CN116863340A - Rice leaf disease identification method based on deep learning - Google Patents
Rice leaf disease identification method based on deep learning Download PDFInfo
- Publication number
- CN116863340A CN116863340A CN202311033402.7A CN202311033402A CN116863340A CN 116863340 A CN116863340 A CN 116863340A CN 202311033402 A CN202311033402 A CN 202311033402A CN 116863340 A CN116863340 A CN 116863340A
- Authority
- CN
- China
- Prior art keywords
- disease
- rice
- picture
- pictures
- paddy field
- 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
- 201000010099 disease Diseases 0.000 title claims abstract description 218
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 218
- 235000007164 Oryza sativa Nutrition 0.000 title claims abstract description 108
- 235000009566 rice Nutrition 0.000 title claims abstract description 108
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000013135 deep learning Methods 0.000 title claims abstract description 19
- 240000007594 Oryza sativa Species 0.000 title 1
- 241000209094 Oryza Species 0.000 claims abstract description 115
- 238000012549 training Methods 0.000 claims abstract description 28
- 238000012545 processing Methods 0.000 claims abstract description 27
- 239000000463 material Substances 0.000 claims abstract description 23
- 238000013528 artificial neural network Methods 0.000 claims abstract description 8
- 238000012216 screening Methods 0.000 claims description 7
- 230000005484 gravity Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 3
- 238000003384 imaging method Methods 0.000 claims description 3
- 239000002131 composite material Substances 0.000 claims 1
- 239000000523 sample Substances 0.000 description 35
- 238000002372 labelling Methods 0.000 description 8
- 241000196324 Embryophyta Species 0.000 description 6
- 238000001514 detection method Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 3
- 239000003086 colorant Substances 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000003902 lesion Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 241000894007 species Species 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Multimedia (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a rice leaf disease identification method based on deep learning, which belongs to the technical field of image processing, and comprises the following steps: acquiring a material sample, setting a corresponding training set based on the material sample, and establishing a corresponding disease identification model according to the training set and the deep neural network; establishing an operation platform based on a disease identification model; the operation platform is used for receiving the rice pictures uploaded by each client, analyzing the received rice pictures, determining corresponding disease types, and outputting corresponding disease reasons and corresponding measures according to all rice picture analysis results corresponding to the rice field; the user installs and applies the corresponding client; the user edits the corresponding paddy field information through the client, and the paddy field picture is collected after the paddy field information is compiled; and the collected rice pictures are combined with corresponding paddy field information to be sent to an operation platform, and the operation platform outputs corresponding identified disease types, disease reasons and countermeasures according to the received rice pictures.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to a rice leaf disease identification method based on deep learning.
Background
Currently, two main technical means are available for identifying leaf lesions: firstly, establishing an artificial neural network model to identify photos or images of single blade samples; the other is to use a support vector machine to establish feature vectors through the features of colors, textures, shapes and the like for recognition. However, the number of layers of the artificial neural network is small, and the feature extraction capability is insufficient; the support vector machine relies on specific features to classify, however, the specific features do not fully or well represent the lesion feature information, and therefore the classification accuracy is limited. Moreover, the two methods are used for identifying the plant by utilizing the collected limited blade samples indoors, and the research work is not converted into field application. On the other hand, training samples are fewer, the sample size is mostly expanded through some simple image processing modes such as rotation scaling, and the model generalization capability is poor.
Disclosure of Invention
In order to solve the problems of the scheme, the invention provides a rice leaf disease identification method based on deep learning, so as to solve the existing rice leaf disease identification problem.
The aim of the invention can be achieved by the following technical scheme:
the rice leaf disease identification method based on deep learning comprises the following steps:
step S1: acquiring a material sample, setting a corresponding training set based on the material sample, and establishing a corresponding disease identification model according to the training set and the deep neural network;
further, the method for acquiring the material sample comprises the following steps:
setting a target search formula, carrying out real-time search on an initial picture based on the target search formula, and carrying out de-duplication, check and screening on the searched initial picture to obtain a corresponding qualified picture; identifying qualified picture characteristics corresponding to the qualified picture, and carrying out real-time picture processing on the qualified picture based on the qualified picture characteristics and a preset target image processing mode to obtain a corresponding expanded picture; updating the characteristics of the combined picture in real time according to the obtained disease types and the picture quantity corresponding to the extended picture; and stopping searching when the characteristics of the qualified picture meet the training requirement, and integrating the existing qualified picture and the expanded picture into a material sample.
Further, the method for processing the combined picture in real time comprises the following steps:
identifying the number of qualified pictures and the number of expanded pictures in the qualified picture characteristics corresponding to each disease type in real time, calculating the proportion of the number of expanded pictures in each disease type, marking the proportion as real-time proportion, and when the real-time proportion is lower than a threshold value X1, performing corresponding target image processing on the qualified pictures corresponding to the disease type; when the real-time specific gravity is not lower than the threshold value X1, the corresponding target image processing is not carried out on the qualified picture corresponding to the disease type.
Step S2: establishing an operation platform based on a disease identification model; the operation platform is used for receiving the rice pictures uploaded by each client, analyzing the received rice pictures, determining corresponding disease types, and outputting corresponding disease reasons and corresponding measures according to all rice picture analysis results corresponding to the rice field;
further, the operation platform comprises a disease identification module and a comprehensive analysis module;
the disease identification module is used for identifying the diseases of the received rice pictures to obtain corresponding disease types;
the comprehensive analysis module is used for comprehensively analyzing the paddy field by combining the paddy picture information and the disease identification record corresponding to the disease identification module and outputting the disease reason of the paddy field and corresponding countermeasures.
Further, the working method of the disease recognition module comprises the following steps:
and acquiring the received rice pictures, inputting the rice pictures into a disease identification model for disease identification, and obtaining corresponding disease types.
Further, the working method of the comprehensive analysis module comprises the following steps:
identifying paddy field information corresponding to the paddy pictures, generating disease characteristics of the paddy pictures according to identification records of the disease identification modules on the paddy pictures, analyzing disease values corresponding to the disease characteristics, generating corresponding paddy field disease characteristics according to uploading sequences of the paddy pictures and the disease values, and summarizing the paddy field disease characteristics corresponding to the disease types into a paddy field characteristic set; analyzing the obtained paddy field characteristic set through a preset disease analysis model, and outputting corresponding disease reasons and countermeasures.
Further, the method for calculating the disease value comprises the following steps:
identifying disease types, disease numbers and disease areas in disease characteristics, and marking the disease areas as Mi, wherein i=1, 2, … …, n and n are the disease numbers; matching corresponding adjustment coefficients c according to the number of diseases and the types of the diseases; setting an area initial value A0; the corresponding disease value BT is calculated from the disease value formula bt=c×Σexp (Mi-A0).
Further, when the number of diseases is zero, the disease value bt=0.
Step S3: the user installs and applies the corresponding client;
step S4: the user edits the corresponding paddy field information through the client, and the paddy field picture is collected after the paddy field information is compiled;
further, the method for collecting rice pictures comprises the following steps:
shooting target rice to be subjected to disease identification through an imaging function of a client to obtain a corresponding first shooting picture;
identifying the rice outline in the first shot picture based on a preset rice outline identification model, and displaying the identified rice outline in the first shot picture to obtain a second shot picture;
the user determines a corresponding target rice outline in the second shot picture to obtain a third shot picture;
and processing the region corresponding to the non-target rice outline in the third shot picture to obtain the rice picture.
Step S5: and the collected rice pictures are combined with corresponding paddy field information to be sent to an operation platform, and the operation platform outputs corresponding identified disease types, disease reasons and countermeasures according to the received rice pictures.
Compared with the prior art, the invention has the beneficial effects that:
the identified disease condition can be specific to the corresponding target rice by marking the target rice corresponding to the single plant in the rice picture, so that the disease condition of the single plant can be truly represented without being influenced by other rice plants; meanwhile, interference factors can be reduced, and the corresponding disease identification accuracy is improved; and through carrying out single plant rice discernment, realize mutually supporting with the comprehensive analysis module in the operation platform, know the disease characteristic of single plant rice fast, and then realize the disease analysis of whole paddy field, confirm corresponding disease proportion, information such as constitution, output corresponding disease reason and countermeasure, assist peasant user to carry out the planting of rice, solve current environmental detection restriction, realize peasant user's simple operation can know the actual disease condition of paddy field.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings can be obtained according to these drawings without inventive effort to a person skilled in the art.
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a flowchart of the acquisition of a material sample according to the present invention;
fig. 3 is a functional block diagram of the comprehensive analysis module of the present invention.
Detailed Description
The technical solutions of the present invention will be clearly and completely described in connection with the embodiments, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1 to 3, deep learning is an image recognition means currently emerging, and can recognize a target without depending on specific features, and can improve recognition capability and efficiency of a model by means of adjusting parameters, optimizing a sample structure, and the like. The method is widely applied to the aspects of behavior detection, handwriting font recognition and the like at present, and has fresh application in agriculture, especially in the rice disease recognition direction. The invention provides a rice leaf disease identification method based on deep learning, which utilizes a current advanced deep learning network structure to acquire a large number of rice leaf disease picture samples for training by combining the method provided by the invention, and establishes a corresponding disease identification model to identify rice leaf diseases.
The specific process is as follows:
step S1: acquiring a material sample, setting a corresponding training set based on the acquired material sample, and establishing a corresponding disease identification model according to the set training set and the deep neural network;
the method comprises the steps of obtaining a material sample, namely various rice leaf disease pictures used for building a training set, wherein the obtaining method of the existing material sample is basically divided into two types, one type is that a plurality of sample pictures are obtained manually, and then the sample quantity is expanded by a plurality of simple image processing modes such as rotation scaling and the like to form a sample material, but the training samples obtained by the method are fewer, and the model generalization capability is poor; the other is to collect in the field by means of unmanned aerial vehicle and other tools, but the collection efficiency of this mode is very low, and a lot of time is required to find out the rice with the disease leaves to shoot, and not all rice leaves have disease conditions, so in order to collect a lot of training samples, the collection equipment such as unmanned aerial vehicle needs to be controlled manually to take a lot of time to collect, the efficiency is very low, and a lot of manpower resources are also required to be spent; therefore, according to the advantages and disadvantages of the two modes, the invention provides a method for acquiring a material sample, which realizes the effects of high efficiency, high speed and reduced manual participation of the material sample, and the specific method comprises the following steps:
setting a corresponding target retrieval formula according to the setting requirements of a material sample, wherein the setting requirements such as rice pictures, definition requirements, disease requirements and the like are specifically set according to actual detection requirements, searching pictures in the internet of things by using the set retrieval formula to obtain a large number of initial pictures, performing de-duplication and checking screening on the obtained initial pictures by using the existing image recognition technology, performing de-duplication to remove a large number of repeated pictures, checking the rest initial pictures according to the setting requirements of the material sample, removing the initial pictures which do not meet the setting requirements, realizing checking screening, checking screening by combining a manual mode, and directly removing unqualified initial pictures; marking the initial picture subjected to the duplication and verification screening as a qualified picture, identifying each disease type corresponding to the qualified picture and the number of pictures corresponding to each disease type, integrating each disease type and the number of corresponding pictures as qualified picture characteristics, performing corresponding target image processing on the qualified picture according to the qualified picture characteristics, and expanding the qualified picture; according to the initial picture searching and processing process, updating the qualified picture characteristics in real time, and when the qualified picture characteristics meet the number requirement of the training set, achieving the training requirement; stopping searching, and integrating the existing qualified picture and the expanded picture into a material sample.
The method comprises the steps of carrying out corresponding target image processing on a qualified picture according to the characteristics of the qualified picture, wherein the target image processing is to expand the image by utilizing the existing image processing modes such as rotary scaling and the like, namely a target image processing mode; processing according to the qualified picture characteristics, namely according to the number of pictures corresponding to each disease type in the qualified picture characteristics and the number proportion of the expanded pictures obtained through target image processing, setting a threshold value X1 of the upper limit of the number proportion of each type of expanded pictures, and when the threshold value X1 is not reached, performing corresponding target image processing, otherwise, not performing target image processing on the qualified pictures corresponding to the disease type; the threshold X1 is mainly set according to the acquisition difficulty degree and expansion limit of qualified pictures of various disease types; and with the progress of searching, de-duplication and checking screening, after the number of pictures of a certain disease type reaches the training requirement, the specific gravity of the pictures can be reduced by adopting a mode of eliminating corresponding expansion pictures one by one according to actual needs, or the pictures can be eliminated.
Corresponding disease identification models are established according to the set training set and the deep neural network, and corresponding disease identification model establishment can be realized according to the existing deep neural network establishment and training process; the specific setup and training procedures that follow are not described in detail in this invention.
The disease identification model is illustratively established by the following steps:
creating a sample data set, wherein the sample data set comprises a material sample and a manual labeling sample set, the material sample is an original picture set, the manual labeling sample set is a picture set obtained by performing format conversion on an original picture and manually labeling the disease of a leaf, and the ratio of the material sample to the manual labeling sample set in the sample data set is 2:1; binarizing the pictures in the manual labeling sample set, storing the pictures in a single-channel mode, and cutting the photo sample set and the manual labeling sample set according to a proportion to form a second photo sample set and a second manual labeling sample set; setting up a Linknet network model based on a Linknet network structure under a Pytorch deep learning framework, setting parameters of the Linknet network model, inputting a second photo sample set and a second manual labeling sample set into the Linknet network model, training the Linknet network model based on the Pytorch deep learning framework, storing a plurality of models in the training process, and selecting the model with the minimum error by using verification set data as a disease identification model.
Step S2: establishing an operation platform based on a disease identification model;
the operation platform is used for receiving the rice pictures uploaded by each client and analyzing the received rice pictures through the disease identification model to determine the corresponding disease types;
the operation platform mainly comprises a disease identification module and a comprehensive analysis module;
the disease recognition module is established based on a disease recognition model and is used for recognizing the received rice image for disease and obtaining the corresponding disease type.
The comprehensive analysis module is used for comprehensively analyzing the paddy field by combining the paddy picture information and the disease identification record corresponding to the disease identification module and outputting the reason of the paddy field disease and corresponding countermeasures; the specific working method comprises the following steps:
identifying paddy field information corresponding to the paddy field picture, setting information such as the name, the number and the like of the paddy field in advance before a user uses a client to collect the paddy field picture, and automatically endowing the collected and uploaded paddy field picture with the corresponding paddy field information when the user collects the picture so as to indicate which paddy field the paddy field picture belongs to; generating disease characteristics of the rice picture according to the identification record of the disease identification module to the rice picture, wherein the disease characteristics comprise disease types, the number of the diseases on the rice plant and the size of each disease, evaluating disease values corresponding to the rice plant according to the disease characteristics of the rice picture, and generating rice field disease characteristics of the disease types of the rice field according to the uploading sequence of the rice picture and the disease values, if the disease values corresponding to the sequence are a1, a2, a3,0 and a4, the rice field disease characteristics of the disease types are as follows: (a, a1, a2, a3,0, a 4), a representing a corresponding disease species; summarizing the rice field disease characteristics corresponding to various disease types into a rice field characteristic set; the method comprises the steps of establishing a corresponding disease analysis model based on a CNN network or a DNN network, establishing a corresponding training set by a manual mode for training, wherein the training set comprises a simulated rice field characteristic set and disease reasons and countermeasures which are correspondingly set, the disease reasons and the countermeasures are mutually related, setting the training set according to a large amount of existing historical disease data, setting the corresponding disease reasons according to the combination of different types of diseases, setting the corresponding countermeasures according to the combination of the disease reasons, establishing a corresponding disease reason and countermeasures database, and analyzing the obtained rice field characteristic set through the disease analysis model after successful training to obtain the corresponding disease reasons and countermeasures.
The disease value calculating method comprises the following steps:
identifying disease types, disease numbers and sizes of various diseases in disease characteristics, and marking corresponding diseases in the disease types as i, wherein i=1, 2, … …, n and n are the disease numbers; marking the disease size as Mi, i.e. area; according to the number of diseases and the adjustment coefficient corresponding to the disease types, according to the number of diseases possibly possessed by the disease types on a plant of rice, setting the corresponding adjustment coefficient c for each number in a discussion mode by an expert group to form an adjustment coefficient matching table corresponding to the number of diseases corresponding to each disease type, wherein the number of diseases is limited, the corresponding adjustment coefficient is set more quickly in a manual mode, and the adjustment coefficient gradually increases with the increase of the number of diseases; matching corresponding adjustment coefficients according to the corresponding disease quantity and disease types; setting an area initial value A0, wherein the area initial value is an average value calculated according to a large number of disease areas; the corresponding disease value is calculated according to the disease value formula bt=c×Σexp (Mi-A0), wherein when there is no disease number, i.e. no disease, the disease value bt=0.
Step S3: the method comprises the steps that a user downloads and installs a client on mobile equipment such as a mobile phone and the like, and carries out corresponding registration, wherein the client is used for collecting and uploading rice pictures and correspondingly receiving feedback information of an operation platform;
step S4: the user edits the corresponding paddy field information through the client, and the paddy field picture is collected after the paddy field information is compiled;
the method for collecting the rice pictures comprises the following steps:
shooting rice to be subjected to disease identification through an imaging function of a client to obtain a corresponding first shooting picture;
the method comprises the steps that a corresponding rice contour recognition model is established based on the existing image recognition technology and contour recognition technology, the rice contour recognition model is used for recognizing rice contours in pictures, the existing contour recognition model can be directly used for recognition, only the corresponding recognition accuracy is guaranteed, a corresponding verification set can be set for verifying the existing contour recognition model, and the contour recognition model with high accuracy is selected as the rice contour recognition model; identifying the rice outline in the first shot image through a rice outline identification model, and displaying the identified rice outline, such as measuring edge display and outline representation with different colors; marking a first shot picture for displaying the rice outline as a second shot picture, canceling the mark of a non-identified rice outline in the second shot picture by a user to obtain the outline of the rice corresponding to the need for disease identification, namely, clicking a certain outline once to cancel the outline display, canceling the outline, indicating canceling the outline, clicking the outline again to display the outline again, canceling and supplementing the outline, clicking the unidentified outline to supplement identification, finally obtaining the outline of the target rice, marking the second shot picture for displaying only the outline of the target rice as a third shot picture, processing the area corresponding to the outline of the non-target rice in the third shot picture, such as deleting, and the like, if not deleting, determining the corresponding target rice according to the outline of the target rice; and marking the processed third shot picture as a rice picture.
The identified disease condition can be specific to the corresponding target rice by marking the target rice corresponding to the single plant in the rice picture, so that the disease condition of the single plant can be truly represented without being influenced by other rice plants; meanwhile, interference factors can be reduced, and the corresponding disease identification accuracy is improved; and through carrying out single plant rice discernment, realize mutually supporting with the comprehensive analysis module in the operation platform, know the disease characteristic of single plant rice fast, and then realize the disease analysis of whole paddy field, confirm corresponding disease proportion, information such as constitution, output corresponding disease reason and countermeasure, assist peasant user to carry out the planting of rice, solve current environmental detection restriction, realize peasant user's simple operation can know the actual disease condition of paddy field.
Step S5: and the collected rice pictures are combined with corresponding paddy field information to be sent to an operation platform, and the operation platform outputs corresponding identified disease types, disease reasons and countermeasures according to the received rice pictures.
The above formulas are all formulas with dimensions removed and numerical values calculated, the formulas are formulas which are obtained by acquiring a large amount of data and performing software simulation to obtain the closest actual situation, and preset parameters and preset thresholds in the formulas are set by a person skilled in the art according to the actual situation or are obtained by simulating a large amount of data.
The above embodiments are only for illustrating the technical method of the present invention and not for limiting the same, and it should be understood by those skilled in the art that the technical method of the present invention may be modified or substituted without departing from the spirit and scope of the technical method of the present invention.
Claims (8)
1. The rice leaf disease identification method based on deep learning is characterized by comprising the following steps:
acquiring a material sample, setting a corresponding training set based on the material sample, and establishing a corresponding disease identification model according to the training set and the deep neural network;
establishing an operation platform based on a disease identification model; the operation platform is used for receiving the rice pictures uploaded by each client, analyzing the received rice pictures, determining corresponding disease types, and outputting corresponding disease reasons and corresponding measures according to all rice picture analysis results corresponding to the rice field;
the user installs and applies the corresponding client;
the user edits the corresponding paddy field information through the client, and the paddy field picture is collected after the paddy field information is compiled;
the collected rice pictures are combined with corresponding rice field information to be sent to an operation platform, and the operation platform outputs corresponding identified disease types, disease reasons and countermeasures according to the received rice pictures;
the method for acquiring the material sample comprises the following steps:
setting a target search formula, carrying out real-time search on an initial picture based on the target search formula, and carrying out de-duplication, check and screening on the searched initial picture to obtain a corresponding qualified picture;
identifying qualified picture characteristics corresponding to the qualified picture, and carrying out real-time picture processing on the qualified picture based on the qualified picture characteristics and a preset target image processing mode to obtain a corresponding expanded picture;
updating the characteristics of the combined picture in real time according to the obtained disease types and the picture quantity corresponding to the extended picture; and stopping searching when the characteristics of the qualified picture meet the training requirement, and integrating the existing qualified picture and the expanded picture into a material sample.
2. The method for identifying rice leaf diseases based on deep learning according to claim 1, wherein the method for performing real-time image processing on the composite image comprises the steps of:
identifying the number of qualified pictures and the number of expanded pictures in the qualified picture characteristics corresponding to each disease type in real time, calculating the proportion of the number of expanded pictures in each disease type, and marking the proportion as real-time proportion;
when the real-time specific gravity is lower than a threshold value X1, performing corresponding target image processing on qualified pictures corresponding to disease types;
when the real-time specific gravity is not lower than the threshold value X1, the corresponding target image processing is not carried out on the qualified picture corresponding to the disease type.
3. The deep learning-based rice leaf disease identification method according to claim 1, wherein the operation platform comprises a disease identification module and a comprehensive analysis module;
the disease identification module is used for identifying the diseases of the received rice pictures to obtain corresponding disease types;
the comprehensive analysis module is used for comprehensively analyzing the paddy field by combining the paddy picture information and the disease identification record corresponding to the disease identification module and outputting the disease reason of the paddy field and corresponding countermeasures.
4. A method for identifying rice leaf diseases based on deep learning according to claim 3, wherein the working method of the disease identification module comprises:
and acquiring the received rice pictures, inputting the rice pictures into a disease identification model for disease identification, and obtaining corresponding disease types.
5. The method for identifying rice leaf diseases based on deep learning according to claim 3, wherein the working method of the comprehensive analysis module comprises the following steps:
identifying paddy field information corresponding to the paddy pictures, and generating disease characteristics of the paddy pictures according to identification records of the disease identification modules on the paddy pictures;
analyzing disease values corresponding to the disease features, generating corresponding paddy field disease features according to the uploading sequence of the paddy pictures and the disease values, and summarizing the paddy field disease features corresponding to the disease types into a paddy field feature set;
analyzing the obtained paddy field characteristic set through a preset disease analysis model, and outputting corresponding disease reasons and countermeasures.
6. The method for identifying rice leaf diseases based on deep learning according to claim 5, wherein the calculating method of the disease value comprises:
identifying disease types, disease numbers and disease areas in disease characteristics, and marking the disease areas as Mi, wherein i=1, 2, … …, n and n are the disease numbers; matching corresponding adjustment coefficients c according to the number of diseases and the types of the diseases; setting an area initial value A0; the corresponding disease value BT is calculated from the disease value formula bt=c×Σexp (Mi-A0).
7. The method for identifying rice leaf diseases based on deep learning according to claim 6, wherein the disease value bt=0 when the number of diseases is zero.
8. The method for identifying rice leaf diseases based on deep learning according to claim 1, wherein the method for collecting rice pictures comprises the following steps:
shooting target rice to be subjected to disease identification through an imaging function of a client to obtain a corresponding first shooting picture;
identifying the rice outline in the first shot picture based on a preset rice outline identification model, and displaying the identified rice outline in the first shot picture to obtain a second shot picture;
the user determines a corresponding target rice outline in the second shot picture to obtain a third shot picture;
and processing the region corresponding to the non-target rice outline in the third shot picture to obtain the rice picture.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311033402.7A CN116863340A (en) | 2023-08-16 | 2023-08-16 | Rice leaf disease identification method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311033402.7A CN116863340A (en) | 2023-08-16 | 2023-08-16 | Rice leaf disease identification method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116863340A true CN116863340A (en) | 2023-10-10 |
Family
ID=88232481
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311033402.7A Pending CN116863340A (en) | 2023-08-16 | 2023-08-16 | Rice leaf disease identification method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116863340A (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109165623A (en) * | 2018-09-07 | 2019-01-08 | 北京麦飞科技有限公司 | Rice scab detection method and system based on deep learning |
| CN109472252A (en) * | 2018-12-28 | 2019-03-15 | 华南农业大学 | A kind of field crops insect pest automatic identification and job management system |
| CN109886155A (en) * | 2019-01-30 | 2019-06-14 | 华南理工大学 | Man power single stem rice detection localization method, system, equipment and medium based on deep learning |
| CN111814622A (en) * | 2020-06-29 | 2020-10-23 | 华南农业大学 | Crop pest type identification method, system, equipment and medium |
| AU2020103613A4 (en) * | 2020-11-23 | 2021-02-04 | Agricultural Information and Rural Economic Research Institute of Sichuan Academy of Agricultural Sciences | Cnn and transfer learning based disease intelligent identification method and system |
| CN112507770A (en) * | 2020-08-13 | 2021-03-16 | 华南农业大学 | Rice disease and insect pest identification method and system |
| CN115019044A (en) * | 2022-06-16 | 2022-09-06 | 中山大学 | Individual plant segmentation method and device, terminal device and readable storage medium |
| CN116310541A (en) * | 2023-03-09 | 2023-06-23 | 浙江托普云农科技股份有限公司 | Insect classification method and system based on convolutional network multidimensional learning |
| CN116342919A (en) * | 2022-09-29 | 2023-06-27 | 安徽省农业科学院 | Rice disease identification method based on attention and gating mechanism |
-
2023
- 2023-08-16 CN CN202311033402.7A patent/CN116863340A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109165623A (en) * | 2018-09-07 | 2019-01-08 | 北京麦飞科技有限公司 | Rice scab detection method and system based on deep learning |
| CN109472252A (en) * | 2018-12-28 | 2019-03-15 | 华南农业大学 | A kind of field crops insect pest automatic identification and job management system |
| CN109886155A (en) * | 2019-01-30 | 2019-06-14 | 华南理工大学 | Man power single stem rice detection localization method, system, equipment and medium based on deep learning |
| CN111814622A (en) * | 2020-06-29 | 2020-10-23 | 华南农业大学 | Crop pest type identification method, system, equipment and medium |
| CN112507770A (en) * | 2020-08-13 | 2021-03-16 | 华南农业大学 | Rice disease and insect pest identification method and system |
| AU2020103613A4 (en) * | 2020-11-23 | 2021-02-04 | Agricultural Information and Rural Economic Research Institute of Sichuan Academy of Agricultural Sciences | Cnn and transfer learning based disease intelligent identification method and system |
| CN115019044A (en) * | 2022-06-16 | 2022-09-06 | 中山大学 | Individual plant segmentation method and device, terminal device and readable storage medium |
| CN116342919A (en) * | 2022-09-29 | 2023-06-27 | 安徽省农业科学院 | Rice disease identification method based on attention and gating mechanism |
| CN116310541A (en) * | 2023-03-09 | 2023-06-23 | 浙江托普云农科技股份有限公司 | Insect classification method and system based on convolutional network multidimensional learning |
Non-Patent Citations (1)
| Title |
|---|
| 姜敏;沈一鸣;张敬尧;饶元;董伟;: "基于深度学习的水稻病虫害诊断方法研究", 洛阳理工学院学报(自然科学版), no. 04 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107665492B (en) | Colorectal panoramic digital pathological image tissue segmentation method based on depth network | |
| CN109583483B (en) | Target detection method and system based on convolutional neural network | |
| CN104700099B (en) | The method and apparatus for recognizing traffic sign | |
| CN109509187B (en) | Efficient inspection algorithm for small defects in large-resolution cloth images | |
| CN103324937B (en) | The method and apparatus of label target | |
| CN109635875A (en) | A kind of end-to-end network interface detection method based on deep learning | |
| CN110929760A (en) | Garbage classification software based on computer vision | |
| CN109284779A (en) | Object detecting method based on the full convolutional network of depth | |
| CN112200211A (en) | Small sample fish identification method and system based on residual error network and transfer learning | |
| CN113850799B (en) | YOLOv 5-based trace DNA extraction workstation workpiece detection method | |
| CN111652835A (en) | Method for detecting insulator loss of power transmission line based on deep learning and clustering | |
| CN111382766A (en) | Equipment fault detection method based on fast R-CNN | |
| CN114066848A (en) | FPCA appearance defect visual inspection system | |
| CN114140665A (en) | Dense small target detection method based on improved YOLOv5 | |
| CN112560748A (en) | Crop shape analysis subsystem and method | |
| CN114596509A (en) | Machine vision-based rice seedling leaf age period identification method | |
| CN110728269A (en) | High-speed rail contact net support pole number plate identification method | |
| CN112966698A (en) | Freshwater fish image real-time identification method based on lightweight convolutional network | |
| CN113793385A (en) | Method and device for positioning fish head and fish tail | |
| CN112560749A (en) | Crop analysis system and analysis method | |
| CN112581459A (en) | Crop classification system and method | |
| CN110472092B (en) | Geographical positioning method and system of street view picture | |
| CN112132137A (en) | FCN-SPP-Focal Net-based method for identifying correct direction of abstract picture image | |
| CN117011759A (en) | Method and system for analyzing multi-element geological information of surrounding rock of tunnel face by drilling and blasting method | |
| CN116645351A (en) | Online defect detection method and system for complex scene |
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 |