CN111652867A - Heavy-environment-free plant growth method based on deep morphological belief network - Google Patents
Heavy-environment-free plant growth method based on deep morphological belief network Download PDFInfo
- Publication number
- CN111652867A CN111652867A CN202010484897.5A CN202010484897A CN111652867A CN 111652867 A CN111652867 A CN 111652867A CN 202010484897 A CN202010484897 A CN 202010484897A CN 111652867 A CN111652867 A CN 111652867A
- Authority
- CN
- China
- Prior art keywords
- environment
- deep
- network
- training
- data
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a heavy-environment-free plant growth method based on a deep morphological confidence network, which comprises the following steps: collecting and preprocessing multi-source heterogeneous data of a crop growth environment without a heavy environment; performing feature extraction on the training set data by using a mathematical morphology operator; constructing and training a deep belief network; collecting field data of crops to be tested, and carrying out pretreatment and feature extraction to obtain a test data set; using the trained deep belief network to automatically predict the growth situation of the test data set; and automatically generating a crop growth situation monitoring analysis report under the environment according to the prediction result of the depth confidence network. The invention has high precision of monitoring and analyzing the growth situation of the crops in the weightless environment, wide application range and strong robustness.
Description
Technical Field
The invention relates to a heavy-environment-free plant growth method based on a deep morphological confidence network, and belongs to the technical field of crop monitoring.
Background
In long-term space flight, proper food replenishment is very important because pilots are in a weightless environment for a long time. Although the variety of space food is increasingly abundant at present, because about 14000 pounds of food is consumed and the price is high when every 1 kilogram of food is conveyed to an international space station, the space food is mainly high-calorie food such as dry food, rehydrated food, canned food and the like which can be stored for a long time, and enough fresh fruits and vegetables are lacked. This is detrimental to the pilot's physical and psychological health over long-term space flight.
To solve this problem, researchers in various countries have gradually investigated the possibility of growing crops in space under zero gravity, and some countries such as china and the united states have started practical tests. The attempt to plant crops in space is of great significance. On one hand, fresh fruits and vegetables are more beneficial to the physical health of astronauts, and meanwhile, things from the earth exist in the long-term space living environment, so that the psychological health of astronauts is influenced positively; on the other hand, growing crops in space can reduce food delivery costs. In addition, if plants are able to survive in space, there are more areas that can be utilized, for example, plants can consume carbon dioxide and produce oxygen, while also purifying the air of the presence of contaminants. The cost of space life of astronauts can be reduced, and the development of the technology is mature, so that the method has great significance for long-term life of human beings in space.
Although space planting has great significance, factors such as weight loss and cosmic rays in space can have unique influence on the survival, development, aging and variation of crops. In order to further research the possibility of space planting and improve the survival rate of space crop planting, the growth situation of crops in a weightless environment needs to be effectively monitored, and scientific and effective planting is carried out according to the monitoring condition. However, at present, no method related to crop growth situation monitoring under the weightlessness condition is available, so that a method for monitoring the crop growth situation under the space unique weightlessness environment needs to be designed in an urgent need.
Therefore, space crop planting is of great significance, and monitoring and analysis of growth situation of crops without heavy environment is a key ring for realizing space planting, but no relevant method is available at present. In order to solve the problems, improve the survival rate of space crop planting and further discuss the possibility of space planting, the invention provides a heavy-environment-free crop growth situation monitoring and analyzing method of a deep morphological confidence network.
Disclosure of Invention
The invention aims to overcome the defects of the existing method and provide a method for monitoring and analyzing the growth situation of the crops without the heavy environment, which has high precision, wide application range and strong robustness.
In order to achieve the purpose, the invention provides the following technical scheme:
a heavy-environment-free plant growing method based on a deep morphological confidence network comprises the following steps:
(1) collecting and preprocessing original multi-source heterogeneous data of crop growth situation in a weightless environment;
(2) performing feature extraction on the training set data by using a mathematical morphology operator;
(3) constructing and training a deep belief network;
(4) collecting field data of crops to be tested in a weightless environment, and carrying out pretreatment and feature extraction to obtain a test data set;
(5) monitoring and analyzing the crop growth situation of the test data set by using the trained deep confidence network;
(6) and automatically generating a monitoring and analyzing report of the growth situation of the crops without the heavy environment according to the diagnosis result of the deep confidence network.
The crop growth situation monitoring and analyzing method based on the deep morphology confidence network comprises the following steps of (1):
(1-1) arranging various sensors, collecting crop non-heavy environment related parameters, carbon dioxide on the upper part of a leaf surface, ultrasonic waves, small molecular atomized water mist, temperature, illumination intensity, humidity and the like, and real-scene images, infrared images and the like of the actual growth state of crops;
(1-2) cleaning, cutting and normalizing all the collected data to obtain independent variables of an original training data set;
(1-3) identifying and marking the growth situation of the crops in the next time period aiming at each group of independent variables under the same time domain degree to form a label of an original training data set;
the existing crop monitoring and analyzing technology generally only adopts one-dimensional crop growth environment parameters or two-dimensional crop growth state images as sensitive variables, because the existing mature classification model can only process source data with a single structure. However, aiming at the complexity of crop growth in a non-heavy environment, strong related information is easy to ignore and omit only by adopting data of a certain single structure, so that the monitoring and analyzing accuracy of the crop growth situation in the non-heavy environment is low, and the method cannot be practically applied. The invention considers the requirement of practical application, adopts multi-source heterogeneous data comprising one dimension and two dimensions, and ensures the integrity and accuracy of the original information base to the maximum extent.
The heavy-environment-free plant growing method based on the deep morphological belief network specifically comprises the following steps of (2):
(2-1) cutting, unifying the size and graying all the two-dimensional image data extracted in the step (1);
(2-2) extracting features of the two-dimensional image data set by using two-dimensional morphological operators of different kernels;
(2-3) extracting features of the one-dimensional data set by using one-dimensional morphological operators of different kernels;
(2-4) performing maximum pooling on the extracted feature data sets;
and (2-5) flattening the pooled feature data through a Flatten layer, and connecting the flattened feature data in series to form a uniform one-dimensional feature data set.
Preferably, the two-dimensional image pixel size is 256 × 256, the two-dimensional morphological operator size is 16 × 16, the one-dimensional morphological operator size is 1 × 16, and the pooling region size is 4 × 4.
The conventional method usually uses the conventional Fourier transform correlation technique for feature extraction, the core of the Fourier transform adopts convolution operation, and more consideration is given from the aspect of frequency domain, but the method usually consumes longer time in practical application, has larger data processing load and higher requirement on a central processing unit, so that image data can not be locally processed usually, and the data needs to be transmitted back to a data background and then uniformly processed by a high-performance central processing unit, so that the method has higher requirement on communication bandwidth. The method adopts a morphological method to extract the characteristics, is mainly considered based on a time domain angle, has high real-time calculation speed and light calculation load, can be directly arranged in field hardware, is uploaded after being quickly processed locally, reduces the communication bandwidth pressure to the maximum extent while accurately extracting the characteristics, and is beneficial to reducing the arrangement cost and expanding the application range.
The crop growth situation monitoring and analyzing method based on the deep morphology confidence network specifically comprises the following steps of (3):
(3-1) constructing a five-layer deep confidence network based on a restricted Boltzmann machine, wherein the five-layer deep confidence network comprises an input layer, three hidden layers and an output layer.
(3-2) performing forward unsupervised pre-training on the deep confidence network by using a CD-K algorithm, wherein K is generally 1, and the parameter updating formula is as follows:
W←W+×[P(h=1|v)vT-P(h*=1|v*)v*T](1)
b←b+×(v-v*) (2)
c←c+×[P(h=1|v)-P(h*=1|v*)](3)
wherein, W, b and c are respectively a connecting layer weight matrix, a bias vector of a visible layer and a bias vector of a hidden layer. v is a visible layer neuron vector, v is a visible layer neuron reconstruction vector, and is learning efficiency;
(3-3) carrying out supervised reverse fine tuning training on the deep confidence network by using a BP algorithm, wherein a parameter updating formula is as follows:
wherein, α is the learning efficiency,the neural weights are connected for the l-th layer,biasing weights for the l-th layer;
(3-4) storing the training optimized deep confidence network model;
the invention uses the field data with various sources and various data structures to monitor and analyze the crop growth situation in the weightless environment, and furthest ensures the integrity and the accuracy of the original information base. Meanwhile, the constructed deep morphology confidence network organically combines mathematical morphology with deep learning, and a morphology operator is used for carrying out feature extraction on multi-source heterogeneous data in a non-heavy environment, so that the influence of background and noise data is effectively weakened, and the effectiveness of the extracted features is greatly improved. In addition, the used deep confidence network has a plurality of hidden layers, has more excellent feature extraction capability compared with a common shallow network, and has excellent feature mining and recognition capability on crop growth situation related data in a weightless environment by combining a special unsupervised and supervised combined training mode. The method has better monitoring and predicting performance for monitoring and analyzing the growth situation of crops with dark light and complex background under the weightless environment, and has stronger robustness and more excellent practical application level.
Drawings
FIG. 1 is a flow chart of a method for monitoring and analyzing the growth situation of crops in a weightless environment.
Detailed Description
In order to more clearly describe the technical contents of the present invention, the following further description is made in conjunction with the accompanying drawings and embodiments.
As shown in fig. 1, the invention provides a method for monitoring and analyzing the growth situation of crops in a weightless environment based on a deep morphological confidence network, which comprises the following steps:
step (1):
and collecting original multi-source heterogeneous data of the crop growth environment in the weightless environment, and performing normalization pretreatment to form a training set. The method comprises the following specific steps:
(1-1) arranging various sensors on site, and acquiring relevant parameters of the crop growth environment, including carbon dioxide on the upper part of the leaf surface, ultrasonic waves, small molecular atomized water mist, temperature, illumination intensity, humidity and the like, and real-scene images, infrared images and the like of the actual growth state of crops. Sensor data can be transmitted to a data background in real time through a communication network for storage;
(1-2) cleaning, cutting and normalizing all the acquired one-dimensional and two-dimensional data at a data background to obtain independent variables of an original training data set;
(1-3) identifying and marking the growth situation of the crops in the next time period aiming at each group of independent variables under the same time domain degree to form a label of an original training data set;
step (2):
designing mathematical morphology operators of different kernels according to the collected data of different types for feature extraction, wherein the method comprises the following steps:
(2-1) cutting all the extracted two-dimensional image data, unifying the sizes, wherein optionally, the sizes of the pictures are unified to 256 × 256, and then performing graying processing on the pictures by using an averaging method;
(2-2) extracting features from the two-dimensional image dataset using two-dimensional morphological operators of different kernels, wherein optionally the morphological operator size is 16 x 16;
(2-3) extracting features from the one-dimensional dataset using one-dimensional morphological operators of different kernels, wherein optionally the morphological operator size is 1 x 16;
(2-4) performing maximum pooling on the extracted feature data set, wherein the size of a pooling area is 4 multiplied by 4;
and (2-5) flattening the pooled feature data through a Flatten layer, and connecting the flattened feature data in series to form a uniform one-dimensional feature data set.
Step (3)
The method comprises the following steps of constructing and training a deep confidence network, and specifically comprises the following steps:
and (3-1) constructing a five-layer deep confidence network, wherein the structure of the five-layer deep confidence network is formed by sequentially stacking limited Boltzmann machines and comprises an input layer, three hidden layers and an output layer. Optionally, the number of neurons in the input layer is consistent with the dimensionality of the feature data set. Optionally, the number of neurons in the intermediate hidden layer is 128, 256, 512, respectively. Optionally, the output layer is a softmax regression layer, and the number of neurons of the output layer is consistent with the number of categories of the plant diseases and insect pests to be detected.
(3-2) firstly carrying out unsupervised forward pre-training on the depth confidence network, carrying out greedy training on each restricted Boltzmann machine in the depth confidence network layer by using a CD-K algorithm, and calculating the weight and bias of each layer and the output values of three hidden layers, wherein the specific updating formula is as follows:
W←W+×[P(h=1|v)vT-P(h*=1|v*)v*T](6)
b←b+×(v-v*) (7)
c←c+×[P(h=1|v)-P(h*=1|v*)](8)
wherein, W, b and c are respectively a connecting layer weight matrix, a bias vector of a visible layer and a bias vector of a hidden layer. v is a visible layer neuron vector, v is a visible layer neuron reconstruction vector, and is learning efficiency;
(3-3) introducing a data set label, performing supervised reverse fine tuning training on the deep belief network by adopting a BP algorithm, adjusting the weight and the offset of all layers at one time, optimizing the parameters of the deep belief network, and completing global training, wherein a specific updating formula is as follows:
wherein, α is the learning efficiency,the neural weights are connected for the l-th layer,biasing weights for the l-th layer;
and (3-4) completely storing the deep confidence network after training optimization, wherein the deep confidence network comprises a network structure, network weight and bias, neuron number, an activation mode and the like.
Step (4)
Collecting field data of crops to be tested, and sequentially carrying out pretreatment and morphological feature extraction according to the operations in the steps (1) and (2) to obtain a test data set;
step (5)
Using the trained deep belief network to analyze and predict the growth situation of the test data set;
step (6)
And automatically generating a crop growth situation monitoring and analyzing report under the weightless environment according to the prediction result of the depth confidence network, accurately describing the future growth situation of the crops under the weightless environment by the report, providing corresponding measures for controlling growth environment parameters according to the growth situation, and improving the survival rate of the crops.
The method overcomes the defect that the existing mature classification model can only process source data with a single structure generally, and uses multi-source heterogeneous data for analysis, thereby ensuring the integrity and accuracy of the original information base to the maximum extent; meanwhile, the morphological operator is used for extracting the features, so that the influence of background and noise in the source data is effectively reduced, the feature extraction speed and precision are improved, and the requirements of various embedded devices can be well met; the confidence network with a deep structure gives full play to the strong characteristic extraction and analysis capability of the confidence network, and quickly and accurately judges the type of the plant diseases and insect pests.
In conclusion, the crop growth situation monitoring and analyzing method provided by the invention is comprehensive in consideration, high in judgment speed, low in requirement on a communication network, wide in application range and accurate in monitoring and analyzing result. The method well meets the requirements of future space planting on monitoring and analyzing the crop growth situation, has the condition of large-area popularization and use, and has important significance for scientifically guiding space crop planting, improving the crop survival rate under the weightless condition and achieving the purpose of space planting.
In this specification, the invention has been described with reference to specific embodiments thereof. Various modifications and alterations may be made by those skilled in the art without departing from the spirit and scope of the invention. Therefore, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (4)
1. A heavy-environment-free plant growing method based on a deep morphological confidence network is characterized by comprising the following steps:
(1) collecting and preprocessing multi-source heterogeneous data of a crop growth environment without a heavy environment;
(2) performing feature extraction on the training set data by using a mathematical morphology operator;
(3) constructing and training a deep belief network;
(4) collecting field data of crops to be tested, and carrying out pretreatment and feature extraction to obtain a test data set;
(5) using the trained deep belief network to perform growth situation automatic prediction analysis on the test data set;
(6) and automatically generating a crop growth situation monitoring analysis report under the environment according to the prediction result of the depth confidence network.
2. The deep morphology belief network-based gravity-free environment plant growing method according to claim 1, wherein the step (1) specifically comprises:
(1-1) arranging various sensors, collecting crop non-heavy environment related parameters, carbon dioxide on the upper part of a leaf surface, ultrasonic waves, small molecular atomized water mist, temperature, illumination intensity, humidity and the like, and real-scene images, infrared images and the like of the actual growth state of crops;
(1-2) cleaning, cutting and normalizing all the collected data to obtain independent variables of an original training data set;
and (1-3) identifying and marking the growth situation of the crops in the next time period aiming at each group of independent variables under the same time domain degree to form a label of an original training data set.
3. The deep morphology belief network-based gravity-free environment plant growing method according to claim 1, wherein the step (2) specifically comprises:
(2-1) cutting, unifying the size and graying all the two-dimensional image data extracted in the step (1);
(2-2) extracting features of the two-dimensional image data set by using two-dimensional morphological operators of different kernels;
(2-3) extracting features of the one-dimensional data set by using one-dimensional morphological operators of different kernels;
(2-4) performing maximum pooling on the extracted feature data sets;
and (2-5) flattening the pooled feature data through a Flatten layer, and connecting the flattened feature data in series to form a uniform one-dimensional feature data set.
4. The deep morphology belief network-based gravity-free environment plant growing method according to claim 1, wherein the step (3) specifically comprises:
(3-1) constructing a five-layer deep confidence network based on a restricted Boltzmann machine, wherein the five-layer deep confidence network comprises an input layer, three hidden layers and an output layer.
(3-2) performing forward unsupervised pre-training on the deep confidence network by using a CD-K algorithm;
(3-3) carrying out supervised reverse fine tuning training on the deep confidence network by using a BP algorithm;
and (3-4) storing the deep confidence network model after training optimization.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010484897.5A CN111652867A (en) | 2020-06-01 | 2020-06-01 | Heavy-environment-free plant growth method based on deep morphological belief network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010484897.5A CN111652867A (en) | 2020-06-01 | 2020-06-01 | Heavy-environment-free plant growth method based on deep morphological belief network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111652867A true CN111652867A (en) | 2020-09-11 |
Family
ID=72348935
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010484897.5A Pending CN111652867A (en) | 2020-06-01 | 2020-06-01 | Heavy-environment-free plant growth method based on deep morphological belief network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111652867A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112800929A (en) * | 2021-01-25 | 2021-05-14 | 安徽农业大学 | On-line monitoring method for bamboo shoot quantity and high growth rate based on deep learning |
-
2020
- 2020-06-01 CN CN202010484897.5A patent/CN111652867A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112800929A (en) * | 2021-01-25 | 2021-05-14 | 安徽农业大学 | On-line monitoring method for bamboo shoot quantity and high growth rate based on deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109191455A (en) | A kind of field crop pest and disease disasters detection method based on SSD convolutional network | |
CN107169621A (en) | A kind of Dissolved Oxygen in Water Forecasting Methodology and device | |
CN110119169A (en) | A kind of tomato greenhouse temperature intelligent early warning system based on minimum vector machine | |
Wang et al. | Digital image processing technology under backpropagation neural network and K-Means Clustering algorithm on nitrogen utilization rate of Chinese cabbages | |
Hariri et al. | Tipburn disorder detection in strawberry leaves using convolutional neural networks and particle swarm optimization | |
Diao et al. | Navigation line extraction algorithm for corn spraying robot based on improved YOLOv8s network | |
Xiao et al. | Leaf disease detection based on lightweight deep residual network and attention mechanism | |
CN111652867A (en) | Heavy-environment-free plant growth method based on deep morphological belief network | |
Sharma et al. | A New Dawn for Tomato-spotted wilt virus Detection and Intensity Classification: A CNN and LSTM Ensemble Model | |
CN114862596A (en) | Crop bank loan data risk analysis method and device | |
Sudha et al. | Real time riped fruit detection using faster R-CNN deep neural network models | |
Yang et al. | Identification of high-oil content soybean using hyperspectral reflectance and one-dimensional convolutional neural network | |
Liu et al. | Tomato detection based on convolutional neural network for robotic application | |
Sun et al. | A recognition method of cattle and sheep based on convolutional neural network | |
CN116863403A (en) | Crop big data environment monitoring method and device and electronic equipment | |
Cum et al. | Machine learning models comparison for water stress detection based on stem electrical impedance measurements | |
Aljaafreh et al. | A Real-Time Olive Fruit Detection for Harvesting Robot Based on YOLO Algorithms | |
Huang et al. | Application of data augmentation and migration learning in identification of diseases and pests in tea trees | |
Wei et al. | Small sample and efficient crop pest recognition method based on transfer learning and data transformation | |
Balaji et al. | Plant Infirmity Detection Using Vgg-16 Convolutional Neural Network | |
Tan et al. | Pepper Target Recognition and Detection Based on Improved YOLO v4 | |
Dong et al. | Measurement method of plant phenotypic parameters based on image deep learning | |
CN110717506A (en) | Hyperspectral remote sensing image classification method based on convolutional neural network | |
Mahalakshmi et al. | Agricultural Crop and Fertilizer Recommendations based on Various Parameters | |
Taji et al. | A Systematic Literature Review of Computational Studies in Aquaponic System |
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 |