CN-121074437-B - Intelligent agricultural crop growth analysis system based on big data

CN121074437BCN 121074437 BCN121074437 BCN 121074437BCN-121074437-B

Abstract

The invention discloses a big data-based intelligent agricultural crop growth analysis system, which is characterized in that initial crop data are acquired through a sensor array, noise reduction processing is carried out on the initial crop data through a wavelet denoising-Kalman filtering coupling algorithm, feature extraction is carried out on image data based on an improved MobileNetV network, crop morphological feature vectors are output, association analysis is carried out on environment and physiology through an LSTM (link state model) construction time sequence network, dynamic relation between learning parameters and photosynthetic efficiency is established, a crop state prediction model is established, the noise reduction crop data are input into the prediction model for prediction, growth health scores are output, crop health grade judgment is carried out according to the growth health scores, and a crop growth intervention strategy is generated according to judgment results. The growth health score of crops can be predicted more accurately, and the resource utilization efficiency is improved.

Inventors

HOU WENFANG
CHENG SHI
YANG TINGTING
XIA FEI
LI XIAODONG
WANG HUIYUN
CHEN HONG
LIU TAO
WANG BENXIANG
DUAN JIFENG

Assignees

南通大学

Dates

Publication Date: 20260505
Application Date: 20250801

Claims (8)

1. The intelligent agricultural crop growth analysis system based on big data is characterized by comprising the following modules: The data acquisition processing module is used for acquiring image data, environment data and physiological data of crops through the sensor array to obtain multidimensional crop data, and carrying out data preprocessing on the multidimensional crop data to obtain initial crop data; The crop data denoising module is used for dividing continuous initial crop data flow into data blocks according to the data type and the time stamp of the initial crop data, carrying out time alignment on the divided data blocks to obtain aligned crop data, carrying out wavelet transformation on each data block in the aligned crop data by using a Daubechies wavelet function, decomposing an original signal into different frequency layers to obtain decomposed crop data, analyzing amplitude distribution of the decomposed crop data, estimating noise level according to statistical characteristics, carrying out data screening according to a threshold value to obtain screened crop data, synthesizing an approximation coefficient and a detail coefficient of the screened crop data through wavelet inverse transformation to obtain a signal block after wavelet denoising to obtain filtered crop data, establishing a dynamic model based on the environmental temperature, the humidity and the illumination intensity in environmental parameters and the soil moisture and the stem flow rate in the physiological parameters, defining process noise and observation noise, predicting a state value and uncertainty of the parameter at the current moment according to the best estimation and the defined dynamic model at the previous moment, obtaining an actual observation value of the filtered crop data channel at the current moment, carrying out data fusion on the predicted value and the best estimation and the observation value at the current moment, and obtaining the best variance of the filtered crop data as the final estimation and the best variance-reduction result; The prediction model building module is used for carrying out feature extraction on image data based on an improved MobileNetV network and outputting crop morphology feature vectors, carrying out association analysis on environment and physiology through an LSTM (least squares) construction time sequence network, learning the dynamic relation between parameters and photosynthetic efficiency, and building a MobileNetV-LSTM crop state prediction model; the crop state identification module is used for inputting the noise reduction crop data into the MobileNetV-LSTM crop state prediction model for prediction and outputting a growth health score; And the crop growth intervention module is used for judging the health grade of the crop according to the growth health score and generating a crop growth intervention strategy according to a judging result.
2. The intelligent agricultural crop growth analysis system based on big data according to claim 1, wherein the data acquisition processing module comprises the following sub-modules: the conversion sub-module is used for carrying out gray level conversion on the image data in the multi-dimensional crop data and converting the multi-spectral image into a gray level image; the enhancer module is used for enhancing the contrast of the image data in the gray level image by using a histogram equalization method to obtain processed image data; the unified sub-module is used for carrying out format conversion on the environmental data and the physiological data and converting signals output by different sensors into a unified digital format; And the obtained submodule is used for detecting abnormal values by adopting a Z-score method, repairing the abnormal values exceeding 3 times of standard deviation by utilizing an adjacent data interpolation method, and obtaining initial crop data.
3. The intelligent agricultural crop growth analysis system based on big data of claim 1, wherein the predictive model building module comprises the following sub-modules: an analysis sub-module for analyzing the input image based on MobileNetV network, discarding redundant visual information unrelated to crop status; The adjusting submodule is used for adjusting a feature extraction layer in the MobileNetV network and extracting crop-making growth features at least comprising leaf size, color, texture uniformity, plant density, plant diseases and insect pests spots and chlorophyll expression; The extraction submodule is used for extracting the crop morphological feature vector representing the appearance form, the growth situation and the potential health problem of the current crop and with high dimensionality.
4. The intelligent agricultural crop growth analysis system based on big data of claim 1, wherein the predictive model building module further comprises the following sub-modules: The input sub-module is used for inputting the environmental time sequence data and the physiological time sequence data in the noise reduction crop data into the LSTM construction time sequence network; the learning sub-module is used for utilizing the LSTM to construct a time sequence network to analyze the light intensity, the CO 2 , the stomatal conductivity index and the chlorophyll level indication characteristic parameters which are directly related to the photosynthetic efficiency, and the law of evolution of the learning parameters along with time and the mutual influence mode are utilized; and the integration sub-module is used for integrating and encoding the environmental physiological history information and the dynamic association information in the whole time window into a context state vector containing time sequence information.
5. A smart agricultural crop growth analysis system based on big data as claimed in claim 1, wherein the crop growth intervention module comprises the following sub-modules: And the matching sub-module is used for judging the health grade of the crops according to the growth health score, establishing an agricultural expert knowledge base, and matching corresponding intervention strategies from the knowledge base according to the health grade judgment result of the crops and combining current environmental data and physiological data.
6. A method of implementing a big data based intelligent agricultural crop growth analysis system according to claim 1, the method comprising the steps of: acquiring image data, environment data and physiological data of crops through a sensor array to obtain multidimensional crop data, and performing data preprocessing on the multidimensional crop data to obtain initial crop data; Dividing continuous initial crop data flow into data blocks according to the data type and time stamp of the initial crop data, carrying out time alignment on the divided data blocks to obtain aligned crop data, carrying out wavelet transformation on each data block in the aligned crop data by utilizing Daubechies wavelet function, decomposing an original signal into different frequency layers to obtain decomposed crop data, analyzing amplitude distribution of the decomposed crop data, estimating noise level according to statistical characteristics, carrying out data screening according to a threshold value to obtain screened crop data, synthesizing approximation coefficients and detail coefficients of the screened crop data through wavelet inverse transformation to obtain a signal block subjected to wavelet noise reduction to obtain filtered crop data, establishing a dynamic model based on environmental temperature, humidity and illumination intensity in environmental parameters and soil moisture and stem flow rate in physiological parameters, defining process noise and observation noise, predicting a state value and uncertainty of the parameter at the current moment according to the optimal estimation and the defined dynamic model at the previous moment, obtaining an actual value of a corresponding data channel of the filtered crop data at the current moment, carrying out data screening according to a statistical characteristic, carrying out inverse transformation on the approximation value and the weighted value to obtain a final estimated result of the filtered crop data at the current moment, and obtaining the optimal estimated result of noise reduction in a final estimated result of the current crop noise reduction; Performing association analysis on environment and physiology through an LSTM construction time sequence network, learning the dynamic relationship between parameters and photosynthetic efficiency, and establishing a MobileNetV-LSTM crop state prediction model; Inputting the noise reduction crop data into the MobileNetV-LSTM crop state prediction model for prediction, and outputting a growth health score; and judging the health grade of the crops according to the growth health score, and generating a crop growth intervention strategy according to a judging result.
7. A method of implementing a big data based intelligent agricultural crop growth analysis system according to claim 6, said method comprising the steps of: performing gray level conversion on image data in the multi-dimensional crop data, and converting the multi-spectral image into a gray level image; enhancing the contrast of the image data by using a histogram equalization method to obtain processed image data; Performing format conversion on the environmental data and the physiological data, and converting signals output by different sensors into a uniform digital format; And detecting abnormal values by adopting a Z-score method, and repairing abnormal values exceeding 3 times of standard deviation by utilizing an adjacent data interpolation method to obtain initial crop data.
8. A method of implementing a big data based intelligent agricultural crop growth analysis system of claim 7, the method comprising the steps of: Dividing continuous initial crop data streams into data blocks according to the data types and the time stamps of the initial crop data, and performing time alignment on the divided data blocks to obtain aligned crop data; Performing wavelet transformation on each data block in the aligned crop data by using a Daubechies wavelet function, and decomposing an original signal into different frequency layers to obtain decomposed crop data; Analyzing the amplitude distribution of the decomposed crop data, estimating the noise level according to the statistical characteristics, and carrying out data screening according to a threshold value to obtain screened crop data; And synthesizing the approximate coefficient and the detail coefficient of the screened crop data through wavelet inverse transformation to obtain a signal block after wavelet noise reduction to obtain filtered crop data.

Description

Intelligent agricultural crop growth analysis system based on big data Technical Field The invention relates to the technical field of crop growth analysis, in particular to an intelligent agricultural crop growth analysis system based on big data. Background At present, the data acquisition in the agricultural production has various problems. The traditional feature extraction and prediction model has limited capability of excavating the growth rule of crops, and can not effectively solve the problem of fusion of image data with environmental and physiological time sequence data. The convolutional neural network is difficult to capture the dynamic influence of environmental factors on crop growth, and the time sequence model is only relied on, so that the crop morphological characteristic information cannot be fully utilized. Meanwhile, the existing crop health assessment and intervention strategies lack of accuracy, and most of the existing crop health assessment and intervention strategies adopt a one-cut management mode, so that personalized and refined growth requirements of crops are difficult to meet, and resource waste and yield loss are caused. Disclosure of Invention The invention aims to solve the problems, and designs an intelligent agricultural crop growth analysis system based on big data. The technical scheme for achieving the purpose is that in the intelligent agricultural crop growth analysis system based on big data, the intelligent agricultural crop growth analysis system method comprises the following steps: The data acquisition processing module is used for acquiring image data, environment data and physiological data of crops through the sensor array to obtain multidimensional crop data, and carrying out data preprocessing on the multidimensional crop data to obtain initial crop data; The crop data denoising module is used for denoising the initial crop data by utilizing a wavelet denoising-Kalman filtering coupling algorithm to obtain denoising crop data; the prediction model building module is used for carrying out feature extraction on image data based on an improved Mobil eNetV network and outputting crop morphology feature vectors, carrying out association analysis on environment and physiology through an LSTM (least squares) construction time sequence network, learning the dynamic relation between parameters and photosynthetic efficiency, and building a Mobil eNetV-LSTM crop state prediction model; The crop state identification module is used for inputting the noise reduction crop data into the Mobil eNetV-LSTM crop state prediction model for prediction and outputting a growth health score; And the crop growth intervention module is used for judging the health grade of the crop according to the growth health score and generating a crop growth intervention strategy according to a judging result. Further, in the intelligent agricultural crop growth analysis system based on big data, the data acquisition processing module comprises the following submodules: the conversion sub-module is used for carrying out gray level conversion on the image data in the multi-dimensional crop data and converting the multi-spectral image into a gray level image; an enhancer module for enhancing the contrast of the image data by using a histogram equalization method to obtain processed image data; the unified sub-module is used for carrying out format conversion on the environmental data and the physiological data and converting signals output by different sensors into a unified digital format; And the obtained submodule is used for detecting abnormal values by adopting a Z-score method, repairing the abnormal values exceeding 3 times of standard deviation by utilizing an adjacent data interpolation method, and obtaining initial crop data. Further, in the intelligent agricultural crop growth analysis system based on big data, the crop data noise reduction module comprises the following submodules: an alignment submodule, configured to divide a continuous initial crop data stream into data blocks according to a data type and a time stamp of the initial crop data, and time align the divided data blocks to obtain aligned crop data; the decomposition sub-module is used for carrying out wavelet transformation on each data block in the aligned crop data by utilizing Daubech i es wavelet functions, and decomposing an original signal into different frequency layers to obtain decomposed crop data; the screening submodule is used for analyzing the amplitude distribution of the decomposed crop data, estimating the noise level according to the statistical characteristics, and carrying out data screening according to the threshold value to obtain screened crop data; and the obtaining submodule is used for synthesizing the approximate coefficient and the detail coefficient of the screened crop data through wavelet inverse transformation to obtain a signal block after wavelet noise reduction to obtain the filtered c