people in real life often plant, buy and wear flowers in various ways due to beautiful appearance and pleasant fragrance of flowers.
In the whole growth period of flowers, a grower needs to ensure that the growth environmental parameters of flower seedlings are in a proper range, and needs to regularly deinsectize and fertilize the flower seedlings to ensure the healthy and strong growth of the flowers. However, if the growth environment parameters of the seedlings are out of the proper ranges during the growth cycle of the seedlings, or the seedlings are subjected to serious insect damage, or the soil contains too much fertilizer, these causes increase the death rate of the flower seedlings in the future growth cycle, resulting in the flower seedlings failing to thrive and die. How to predict the death rate of the next growth cycle of the seedling based on the characteristic parameters of the body of the current flower seedling and the environmental parameters of the environment where the seedling is located, and further, measures for preventing flower death are made in advance, which becomes a big problem in the technical field of flower cultivation.
Disclosure of Invention
The invention aims to provide a seedling mortality prediction method in a flower cultivation process based on deep learning, so as to solve one or more technical problems in the prior art and provide at least one beneficial selection or creation condition.
The solution of the invention for solving the technical problem is as follows: the method for predicting the death rate of the seedlings in the flower cultivation process based on deep learning is provided, and comprises the following steps:
s100, acquiring characteristic data of seedlings in the whole growth period and corresponding mortality data thereof based on flower big data, preprocessing the characteristic data of the seedlings and the corresponding mortality data thereof, and generating a death characteristic data set of the seedlings;
wherein the characteristic data comprises ambient temperature, ambient pH, soil polyion concentration, soil enzyme biological concentration, seedling wormhole number and seedling chlorophyll concentration;
s200, establishing a first mortality prediction model and a second mortality prediction model according to the death characteristic data set;
s300, acquiring environmental data and growth data of the seedlings in the current growth cycle, preprocessing the environmental data and the growth data of the seedlings in the current growth cycle, and generating a current characteristic data set of the seedlings;
s400, according to the current characteristic data set, the first mortality prediction model predicts the mortality of the seedlings in the next growth cycle and outputs a first predicted mortality, the second mortality prediction model predicts the mortality of the seedlings in the next growth cycle and outputs a second predicted mortality;
s500, giving the first predicted death rate to a first weight, giving the second predicted death rate to a second weight, and accumulating the first predicted death rate given to the first weight and the second predicted death rate given to the second weight to obtain the comprehensive death rate of the seedling in the next growth cycle.
As a further improvement of the above technical solution, in step S100, the step of preprocessing the characteristic data of the seedling and the mortality data corresponding thereto is:
s110, finding out characteristic data of seedlings and a missing value of corresponding mortality data of the seedlings, and completing the missing value through a regression algorithm;
s120, searching for the characteristic data of the seedlings and the repetition values and abnormal values of the mortality data corresponding to the characteristic data, and discarding the repetition values and the abnormal values;
s130, reducing dimensions of the characteristic data of the seedlings and the mortality data corresponding to the characteristic data;
s140, standardizing the characteristic data of the seedlings and the corresponding mortality data thereof to generate a death characteristic data set of the seedlings.
As a further improvement of the above technical solution, in step S200, the establishing of the first mortality prediction model and the second mortality prediction model according to the death characteristic data set includes a first mortality prediction model establishing step and a second mortality prediction model establishing step;
the first mortality prediction model establishing step comprises the steps of:
s210, converting the format of the death characteristic data set into an input format conforming to a convolutional neural network, and dividing the death characteristic data set after the format conversion into a first training set and a first testing set according to the ratio of 8: 2;
s211, inputting the first training set into a convolutional neural network, training the convolutional neural network, and obtaining a first feature vector;
s212, converting the format of the first feature vector into an input format conforming to the bidirectional long and short term memory neural network, inputting the first feature vector after format conversion into the bidirectional long and short term memory neural network, training the bidirectional long and short term memory neural network, and obtaining a second feature vector;
s213, constructing a result classification layer and constructing an attention layer based on an attention mechanism, wherein the second feature vector is input into the attention layer, the output result of the attention layer is input into the result classification layer, and the result classification layer outputs to generate a first mortality prediction model;
s214, setting a first evaluation index, and evaluating the performance of the first mortality prediction model according to the first test set and the first evaluation index;
the second mortality prediction model establishing step comprises:
s220, performing feature selection on the death feature data set;
s221, dividing the death data set after feature selection into a second training set and a second testing set according to the ratio of 8:2, inputting the second training set into the XGboost model, training the XGboost model, and generating a second mortality prediction model;
s222, evaluating the performance of the second mortality prediction model according to the second test set and the second evaluation index.
As a further improvement of the above technical solution, the attention mechanism satisfies the following formula:
Figure BDA0003668705040000041
wherein,
Figure BDA0003668705040000042
representing the output result of the attention layer, K is a second feature vector, ω is a preset training parameter vector, and softmax represents the operation in the softmax classifier.
As a further improvement of the above technical solution, the step of performing feature selection on the death feature data set includes: inquiring a null value in the death characteristic data set, and calling a median of a column where the null value is located to fill; and calling a scoring function of the sklern library to perform univariate feature selection on the death feature data set filled with the null value, wherein the scoring function is an f _ regression () function.
As a further improvement of the above technical solution, the first evaluation index satisfies the following formula:
Figure BDA0003668705040000051
Figure BDA0003668705040000052
α 1 ＝30％×α 12 +70％×α 11 ；
wherein alpha is 11 Average absolute error value, alpha, for the first mortality prediction model 12 Root mean square error value, α, for the first mortality prediction model 1 Error value, n, for the first mortality prediction model 1 Is the amount of data in the first test set,
Figure BDA0003668705040000056
to input firstA mortality prediction value, m, output by the first mortality prediction model after the test set i The death rate true value corresponding to the first test set;
the second evaluation index satisfies the following formula:
Figure BDA0003668705040000053
Figure BDA0003668705040000054
α 2 ＝30％×α 22 +70％×α 21 ；
wherein alpha is 21 Average absolute error value, alpha, for the second mortality prediction model 22 Root mean square error value, α, for the second mortality prediction model 2 Error value, n, for the second mortality prediction model 2 Is the amount of data in the second test set,
Figure BDA0003668705040000055
predicted mortality value, k, output for the second mortality prediction model after input of the second test set i The mortality true value corresponding to the second test set.
As a further improvement of the above technical solution, in step S500, the first weight satisfies the following formula:
Figure BDA0003668705040000061
wherein, delta 1 Representing a first weight, α, assigned to a first mortality prediction model 11 Is the mean absolute error value, α, of the first mortality prediction model 12 Root mean square error value, α, for the first mortality prediction model 21 Average absolute error value, alpha, for the second mortality prediction model 22 A root mean square error value for the second mortality prediction model;
the second weight satisfies the following formula:
Figure BDA0003668705040000062
wherein, delta 2 Representing a second weight, α, assigned to a second mortality prediction model 11 Is the mean absolute error value, α, of the first mortality prediction model 12 Root mean square error value, α, for the first mortality prediction model 21 Mean absolute error value, α, for the second mortality prediction model 22 A root mean square error value for the second mortality prediction model;
the comprehensive mortality rate of the seedlings in the next growth cycle satisfies the following formula:
D＝δ 1 y 1 +δ 2 y 2 ；
wherein D represents the overall mortality, δ 1 Representing a first weight, δ, assigned to a first mortality prediction model 2 Representing a second weight, y, assigned to a second mortality prediction model 1 For the first prediction of mortality, y 2 Mortality was predicted for the second.
As a further improvement of the above technical solution, in step S300, the environmental data is the environmental data obtained and output by a temperature sensor, a pH sensor, a sodium ion sensor, a residual chlorine sensor, an enzyme biosensor and a nitrate sensor, the temperature sensor is used for collecting the temperature data of the environment where the flowers are located, the pH sensor is used for collecting the pH data of the environment where the flowers are located, the sodium ion sensor is used for collecting the sodium ion concentration of the soil where the flowers are located, the residual chlorine sensor is used for collecting the chlorine ion concentration of the soil where the flowers are located, the enzyme biosensor is used for collecting the enzyme concentration of the soil where the flowers are located, the nitrate sensor is used for collecting the nitrate ion concentration of the soil where the flowers are located, the environmental data include physical quantities of temperature, pH, sodium ion concentration, chloride ion concentration, nitrate example concentration, enzyme concentration;
the growth data is output by a first image sensor, a second image sensor and a processing module, the first image sensor is a fluorescence image sensor, and the first image sensor is used for detecting the chlorophyll concentration in the seedling; the second image sensor is a high image sensor and is used for detecting the spectral characteristics of the seedlings and outputting the spectral characteristics to the processing module, and the processing module is used for obtaining the quantity of wormholes of the seedlings according to the spectral characteristics; the characteristic key data comprise chlorophyll concentration and seedling wormhole number.
A computer-readable storage medium having stored thereon a computer program which, when being executed by a processor, implements the steps of the method for predicting seedling mortality in a deep learning-based floral cultivation process.
A system for predicting seedling mortality in a flower cultivation process based on deep learning, the system comprising:
at least one processor;
at least one memory for storing at least one program;
when the at least one program is executed by the at least one processor, the at least one processor may implement the method for predicting seedling mortality in a deep learning-based flower cultivation process.
The invention has the beneficial effects that: the invention provides a seedling mortality prediction method in a flower cultivation process based on deep learning, which is characterized in that a first mortality prediction model is constructed based on CNN-BilSTM, a second mortality prediction model is constructed based on XGboost algorithm, a first predicted mortality is obtained through the first mortality prediction model according to the collected current characteristic data set of seedlings in the current growth cycle, and a second predicted mortality is obtained through the second mortality prediction model; by combining the first predicted mortality and the second predicted mortality based on the residual reciprocal method, a more accurate total mortality of the seedlings in the next growth cycle is obtained. The method can predict the death rate of the flower seedlings in advance, and cultivators can take rescue measures to the flower seedlings according to the comprehensive death rate, so that the survival rate of the flower seedlings is improved.
Drawings
In order to more clearly illustrate the technical solution in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly described below. It is clear that the described figures are only some embodiments of the invention, not all embodiments, and that a person skilled in the art can also derive other designs and figures from them without inventive effort.
FIG. 1 is a flow chart of a method for predicting the death rate of seedlings in a flower cultivation process based on deep learning;
FIG. 2 is a flow chart of the working of pre-processing seedling characteristic data and mortality data corresponding thereto for a seedling mortality prediction method in a flower cultivation process based on deep learning;
FIG. 3 is a structural diagram of a seedling mortality prediction method in a flower cultivation process based on deep learning;
FIG. 4 is a flowchart of a first mortality prediction model building step of a seedling mortality prediction method in a flower cultivation process based on deep learning;
FIG. 5 is a flowchart of a second mortality prediction model building step of the seedling mortality prediction method in the flower cultivation process based on deep learning.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
It should be noted that although functional block divisions are provided in the system drawings and logical orders are shown in the flowcharts, in some cases, the steps shown and described may be performed in different orders than the block divisions in the systems or in the flowcharts. The terms first, second and the like in the description and in the claims, and the drawings described above, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
Referring to fig. 1 to 5, the present disclosure provides a seedling mortality prediction method in a floriculture process based on deep learning, the method comprising the steps of:
s100, acquiring characteristic data of flower seedlings and corresponding mortality data thereof in the whole growth cycle based on flower big data, preprocessing the characteristic data of the seedlings and the corresponding mortality data thereof, and generating a death characteristic data set of the seedlings;
wherein the characteristic data comprises ambient temperature, ambient pH, soil polyion concentration, soil enzyme biological concentration, seedling wormhole number, seedling chlorophyll concentration and corresponding mortality;
s200, establishing a first mortality prediction model and a second mortality prediction model according to the death characteristic data set;
s300, acquiring environmental data and growth data of the seedlings in the current growth cycle, preprocessing the environmental data and the growth data of the seedlings in the current growth cycle, and generating a current characteristic data set of the seedlings;
s400, according to the current characteristic data set, the first mortality prediction model predicts the mortality of the seedlings in the next growth cycle and outputs a first predicted mortality, and the second mortality prediction model predicts the mortality of the seedlings in the next growth cycle and outputs a second predicted mortality;
s500, giving the first predicted death rate to a first weight, giving the second predicted death rate to a second weight, and accumulating the first predicted death rate given to the first weight and the second predicted death rate given to the second weight to obtain the comprehensive death rate of the seedling in the next growth cycle.
In this embodiment, step S100 is to obtain feature data of a seedling of a flower and corresponding mortality data thereof in the whole growth cycle based on flower big data, and before the feature data of the seedling and the corresponding mortality data thereof are input as a model, the feature data of the seedling and the corresponding mortality data thereof need to be preprocessed, where the preprocessing operation includes data cleaning, data dimension reduction, and standardization; the data cleaning is to carry out operations of discarding, filling, replacing, removing duplicate and the like on the characteristic data of the seedlings and the corresponding mortality data thereof so as to achieve the purposes of removing abnormal values, correcting error values and complementing missing values; the data dimensionality reduction is to reduce the dimensionality of the characteristic data of the seedlings and the corresponding mortality data thereof so as to improve the efficiency of subsequent prediction work; but standardized to a specification unifying the characteristic data of said seedlings and their corresponding mortality data.
Specifically, referring to fig. 2, the step of preprocessing the characteristic data of the seedling and the mortality data thereof is as follows:
s110, finding characteristic data of seedlings and missing values of corresponding mortality data of the seedlings, and completing the missing values through a regression algorithm;
specifically, the missing value is completed through a regression algorithm, the characteristic data of the seedling and the missing field of the corresponding mortality data of the seedling are used as target variables to predict, and the highest possible completing value is obtained. As the characteristic data of the seedlings and the corresponding mortality data are numerical variables, the method adopts a regression algorithm to fill up the missing values.
Preferably, the regression algorithm is a random forest.
S120, searching for the characteristic data of the seedlings and the repetition values and abnormal values of the mortality data corresponding to the characteristic data, and discarding the repetition values and the abnormal values;
specifically, the abnormal values of the characteristic data of the seedling and the mortality data corresponding thereto are defined as characteristic data outside a set range, the abnormal values also being referred to as noise data; and a repetition value is defined as a plurality of characteristic data whose data values are identical. The method is used for processing the characteristic data of the seedlings and the abnormal values and the repeated values of the mortality data corresponding to the characteristic data by adopting a removing method.
S130, reducing the dimension of the characteristic data of the seedlings and the corresponding mortality data of the seedlings;
specifically, the dimensionality reduction operation is to reduce the number of features in feature data of seedlings and corresponding mortality data, and the purpose of dimensionality reduction is to improve the efficiency and effect of subsequent prediction work. The method is used for reducing the dimension of characteristic data of seedlings and corresponding mortality data of the seedlings through Principal Component Analysis (PCA).
S140, standardizing the characteristic data of the seedlings and the corresponding mortality data thereof to generate a death characteristic data set of the seedlings.
Specifically, the normalization process is to convert all the feature data of the seedlings and the corresponding mortality data into the same specification, the feature data after the normalization process are all in the interval of [0,1], and the normalization process satisfies the following formula:
Figure BDA0003668705040000111
wherein x' represents the characteristic data of the standardized seedlings and the corresponding mortality data thereof, x is the characteristic data of the seedlings needing the standardized treatment and the corresponding mortality data thereof, and x min Is the minimum value, x, of the characteristic data of the seedling and its corresponding mortality data max Is the maximum value of the characteristic data of the seedling and the corresponding mortality data.
Referring to fig. 3, the present application obtains the comprehensive mortality of the seedling in the next growth cycle by a combined model method, which combines the results of a plurality of prediction models, thereby making full use of the information of each prediction model and effectively improving the accuracy of the final prediction result. According to the method, a first mortality prediction model and a second mortality prediction model are established in step S200, the first mortality prediction model and the second mortality prediction model jointly form a combined model, the two prediction models respectively predict the mortality of seedlings in the next growth cycle according to the characteristic data of the seedlings in the current growth cycle, the prediction results of the prediction models are given different weight values, and finally the final comprehensive mortality is obtained. The method can effectively improve the precision of the final comprehensive death rate, and according to the comprehensive death rate, a grower can take rescue measures to the seedlings with high death rate in advance so as to improve the survival rate of the seedlings.
Referring to fig. 4, step S200 includes a first mortality prediction model establishing step, where the first mortality prediction model establishing step is:
s210, converting the format of the death characteristic data set into an input format conforming to a convolutional neural network, and dividing the death characteristic data set after the format conversion into a first training set and a first testing set according to the ratio of 8: 2;
s211, inputting the first training set into a convolutional neural network, training the convolutional neural network, and obtaining a first feature vector;
s212, converting the format of the first feature vector into an input format conforming to the bidirectional long and short term memory neural network, inputting the first feature vector after format conversion into the bidirectional long and short term memory neural network, training the bidirectional long and short term memory neural network, and obtaining a second feature vector;
s212, constructing a result classification layer and constructing an attention layer based on an attention mechanism, inputting a second feature vector into the attention layer, inputting an output result of the attention layer into the result classification layer, and outputting by the result classification layer to generate a first mortality prediction model;
and S213, setting a first evaluation index, and evaluating the performance of the first mortality prediction model according to the first test set and the first evaluation index.
According to the method, the format of the death characteristic data set is converted into the input format which accords with the convolutional neural network through centralization processing, and the centralization processing aims to centralize the dimensionality of all data of the death characteristic data set into 0 so as to avoid the phenomenon that the performance of the first mortality prediction model is reduced.
A first mortality prediction model is constructed based on a mixed neural network (CNN-BilSTM), a death characteristic data set after format conversion is divided into a first training set and a first test set according to the ratio of 8:2, the first training set is used for training the neural network, and the first test set is used for testing the performance of the trained neural network. In this embodiment, the process of constructing the first mortality prediction model specifically includes:
firstly, a death feature data set is input into a Convolutional Neural Network (CNN), the Convolutional Neural network is trained to extract local features of the death feature data set, and a first feature vector is output. Convolutional neural networks are generally divided into an input layer, a convolutional layer, a normalization layer, an activation layer, a pooling layer, and a fully-connected layer, where the convolutional layer is a key layer of the convolutional neural network. In the convolutional layer, the convolutional neural network performs convolution, pooling and feature extraction by using a one-dimensional convolutional kernel filter based on a local connection and weight sharing mode, and the convolutional layer meets the following formula:
Figure BDA0003668705040000141
wherein f is an activation function, H i Outputting, for the feature of the i-th layer of the convolutional neural network, an operator
Figure BDA0003668705040000142
Representing a convolution operation, a i Is a bias term, q i Weight values of convolution kernels used for the ith layer of the convolutional neural network.
Before the first feature vector is input into the bidirectional long-short term memory neural network, the format of the first feature vector needs to be converted into the input format of the bidirectional long-short term memory neural network, so as to improve the training efficiency of the bidirectional long-short term memory neural network. And inputting the first feature vector after format conversion into a bidirectional long-short term memory neural network, training the bidirectional long-short term memory neural network, and obtaining a second feature parameter.
The bidirectional Long-Short Term Memory neural network (Bi-directional Long Short-Term Memory, BilSTM) is composed of a forward Long-Short Term Memory neural network hidden layer and a backward Long-Short Term Memory neural network hidden layer, and has a plurality of sharing weights. The forward long and short term memory neural network hiding layer is used for extracting forward features of the first feature vectors, the backward long and short term memory neural network hiding layer is used for extracting reverse features of the first feature vectors, the first feature vectors are input into the forward long and short term memory neural network hiding layer and the backward long and short term memory neural network hiding layer respectively, the forward features and the reverse features of the first feature vectors are extracted from the two hiding layers respectively, and the forward features and the reverse features are accessed into the output layer of the bidirectional long and short term memory neural network together to output second feature vectors.
Compared with a long-short term memory neural network, the long-short term memory neural network cannot extract the features of data in two directions, and the long-short term memory neural network cannot capture more comprehensive features. When the first feature vector is subjected to feature extraction, the whole information hidden in the first feature vector needs to be considered, the bidirectional long-short term memory neural network adopted by the application extracts features in the forward direction and the reverse direction, combines the forward features and the reverse features according to a specific mode, can extract and obtain more data features in the bidirectional mode, and effectively improves the prediction accuracy of the first mortality prediction model.
In this embodiment, the bidirectional long and short term memory neural network includes a forgetting gate, an updating gate, and an output gate, the three gates work together to complete the selection and forgetting of the feature information and the updating of the current cell state, and the feature information is stored in the cell state. Specifically, the updating process of the hidden layer of the forward long-short term memory neural network is as follows:
i t ＝ρ(ω i ·[h t-1 ，x t ]+b i )；
f t ＝ρ(ω f ·[h t-1 ，x t ]+b f )；
Figure BDA0003668705040000151
Figure BDA0003668705040000152
O t ＝ρ(ω o ·[h t-1 ，x t ]+b o )；
Figure BDA0003668705040000153
wherein x is t As input at the current time, i t 、f t 、O t Respectively representing the outputs of an update gate, a forget gate and an output gate of the forward long-short term memory neural network,
Figure BDA0003668705040000154
in order to be a feature of the forward direction,
Figure BDA0003668705040000155
a candidate value that is updated for the current cell state,
Figure BDA0003668705040000156
as the current cell state, b i 、b f 、b c 、b o Are all bias term vectors, ω i 、ω f 、ω c 、ω o Are weight matrixes, and rho is a Sigmod activation function.
Similarly, the backward long-short term memory neural network hidden layer extracts the features to obtain the backward features
Figure BDA0003668705040000157
The forward direction characteristic
Figure BDA0003668705040000158
And the reverse feature
Figure BDA0003668705040000159
And fusing to finally obtain a second feature vector, wherein the second feature vector meets the following formula:
Figure BDA00036687050400001510
wherein h is t Is the second feature vector.
After obtaining the second feature vector output by the bidirectional long-short term memory neural network, we construct a result classification layer and an attention layer based on an attention mechanism, wherein the attention layer is used for filtering key information from the second feature vector and giving higher weight to the key information so as to improve the prediction accuracy of the first mortality prediction model, and the attention mechanism satisfies the following formula:
Figure BDA0003668705040000161
wherein,
Figure BDA0003668705040000162
representing the output result of the attention layer, K is a second feature vector, ω is a preset training parameter vector, and softmax (·) represents the operation in the softmax classifier.
And inputting the second feature vector to an attention layer, inputting an output result of the attention layer to a result classification layer for classification, and finally generating a first mortality prediction model.
In this embodiment, step S213 is to evaluate the performance of the first mortality prediction model based on the first test set and the first evaluation index. According to the method and the device, a first evaluation index is set based on an average absolute error formula and a root mean square error formula, different weights are given to the average absolute error formula and the root mean square error formula, and finally a performance evaluation result is obtained. The first evaluation index specifically is:
Figure BDA0003668705040000163
Figure BDA0003668705040000164
α 1 ＝30％×α 12 +70％×α 11 ；
wherein alpha is 11 Average absolute error value, alpha, for the first mortality prediction model 12 Root mean square error value, α, for the first mortality prediction model 1 Error value, n, for the first mortality prediction model 1 Is the amount of data in the first test set,
Figure BDA0003668705040000171
for the mortality prediction output by the first mortality prediction model after the input of the first test set, m i The mortality true value corresponding to the first test set.
In this embodiment, a first error threshold is set, and if the error value of the first mortality prediction model is not less than or equal to the first error threshold, the hyper-parameters of the convolutional neural network and the bidirectional long-short term memory neural network are reset, and the two networks are retrained to obtain a first mortality prediction model with a smaller error value.
Referring to fig. 5, step S200 further includes a second mortality prediction model establishing step, where the second mortality prediction model establishing step is:
s220, performing feature selection on the death feature data set;
s221, dividing the death data set after feature selection into a second training set and a second testing set according to the ratio of 8:2, inputting the second training set into the XGboost model, training the XGboost model, and generating a second mortality prediction model;
s222, evaluating the performance of the second mortality prediction model according to the second test set and the second evaluation index.
According to the method, a second mortality prediction model is established based on the XGboost algorithm, the XGboost algorithm generates a strong learner through a loss function of a regular term, and the loss function of the regular term reduces the variance of the XGboost model, so that the finally trained model is simpler, and overfitting of the model is effectively prevented.
Before a second mortality prediction model is established, feature selection needs to be performed on a death feature data set to improve the efficiency of the XGBoost algorithm and the prediction accuracy of the second mortality prediction model, and the feature selection performed on the death feature data set specifically includes: inquiring a null value in the death characteristic data set, and calling a median of a column where the null value is located to fill; and calling a scoring function of the sklern library to select the univariate characteristics of the death characteristic data set filled with the null value.
Preferably, the score function is an f _ regression () function.
Inputting the death characteristic data set subjected to characteristic selection into an XGboost model, setting general parameters, booster parameters and learning target parameters of the XGboost model, and training the XGboost model. The general parameters are used for determining boosters selected in the lifting process of the XGboost model, the booster parameters are used for setting the types of the boosters, the learning target parameters are used for controlling the learning scene of the XGboost model, and the type of the booster selected is a tree model.
The XGboost algorithm flow specifically comprises the following steps: initializing a death characteristic data set; defining a loss function, and calculating a first derivative value of each sample value in the death characteristic data set by the loss function; establishing a new decision tree according to the first derivative value; predicting the sample value by using the new decision tree and accumulating the sample value to the original predicted value; and circulating the steps until the set end condition is met, and generating a second mortality prediction model. The XGboost model has an objective function satisfying the following formula:
Figure BDA0003668705040000181
Figure BDA0003668705040000182
Figure BDA0003668705040000183
wherein t is iteration frequency, λ and γ are penalty strength, λ is used for controlling the number of nodes of leaves of each decision tree, λ is used for ensuring that the node fraction of the leaves of each decision tree is within a normal interval, g (·) is a loss function of the XGboost model,
Figure BDA0003668705040000184
for the ith sample in the death feature data set,
Figure BDA0003668705040000191
and B, the predicted value of the ith sample after t-1 iterations, wherein A is the set of nodes of the leaves of the r decision tree.
In step S222 of this embodiment, the performance of the second mortality prediction model is evaluated according to the second test set and the second evaluation index. In order to ensure consistency of weights given to the first mortality prediction model and the second mortality prediction model in the subsequent step S500, a reference of a second evaluation index of the present application is consistent with a reference of the first evaluation index, the second evaluation index is also set based on an average absolute error and a root mean square error formula, and the second evaluation index specifically is:
Figure BDA0003668705040000192
Figure BDA0003668705040000193
α 2 ＝30％×α 22 +70％×α 21 ；
wherein alpha is 21 Average absolute error value, alpha, for the second mortality prediction model 22 Root mean square error value, α, for the second mortality prediction model 2 Error value, n, for the second mortality prediction model 2 Is the amount of data of the second test set,
Figure BDA0003668705040000194
for the mortality prediction value, k, output by the second mortality prediction model after the second test set is input i The mortality true value corresponding to the second test set.
Similarly, a second error threshold value is set, if the error value of the second mortality prediction model is not less than or equal to the second error threshold value, the hyperparameter of the XGboost algorithm is readjusted, and retraining is performed to obtain a second mortality prediction model with smaller error.
In this embodiment, in step S300, in order to obtain the environmental data and the growth data of the seedling in the current growth cycle, the environmental data and the growth data of the seedling in the current growth cycle are preprocessed to generate a current feature data set of the seedling.
Further, the environment data is output environment data obtained by a temperature sensor, a pH sensor, a sodium ion sensor, a residual chlorine sensor, an enzyme biosensor and a nitrate sensor, the temperature sensor is used for collecting temperature data of the environment where flowers are located, the pH sensor is used for collecting pH data of the environment where the flowers are located, the sodium ion sensor is used for collecting sodium ion concentration of soil where the flowers are located, the residual chlorine sensor is used for collecting chloride ion concentration of the soil where the flowers are located, the enzyme biosensor is used for collecting enzyme concentration of the soil where the flowers are located, the nitrate sensor is used for collecting nitrate ion concentration of the soil where the flowers are located, and the environment data comprises physical quantities of temperature, pH value, sodium ion concentration, chloride ion concentration, nitrate example concentration and enzyme concentration;
the growth data is output by a first image sensor, a second image sensor and a processing module, the first image sensor is a fluorescence image sensor, and the first image sensor is used for detecting the chlorophyll concentration in the seedling; the second image sensor is a high image sensor and is used for detecting the spectral characteristics of the seedlings and outputting the spectral characteristics to the processing module, and the processing module is used for obtaining the quantity of wormholes of the seedlings according to the spectral characteristics; the characteristic key data comprise chlorophyll concentration and seedling wormhole number.
In the embodiment, all flower seedlings and soil in which the flower seedlings are located reflect certain light energy to the atmosphere in the flower growing process, the light energy reflected by the flower seedlings is called spectral characteristics, the high-image sensor is used for acquiring and recording the spectral characteristics of each flower seedling, when pests invade the seedlings, the pests absorb the light of the seedlings to cause the spectral characteristics of the seedlings to change, and the high-image sensor detects the spectral characteristics of the seedlings and outputs the spectral characteristics to the processing module for processing, so that the number of wormholes of the seedlings can be obtained.
In this embodiment, the environmental data and growth data of the seedling in the current growth cycle are preprocessed in step S300 in the same way as the characteristic data of the seedling and its corresponding mortality data are preprocessed in step S100.
In this embodiment, step S400 is to use the current feature data set as an input of a first mortality prediction model and an input of a second mortality prediction model, where the first mortality prediction model predicts and outputs a first predicted mortality, and the second mortality prediction model predicts and outputs a second predicted mortality.
In this embodiment, step S500 is to respectively assign weights corresponding to the first predicted mortality and the second predicted mortality, and add the first predicted mortality assigned with the first weight and the second predicted mortality assigned with the second weight to obtain the comprehensive mortality of the seedling in the next growth cycle.
Specifically, the application gives a first weight to the first predicted mortality and a second weight to the second predicted mortality based on the reciprocal residual error method, and the following results can be obtained according to the reciprocal residual error method: when the error value of a mortality prediction model is larger, the weight ratio of the mortality prediction model in the combined model is smaller.
Based on the above, the first weight and the second weight satisfy the following formula:
Figure BDA0003668705040000211
Figure BDA0003668705040000212
wherein, delta 1 Representing a first weight, δ, assigned to a first mortality prediction model 2 Representing a second weight, α, assigned to a second mortality prediction model 11 Is the mean absolute error value, α, of the first mortality prediction model 12 Root mean square error value, α, for the first mortality prediction model 21 Average absolute error value, alpha, for the second mortality prediction model 22 A root mean square error value for the second mortality prediction model.
Thus, the overall mortality rate of the seedlings in the next growth cycle is:
D＝δ 1 y 1 +δ 2 y 2 ；
wherein D represents the overall mortality, δ 1 Representing a first weight, δ, assigned to a first mortality prediction model 2 Representing a second weight, y, assigned to a second mortality prediction model 1 For the first prediction of mortality, y 2 Mortality was predicted for the second.
Based on the above embodiment, the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method for predicting the death rate of seedlings in the flower cultivation process based on deep learning is implemented.
Based on the above embodiment, the present application further provides a flower seedling mortality prediction system, the system includes: at least one processor; at least one memory for storing at least one program; when the at least one program is executed by the at least one processor, the at least one processor implements the method for predicting seedling mortality in a deep learning-based flower cultivation process.
While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that the present invention is not limited to the details of the embodiments shown and described, but is capable of numerous equivalents and substitutions without departing from the spirit of the invention as set forth in the claims appended hereto.

Claims (10)

1. A seedling death rate prediction method in a flower cultivation process based on deep learning is characterized by comprising the following steps:
s100, acquiring characteristic data of seedlings in the whole growth period and corresponding mortality data thereof based on flower big data, preprocessing the characteristic data of the seedlings and the corresponding mortality data thereof, and generating a death characteristic data set of the seedlings;
wherein the characteristic data comprises environmental temperature, environmental pH value, soil multi-ion concentration, soil enzyme biological concentration, seedling wormhole number and seedling chlorophyll concentration;
s200, establishing a first mortality prediction model and a second mortality prediction model according to the death characteristic data set;
s300, acquiring environmental data and growth data of the seedlings in the current growth cycle, preprocessing the environmental data and the growth data of the seedlings in the current growth cycle, and generating a current characteristic data set of the seedlings;
s400, according to the current characteristic data set, the first mortality prediction model predicts the mortality of the seedlings in the next growth cycle and outputs a first predicted mortality, and the second mortality prediction model predicts the mortality of the seedlings in the next growth cycle and outputs a second predicted mortality;
s500, giving the first predicted death rate to a first weight, giving the second predicted death rate to a second weight, and accumulating the first predicted death rate given to the first weight and the second predicted death rate given to the second weight to obtain the comprehensive death rate of the seedling in the next growth cycle.
2. A seedling mortality prediction method in a flower cultivation process based on deep learning according to claim 1, wherein in step S100, the step of preprocessing the characteristic data of the seedling and the corresponding mortality data thereof is as follows:
s110, finding out characteristic data of seedlings and a missing value of corresponding mortality data of the seedlings, and completing the missing value through a regression algorithm;
s120, searching for the characteristic data of the seedlings and the repetition values and abnormal values of the mortality data corresponding to the characteristic data, and discarding the repetition values and the abnormal values;
s130, reducing the dimension of the characteristic data of the seedlings and the corresponding mortality data of the seedlings;
s140, standardizing the characteristic data of the seedlings and the corresponding mortality data thereof to generate a death characteristic data set of the seedlings.
3. A seedling mortality prediction method in a flower cultivation process based on deep learning according to claim 1, wherein in step S200, the establishing of the first mortality prediction model and the second mortality prediction model according to the death feature data set comprises a first mortality prediction model establishing step and a second mortality prediction model establishing step;
the first mortality prediction model establishing step comprises the steps of:
s210, converting the format of the death characteristic data set into an input format conforming to a convolutional neural network, and dividing the death characteristic data set after the format conversion into a first training set and a first testing set according to a ratio of 8: 2;
s211, inputting the first training set into a convolutional neural network, training the convolutional neural network, and obtaining a first feature vector;
s212, converting the format of the first feature vector into an input format conforming to the bidirectional long and short term memory neural network, inputting the first feature vector after format conversion into the bidirectional long and short term memory neural network, training the bidirectional long and short term memory neural network, and obtaining a second feature vector;
s213, constructing a result classification layer and constructing an attention layer based on an attention mechanism, wherein the second feature vector is input into the attention layer, the output result of the attention layer is input into the result classification layer, and the result classification layer outputs to generate a first mortality prediction model;
s214, setting a first evaluation index, and evaluating the performance of the first mortality prediction model according to the first test set and the first evaluation index;
the second mortality prediction model establishing step comprises:
s220, performing feature selection on the death feature data set;
s221, dividing the death data set after feature selection into a second training set and a second testing set according to the ratio of 8:2, inputting the second training set into the XGboost model, training the XGboost model, and generating a second mortality prediction model;
and S222, evaluating the performance of the second mortality prediction model according to the second test set and the second evaluation index.
4. A seedling mortality prediction method in a flower cultivation process based on deep learning according to claim 3, wherein the attention mechanism satisfies the following formula:
Figure FDA0003668705030000031
wherein,
Figure FDA0003668705030000032
representing the output result of the attention layer, K is a second feature vector, ω is a preset training parameter vector, and softmax represents the operation in the softmax classifier.
5. The seedling mortality prediction method in the flower cultivation process based on deep learning of claim 4, wherein the step of feature selection of the death feature data set comprises: inquiring a null value in a death characteristic data set, and calling a median of a column where the null value is located to fill; and calling a scoring function of the sklern library to perform univariate feature selection on the death feature data set filled with the null value, wherein the scoring function is an f _ regression () function.
6. The method for predicting seedling mortality in a flower cultivation process based on deep learning according to claim 5, wherein the first evaluation index satisfies the following formula:
Figure FDA0003668705030000041
Figure FDA0003668705030000042
α 1 ＝30％×α 12 +70％×α 11 ；
wherein alpha is 11 Average absolute error value, alpha, for the first mortality prediction model 12 Root mean square error value, α, for the first mortality prediction model 1 Error value, n, for the first mortality prediction model 1 Is the amount of data in the first test set,
Figure FDA0003668705030000046
for the mortality prediction output by the first mortality prediction model after the input of the first test set, m i The death rate true value corresponding to the first test set;
the second evaluation index satisfies the following formula:
Figure FDA0003668705030000043
Figure FDA0003668705030000044
α 2 ＝30％×α 22 +70％×α 21 ；
wherein alpha is 21 Average absolute error value, alpha, for the second mortality prediction model 22 Root mean square error value, α, for the second mortality prediction model 2 Error value, n, for the second mortality prediction model 2 Is the amount of data in the second test set,
Figure FDA0003668705030000045
for the mortality prediction value, k, output by the second mortality prediction model after the second test set is input i The mortality true value corresponding to the second test set.
7. The method for predicting seedling mortality during flower cultivation based on deep learning of claim 6, wherein in step S500, the first weight satisfies the following formula:
Figure FDA0003668705030000051
wherein, delta 1 Representing a first weight, α, assigned to a first mortality prediction model 11 Average absolute error value, alpha, for the first mortality prediction model 12 Root mean square error value, α, for the first mortality prediction model 21 Mean absolute error value, α, for the second mortality prediction model 22 A root mean square error value for the second mortality prediction model;
the second weight satisfies the following formula:
Figure FDA0003668705030000052
wherein, delta 2 Representing a second weight, α, assigned to a second mortality prediction model 11 Average absolute error value, alpha, for the first mortality prediction model 12 Root mean square error value, α, for the first mortality prediction model 21 Average absolute error value, alpha, for the second mortality prediction model 22 A root mean square error value for the second mortality prediction model;
the comprehensive mortality rate of the seedlings in the next growth cycle satisfies the following formula:
D＝δ 1 y 1 +δ 2 y 2 ；
wherein D represents the overall mortality, δ 1 Representing a first weight, δ, assigned to a first mortality prediction model 2 Representing a second weight, y, assigned to a second mortality prediction model 1 For the first prediction of mortality, y 2 Mortality was predicted for the second.
8. The method for predicting the death rate of seedlings in the flower cultivation process based on the deep learning as claimed in claim 1, wherein in step S300, the environmental data is the environmental data obtained and output by a temperature sensor, a pH sensor, a sodium ion sensor, a residual chlorine sensor, an enzyme biosensor and a nitrate sensor, the temperature sensor is used for collecting the temperature data of the environment where flowers are located, the pH sensor is used for collecting the pH data of the environment where flowers are located, the sodium ion sensor is used for collecting the sodium ion concentration of the soil where flowers are located, the residual chlorine sensor is used for collecting the chloride ion concentration of the soil where flowers are located, the enzyme biosensor is used for collecting the enzyme concentration of the soil where flowers are located, the nitrate sensor is used for collecting the nitrate ion concentration of the soil where flowers are located, and the environmental data comprises the temperature, the pH value, the residual chlorine ion concentration and the nitrate ion concentration, Physical quantities of sodium ion concentration, chloride ion concentration, nitrate example concentration, enzyme concentration;
the growth data is output by a first image sensor, a second image sensor and a processing module, the first image sensor is a fluorescence image sensor, and the first image sensor is used for detecting the chlorophyll concentration in the seedling; the second image sensor is a high image sensor and is used for detecting the spectral characteristics of the seedlings and outputting the spectral characteristics to the processing module, and the processing module is used for obtaining the quantity of wormholes of the seedlings according to the spectral characteristics; the characteristic key data comprise chlorophyll concentration and seedling wormhole number.
9. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method for predicting mortality of seedlings in deep learning-based flower cultivation process according to any one of claims 1 to 8 is implemented.
10. A system for predicting seedling mortality in a flower cultivation process based on deep learning, the system comprising:
at least one processor;
at least one memory for storing at least one program;
when the at least one program is executed by the at least one processor, the at least one processor may implement the method for predicting mortality of seedlings in a deep learning-based flower cultivation process according to any one of claims 1 to 8.
CN202210597524.8A 2022-05-30 2022-05-30 Method for predicting death rate of seedlings in flower cultivation process based on deep learning Pending CN114936699A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210597524.8A CN114936699A (en) 2022-05-30 2022-05-30 Method for predicting death rate of seedlings in flower cultivation process based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210597524.8A CN114936699A (en) 2022-05-30 2022-05-30 Method for predicting death rate of seedlings in flower cultivation process based on deep learning

Publications (1)

Publication Number Publication Date
CN114936699A true CN114936699A (en) 2022-08-23

Family

ID=82867269

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210597524.8A Pending CN114936699A (en) 2022-05-30 2022-05-30 Method for predicting death rate of seedlings in flower cultivation process based on deep learning

Country Status (1)

Country Link
CN (1) CN114936699A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115796324A (en) * 2022-09-08 2023-03-14 呼伦贝尔安泰热电有限责任公司海拉尔热电厂 Method and system for predicting heat supply load in alpine region

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115796324A (en) * 2022-09-08 2023-03-14 呼伦贝尔安泰热电有限责任公司海拉尔热电厂 Method and system for predicting heat supply load in alpine region
CN115796324B (en) * 2022-09-08 2023-11-03 呼伦贝尔安泰热电有限责任公司海拉尔热电厂 Method and system for predicting heat supply load in alpine region

Similar Documents

Publication Publication Date Title
Fdez-Riverola et al. CBR based system for forecasting red tides
Khairunniza-Bejo et al. Application of artificial neural network in predicting crop yield: A review
Huang et al. Development of soft computing and applications in agricultural and biological engineering
Sharma et al. Wheat crop yield prediction using deep LSTM model
Zhang et al. Modeling tree-ring growth responses to climatic variables using artificial neural networks
Campos Banana production in Venezuela: Novel solutions to productivity and plant health
CN115829162B (en) Crop yield prediction method, device, electronic equipment and medium
Bal et al. Review of machine learning and deep learning models in agriculture
CN115222100A (en) Crop yield prediction method based on three-dimensional cyclic convolution neural network and multi-temporal remote sensing image
Gopal et al. Optimum feature subset for optimizing crop yield prediction using filter and wrapper approaches
Jiménez et al. Evaluation of two recurrent neural network methods for prediction of irrigation rate and timing
CN114936699A (en) Method for predicting death rate of seedlings in flower cultivation process based on deep learning
Stathakis et al. Neuro-fuzzy modeling for crop yield prediction
Thamilselvan et al. An extensive review of rainfall prediction using machine learning and deep learning techniques
Vasilakos et al. Granular neural networks for land use classification
Zhao et al. Optimizing Irrigation Scheduling Using Deep Reinforcement Learning
Bodapati et al. Analyzing crop yield using machine learning
El-Telbany et al. Mining the Classification Rules for Egyptian Rice Diseases.
Golian et al. Study of Methods for Determining Types and Measuring of Agricultural Crops due to Satellite Images
Metta et al. Drought prediction using artificial neural network
Anitha 5 Adaptation of XAI
CN118097435B (en) Supergraph neural network-based corn lodging classification method and device
Qiu Rice yield prediction based on LSTM and GRU
Ahlawat et al. A review on analysis and prediction methodologies of crop yield pattern
Zeng et al. Research on Maize Yield Prediction Model Based on CNN-LSTM-PSO

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