CN116612307A - Solanaceae disease grade identification method based on transfer learning - Google Patents
Solanaceae disease grade identification method based on transfer learning Download PDFInfo
- Publication number
- CN116612307A CN116612307A CN202310131684.8A CN202310131684A CN116612307A CN 116612307 A CN116612307 A CN 116612307A CN 202310131684 A CN202310131684 A CN 202310131684A CN 116612307 A CN116612307 A CN 116612307A
- Authority
- CN
- China
- Prior art keywords
- disease
- solanaceae
- model
- image
- transfer learning
- 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 406
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 406
- 241000208292 Solanaceae Species 0.000 title claims abstract description 187
- 238000013526 transfer learning Methods 0.000 title claims abstract description 117
- 238000000034 method Methods 0.000 title claims abstract description 73
- 238000012549 training Methods 0.000 claims abstract description 113
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 35
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 238000013508 migration Methods 0.000 claims description 18
- 230000005012 migration Effects 0.000 claims description 18
- 238000011176 pooling Methods 0.000 claims description 11
- 230000011218 segmentation Effects 0.000 claims description 7
- 230000008014 freezing Effects 0.000 claims description 4
- 238000007710 freezing Methods 0.000 claims description 4
- 230000003313 weakening effect Effects 0.000 claims description 4
- 208000020358 Learning disease Diseases 0.000 abstract description 4
- 230000035945 sensitivity Effects 0.000 abstract description 3
- 235000007688 Lycopersicon esculentum Nutrition 0.000 description 22
- 240000003768 Solanum lycopersicum Species 0.000 description 22
- 238000010586 diagram Methods 0.000 description 17
- 241000196324 Embryophyta Species 0.000 description 16
- 230000008569 process Effects 0.000 description 11
- 238000003709 image segmentation Methods 0.000 description 9
- 238000000605 extraction Methods 0.000 description 8
- 230000007246 mechanism Effects 0.000 description 8
- 238000005457 optimization Methods 0.000 description 8
- 238000012795 verification Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 238000012545 processing Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 230000004927 fusion Effects 0.000 description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 3
- 230000001580 bacterial effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000007423 decrease Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000003902 lesion Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 241000207199 Citrus Species 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- 244000061458 Solanum melongena Species 0.000 description 1
- 235000002597 Solanum melongena Nutrition 0.000 description 1
- 244000061456 Solanum tuberosum Species 0.000 description 1
- 235000002595 Solanum tuberosum Nutrition 0.000 description 1
- 241001454293 Tetranychus urticae Species 0.000 description 1
- 241000607479 Yersinia pestis Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 235000020971 citrus fruits Nutrition 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- 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
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- 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
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- 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
-
- 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
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30004—Biomedical 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention discloses a method for identifying the grade of a solanaceae disease based on transfer learning, which comprises the steps of preprocessing an acquired solanaceae image based on a K-means algorithm, identifying the disease characteristics of the processed image based on a preset transfer learning disease grade identification model, and determining the solanaceae disease image according to the characteristic identification result, wherein the preset transfer learning disease grade identification model is a model obtained by transferring model parameters and parameter weights of an improved MSS-ResNet101 model to a pre-training model to obtain a transferred model and training a classification layer in the transferred model. And counting the area of the disease spots in the solanaceae disease image, and determining the solanaceae disease degree according to the counting result. The invention solves the problems of low precision or overfitting caused by low sensitivity of the classification recognition model of the disease degree of the solanaceae plant to the fine granularity characteristic, and improves the disease recognition rate.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method for identifying the grade of a solanaceae disease based on transfer learning.
Background
Along with the development of agricultural economy, more and more plants are planted and cultivated mechanically, but various factors influence the growth of the plants in the actual planting process, wherein diseases are a great factor influencing the growth of the plants, a great number of crops are damaged by different plant diseases each year, and a great deal of losses are caused, so that in order to ensure the yield of the crops and the service life of the plants, the plant diseases need to be accurately detected and identified;
although a model calculation method for identifying the grade of the solanaceae disease exists in the prior art, the existing method has insufficient fine granularity for processing pictures, has limitation of insufficient quantity of training images, and causes the problem of low precision or overfitting caused by low sensitivity of a network model to fine granularity characteristics, so that disease identification efficiency is affected, and data errors easily occur to cause poor accuracy.
The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present invention and is not intended to represent an admission that the foregoing is prior art.
Disclosure of Invention
The invention mainly aims to provide a solanaceae disease grade identification method based on transfer learning, and aims to solve the technical problems of low precision or over-fitting caused by low sensitivity of a network model to fine granularity characteristics due to insufficient fine granularity of the model for processing pictures and insufficient quantity of training images in the prior art.
In order to achieve the above object, the present invention provides a method for identifying a grade of a disease of the solanaceae based on transfer learning, the method for identifying a grade of a disease of the solanaceae based on transfer learning comprising the steps of:
preprocessing the acquired solanaceae image based on a K-means algorithm to obtain a processed image;
performing disease feature recognition on the processed image based on a disease level recognition model of preset transfer learning, and determining a solanaceae disease image according to a feature recognition result, wherein the disease level recognition model of the preset transfer learning is a model obtained by transferring model parameters and parameter weights of an improved MSS-ResNet101 model to a pre-training model to obtain a transferred model and training a classification layer in the transferred model, and the classification layer is a new classification layer formed by adding a full-connection layer in front of a softmax classification layer;
and counting the area of the disease spots in the solanaceae disease image, and determining the solanaceae disease degree according to the counting result.
Optionally, the step of preprocessing the acquired image of the solanaceae based on the K-means algorithm to obtain a processed image includes:
performing pixel point segmentation on the obtained Solanaceae image based on a K-means algorithm to obtain a pixel point data set;
And dividing the Solanaceae image according to the pixel similarity corresponding to the pixel point data set, and extracting an image containing key disease features from the divided image.
Optionally, the disease grade recognition model based on preset transfer learning performs disease feature recognition on the processed image, and before the step of determining the solanaceae disease image according to the feature recognition result, the method further includes:
adopting an ImageNet image data set as a source data set of migration training;
training the improved MSS-ResNet101 model based on the ImageNet source data set to obtain a pre-training model;
and constructing a disease grade identification model for preset transfer learning based on the parameters and weights corresponding to the pre-training model and the pre-training model.
Optionally, the step of constructing a disease level recognition model for preset transfer learning based on the parameters and weights corresponding to the pre-training model and the pre-training model includes:
initializing the pre-training model by taking the parameters and the weights of the pre-training model as the initialization parameters of the network, and freezing the structure before the global average pooling layer to obtain the pre-training model to be improved;
Adding a full-connection layer in front of the softmax classification layer of the pre-training model to be improved to form a new classification layer, and modifying network parameters corresponding to the softmax classification layer to be suitable for a solanaceae disease degree classification task to obtain a target pre-training model;
training the new classification layer in the target pre-training model to obtain a trained model;
and fine-tuning network parameters corresponding to the trained model to obtain a disease grade identification model for preset transfer learning.
Optionally, the step of identifying the disease characteristics of the processed image based on the disease level identification model of preset transfer learning and determining the solanaceae disease image according to the characteristic identification result includes:
weakening the background in the processed image based on a disease grade identification model of preset transfer learning to obtain a target image with outstanding disease characteristics;
extracting disease features with different scales from the target image through a multi-scale feature structure in a disease grade identification model of preset transfer learning;
and identifying the disease characteristics based on the disease grade identification model of the preset transfer learning, and determining a Solanaceae disease image according to the characteristic identification result.
Optionally, the step of counting the area of the disease spots in the image of the disease of the solanaceae and determining the degree of the disease of the solanaceae according to the statistical result includes:
performing disease identification on the solanaceae disease image based on the disease grade identification model of the preset transfer learning to obtain disease types;
classifying the solanaceae disease images according to the disease categories to obtain solanaceae disease image sets of all disease types;
and counting the disease spot areas in the image collection of the solanaceae diseases of each disease type, and determining the degree of the solanaceae diseases according to the counting result.
Optionally, the step of counting the area of the disease spots in the image set of the solanaceae disease of each disease type and determining the degree of the solanaceae disease according to the statistical result further includes:
acquiring the total pixel quantity of a disease spot area region;
determining the disease degree of a single blade according to the total pixel quantity and the total pixel quantity of the whole blade area;
and counting the single-leaf disease degree in the solanaceae disease image set of each disease type, and determining the solanaceae disease degree grade according to the disease degree counting result.
The method comprises the steps of preprocessing an acquired solanaceae image based on a K-means algorithm to obtain a processed image; disease characteristic recognition is carried out on the processed image based on a disease level recognition model of preset transfer learning, a solanaceae disease image is determined according to a characteristic recognition result, the disease level recognition model of the preset transfer learning is a model obtained by transferring model parameters and parameter weights of an improved MSS-ResNet101 model to a pre-training model to obtain a transferred model and training a classification layer in the transferred model, and the classification layer is a new classification layer formed by adding a full-connection layer in front of a softmax classification layer. And counting the area of the disease spots in the solanaceae disease image, and determining the solanaceae disease degree according to the counting result. According to the invention, the K-means algorithm is used for preprocessing the obtained solanaceae image, disease feature recognition is carried out on the solanaceae image according to the migrated model, and the solanaceae disease image is determined according to the feature recognition result, so that the disease degree is further determined.
Drawings
Fig. 1 is a schematic structural diagram of a solanaceae disease level recognition device based on transfer learning in a hardware operation environment according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of a first embodiment of a method for identifying the grade of the solanaceae disease based on transfer learning;
FIG. 3 is a schematic diagram of K-means elbow method of a first embodiment of the method for identifying the grade of the Solanaceae disease based on transfer learning;
fig. 4 is a schematic image segmentation diagram of a first embodiment of a method for identifying a solanaceae disease level based on transfer learning according to the present invention;
fig. 5 is a schematic diagram of a transfer learning flow of a first embodiment of a method for identifying a grade of a solanaceae disease based on transfer learning in the present invention;
FIG. 6 is a schematic flow chart of a second embodiment of a method for identifying the grade of the solanaceae disease based on transfer learning;
fig. 7 is a schematic diagram of migration training of a second embodiment of a method for identifying a grade of a disease of the solanaceae based on migration learning in the invention;
FIG. 8 is a schematic diagram showing comparison results of recognition rates of verification sets before and after transfer learning according to a second embodiment of a method for recognizing Solanaceae disease level based on transfer learning;
FIG. 9 is a schematic diagram showing the influence of different optimization algorithms of a second embodiment of a method for identifying the grade of the solanaceae disease based on transfer learning on the change of the loss value;
Fig. 10 is a schematic diagram of a confusion matrix for classifying and identifying the degrees of tomato diseases according to a second embodiment of the method for identifying the grade of solanaceae diseases based on transfer learning;
fig. 11 is a block diagram showing the construction of a first embodiment of the device for identifying the grade of the disease of the solanaceae based on the transfer learning.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Referring to fig. 1, fig. 1 is a schematic structural diagram of a solanaceae disease level recognition device based on transfer learning in a hardware operation environment according to an embodiment of the present invention.
As shown in fig. 1, the apparatus for identifying a grade of a disease of the solanaceae based on transfer learning may include: a processor 1001, such as a central processing unit (Central Processing Unit, CPU), a communication bus 1002, a user interface 1003, a network interface 1004, a memory 1005. Wherein the communication bus 1002 is used to enable connected communication between these components. The user interface 1003 may include a Display (Display), and the optional user interface 1003 may also include a standard wired interface, a wireless interface, and the wired interface for the user interface 1003 may be a USB interface in the present invention. The network interface 1004 may optionally include a standard wired interface, a Wireless interface (e.g., a Wireless-Fidelity (Wi-Fi) interface). The Memory 1005 may be a high-speed random access Memory (Random Access Memory, RAM) or a stable Memory (NVM), such as a disk Memory. The memory 1005 may also optionally be a storage device separate from the processor 1001 described above.
It will be appreciated by those skilled in the art that the structure shown in fig. 1 does not constitute a limitation of the solanaceae disease level recognition device based on transfer learning, and may include more or fewer components than shown, or certain components may be combined, or a different arrangement of components.
As shown in fig. 1, an operating system, a network communication module, a user interface module, and a solanaceae disease level recognition program based on transfer learning may be included in a memory 1005, which is considered to be a computer storage medium.
In the solanaceae disease grade identification device based on transfer learning shown in fig. 1, a network interface 1004 is mainly used for connecting a background server, and performing data communication with the background server; the user interface 1003 is mainly used for connecting user equipment; the solanaceae disease grade recognition device based on transfer learning calls a solanaceae disease grade recognition program based on transfer learning stored in a memory 1005 through a processor 1001, and executes the solanaceae disease grade recognition method based on transfer learning provided by the embodiment of the invention.
Based on the hardware structure, the embodiment of the solanaceae disease grade identification method based on transfer learning is provided.
Referring to fig. 2, fig. 2 is a flowchart of a first embodiment of a method for identifying a disease grade of the solanaceae based on transfer learning according to the present invention, and the first embodiment of the method for identifying a disease grade of the solanaceae based on transfer learning is provided.
In this embodiment, the method for identifying the grade of the solanaceae disease based on the transfer learning includes the following steps:
step S10: preprocessing the acquired solanaceae image based on a K-means algorithm to obtain a processed image.
Note that, the execution subject of the present embodiment may be a device for identifying a disease level of the solanaceae, such as: computers, notebooks, computers, tablets and the like, and can also be other solanaceae disease grade identification equipment capable of realizing the same or similar functions. This embodiment is not limited thereto. This embodiment and the following embodiments will be described herein by taking the above-described computer as an example.
It can be understood that the solanaceae image may refer to an solanaceae image which needs to be subjected to disease recognition, the image includes leaf images with disease spots corresponding to the solanaceae plants to be recognized and also includes normal leaf images, and in order to accurately recognize the leaf images with disease spots, the disease image needs to be screened from the solanaceae image before the disease recognition is performed, so that the plant image needs to be preprocessed to accurately recognize the disease image.
It will be appreciated that the plants of the Solanaceae family to be identified include, but are not limited to, tomato, eggplant, pepper, potato etc., which are widely cultivated crops of high economic value, but also routinely cultivated plants. The present embodiment is not particularly limited thereto.
Further, the step S10 further includes: performing pixel point segmentation on the obtained Solanaceae image based on a K-means algorithm to obtain a pixel point data set; and dividing the Solanaceae image according to the pixel similarity corresponding to the pixel point data set, and extracting an image containing key disease features from the divided image.
It should be noted that the gist of the K-means algorithm is to randomly determine K data points as centroids for data to be classified, measure similarity between the centroids and other data sample points according to a certain method, update the centroids by iteration continuously, and finally divide the data samples into K different clusters when the centroids are not changed any more. The K values in the K-means algorithm are generally chosen based on the elbow method, which is schematically shown in FIG. 3 for example in the K-means elbow method. The K value on the x-axis in FIG. 3 represents the number of classifications of data, the y-axisSSEThe value (sum of squares of errors) is a core index for measuring the clustering effect in the elbow method, and the specific form is as follows:
；
wherein, in the formulaRepresent the firstiThe number of clusters is one,jrepresentation->Sample points of->Representation->Average of all samples in the sample. SSE represents the sum of squares of the cluster errors of all sample data, the value of SSE decreases with the increase of the K value, and the descending amplitude also decreases with the increase of the K value KThe value increases and decreases, and when the descending amplitude starts to be gentle, the corresponding K value is the optimal cluster number.
It can be understood that the K-means algorithm is utilized to perform segmentation pretreatment on the image, namely, pixels of the image are used as a data set, and then the image is divided according to the similarity among the pixels, so that the purpose of image segmentation is achieved. The image segmentation aims at extracting needed information from the target image, and the original image is preprocessed by the image segmentation method, so that key features in the disease image can be enhanced, noise interference in the disease image is weakened, and the recognition accuracy of the classification model is effectively improved. The image obtained after the image segmentation pretreatment is the basis for accurately positioning the characteristics and measuring the parameters, and the segmented image is fully utilized to be helpful for the realization of tasks such as image analysis, image classification, target detection and the like. For the unsupervised image clustering problem, the image is often divided into areas with different feature meanings by utilizing the feature similarity of the adjacent areas of the image, so that the image segmentation problem is converted into the unsupervised clustering problem, and the key features of the image are enhanced.
It is understood that in order to improve the identification effect of fine-grained image classification, the method uses an unsupervised K-means segmentation algorithm to preprocess images, and performs primary extraction and filtration on disease features and interference factors in tomato diseases. And the image after image segmentation is used as a new input image to extract key disease features, so that the distinguishing capability of the CNN model on the disease features is further enhanced, and the recognition accuracy of the model is improved. For example: taking bacterial spot disease of tomato leaves as an example, a K-means algorithm is utilized to carry out image segmentation on disease images, as shown in an image segmentation schematic diagram shown in fig. 4, the key characteristics in the disease images are enhanced, and therefore fine granularity of image processing identification is improved.
Step S20: disease characteristic recognition is carried out on the processed image based on a disease level recognition model of preset transfer learning, a solanaceae disease image is determined according to a characteristic recognition result, the disease level recognition model of the preset transfer learning is a model obtained by transferring model parameters and parameter weights of an improved MSS-ResNet101 model to a pre-training model to obtain a transferred model and training a classification layer in the transferred model, and the classification layer is a new classification layer formed by adding a full-connection layer in front of a softmax classification layer.
It should be noted that, the model-based migration learning method migrates the model weights and parameters pre-trained on a large source data set to a target data set with small data volume but similar to the source data set, and a migration learning flow chart is shown in fig. 5. Model migration includes two ways: one is to use a pre-training network as a feature extractor, load all parameters of the pre-training model except the top layer during migration, modify the classification category number of the top layer model classifier to be suitable for new tasks, freeze the feature extraction part of the model, and train the target data domain by using the top layer classifier to obtain a final target task model; the other is to fine tune the model, and the weight or parameter is migrated by continuously updating the pre-training model by using the target task data so as to be better suitable for new tasks.
It can be understood that the disease level recognition model of the preset transfer learning may be a preset model for disease level recognition obtained through model transfer learning, where the preset transfer learning disease level recognition model is a model obtained by transferring model parameters and parameter weights of an improved MSS-res net101 model to a pre-training model to obtain a transferred model, and training a classification layer in the transferred model, where the classification layer refers to a new classification layer formed by adding a full-connection layer before a softmax classification layer.
It should be understood that the modified MSS-res net101 model is a model obtained by modification based on the res net101 model, and the res net101 model is composed of four BottleNeck, each block is respectively composed of 4 residual modules, and at the front end and the back end of the network, each block is respectively composed of 1 convolution layer of 7x7, a maxpool layer and an average pooling layer. The improved MSS-ResNet101 model comprises a multi-scale structure acceptance module replacing 7X7 convolution in the original ResNet101 model and a residual module integrating a convolution kernel attention mechanism, wherein the multi-scale structure acceptance module is connected with the residual module, and the residual module is connected with a global average pooling layer; the multi-scale structure acceptance module is used for extracting characteristics of an input image and inputting an obtained multi-scale structure characteristic diagram to the residual error module; the residual error module is used for carrying out feature fusion on the multi-scale structure feature map according to the convolution kernel attention mechanism, and inputting the fused feature information to a global average pooling layer and a Softmax layer for classification and identification. The convolutional kernel attention mechanism SKNet is introduced into the residual structure of the ResNet101, so that the SKNet can effectively help the model to capture semantic information useful for the recognition task, inhibit the influence of interference factors such as noise and the like, highlight key areas and enhance the expression capability of the model. The convolution kernel receptive field of a single scale is fixed, and the extracted characteristic information is limited. Therefore, an acceptance multi-scale structure acceptance module is embedded on the basis of RseNet101 to perform feature extraction, so that the richness of a feature channel is enhanced, a better recognition effect is obtained, and the recognition accuracy is improved. The improved MSS-ResNet101 model is based on a ResNet101 network model, combines a multiscale structure acceptance module and a convolution kernel attention mechanism SKNet, obtains an improved multiscale network model (Multi-Scale-SK-ResNet 101, hereinafter called MSS-ResNet 101) fused with the convolution kernel attention mechanism, replaces 7X7 convolution in the original ResNet101 by the acceptance module, fuses the convolution kernel attention mechanism in each residual module, and finally reserves a global average pooling layer and a Softmax layer.
In the specific implementation, the processed plant image is subjected to feature extraction through a multiscale structure acceptance module in a disease level recognition model of preset transfer learning to obtain a multiscale structure feature image, the multiscale structure feature image is subjected to feature fusion according to a convolution kernel attention mechanism in a residual error module, the fused feature information is input into a new classification layer to perform disease feature recognition, and a solanaceae disease image is determined according to a feature recognition result.
Further, before the step S20, the method further includes: adopting an ImageNet image data set as a source data set of migration training; training the improved MSS-ResNet101 model based on the ImageNet source data set to obtain a pre-training model; and constructing a disease grade identification model for preset transfer learning based on the parameters and weights corresponding to the pre-training model and the pre-training model.
In the migration training process, the source data set and the target data set need to have certain similarity, so the scheme adopts the image net image data set as the source data set of the migration training, the data set totally comprises 120 ten thousand images, 1000 categories, and is a high-quality large data set, and the image net image data set has SIFT features (scale non-conversion features), so that good recognition effect can be ensured even after the images are subjected to operations such as scaling, rotation, brightness adjustment and the like. The data set is used for pre-training, so that the model can obtain better recognition accuracy in the actual training process, and the trained model can have stronger generalization capability.
It can be appreciated that training the improved MSS-res net101 model based on the ImageNet source data set obtains a pre-training model, which may be derived by training the ImageNet source data with the improved MSS-res net101 model and ending the training after convergence.
It is to be understood that the disease level recognition model of the preset transfer learning is constructed based on the parameters and weights corresponding to the pre-training model and the pre-training model.
Step S30: and counting the area of the disease spots in the solanaceae disease image, and determining the solanaceae disease degree according to the counting result.
The disease degree of the solanaceae plant is determined by the disease area, which may be an area determined according to the disease pixels in the disease image.
It can be understood that the disease spots of different disease types have different characteristics, so that the disease degree of the solanaceae plants of each disease type can be accurately counted by counting the disease spot areas of different disease types.
Further, the step S30 further includes: performing disease identification on the solanaceae disease image based on the disease grade identification model of the preset transfer learning to obtain disease types; classifying the solanaceae disease images according to the disease categories to obtain solanaceae disease image sets of all disease types; and counting the disease spot areas in the image collection of the solanaceae diseases of each disease type, and determining the degree of the solanaceae diseases according to the counting result.
It should be noted that, the disease types include disease types corresponding to solanaceae plants, different disease types corresponding to different solanaceae plants and different expression characteristics, and the disease types are generally determined by disease spot characteristics shown by leaves, for example, if the plants are tomatoes, the disease types corresponding to the tomatoes are determined by identifying disease spot characteristics of the tomato leaves, and the disease types may include bacterial spot, early blight, late blight, leaf mold, spot blight, two-spotted spider mite disease, round spot disease, mosaic disease, yellow leaf curl disease and the like.
The disease type corresponding to the disease features contained in the disease images of the solanaceae can be determined by comparing the disease features contained in the disease images of the solanaceae with the disease features contained in the disease images of the solanaceae in a disease feature library, so that the disease type corresponding to plants can be determined, the later-stage accurate pest removal operation can be ensured, the yield is improved, and the loss of the plants due to the diseases is reduced.
It can be understood that disease categories corresponding to disease images are determined by identifying disease spots in the solanaceae disease images containing disease features, and the solanaceae disease images are classified according to the identified disease categories to obtain solanaceae disease image sets of all disease types, so that statistics can be conveniently carried out according to the disease degree of all disease types in the later period.
In the specific implementation, disease class identification is carried out on the solanaceae disease image based on the disease class identification model of the preset transfer learning, so as to obtain disease class; classifying the solanaceae disease images according to the disease categories to obtain solanaceae disease image sets of all disease types; and counting the disease spot areas in the image collection of the solanaceae diseases of each disease type, and determining the degree of the solanaceae diseases according to the counting result.
Further, the step of counting the disease spot areas in the image set of the solanaceae disease of each disease type and determining the degree of the solanaceae disease according to the statistical result further comprises the following steps: acquiring the total pixel quantity of a disease spot area region; determining the disease degree of a single blade according to the total pixel quantity and the total pixel quantity of the whole blade area; and counting the single-leaf disease degree in the solanaceae disease image set of each disease type, and determining the solanaceae disease degree grade according to the disease degree counting result.
It should be noted that, calculating the proportion of the total pixel amount of the lesion area to the whole blade area can obtain the single blade lesion degree. The specific calculation formula is shown as follows:
；
in the middle ofRThe grading result of the current disease degree of the tomato leaves, S Disease of the patient The total pixel number of the disease spots is represented,S total (S) Representing the number of pixels of the entire leaf.
It is understood that the extent of disease is calculated from the percentage of total leaf area to total leaf area. The solanaceae plants in this example are exemplified by tomatoes, and specific calculation criteria for the degree of tomato disease are shown in table 1:
TABLE 1 tomato leaf disease area fraction criteria
In the embodiment, the acquired solanaceae image is preprocessed based on a K-means algorithm to obtain a processed image; disease characteristic recognition is carried out on the processed image based on a disease level recognition model of preset transfer learning, a solanaceae disease image is determined according to a characteristic recognition result, the disease level recognition model of the preset transfer learning is a model obtained by transferring model parameters and parameter weights of an improved MSS-ResNet101 model to a pre-training model to obtain a transferred model and training a classification layer in the transferred model, and the classification layer is a new classification layer formed by adding a full-connection layer in front of a softmax classification layer. And counting the area of the disease spots in the solanaceae disease image, and determining the solanaceae disease degree according to the counting result. According to the invention, the K-means algorithm is used for preprocessing the obtained solanaceae image, disease feature recognition is carried out on the solanaceae image according to the migrated model, and the solanaceae disease image is determined according to the feature recognition result, so that the disease degree is further determined.
Referring to fig. 6, fig. 6 is a flowchart of a second embodiment of the method for identifying a disease grade of the solanaceae based on the transfer learning according to the present invention, and the second embodiment of the method for identifying a disease grade of the solanaceae based on the transfer learning according to the first embodiment shown in fig. 2 is provided.
In this embodiment, the step of constructing the disease level recognition model for preset transfer learning based on the parameters and weights corresponding to the pre-training model and the pre-training model includes: initializing the pre-training model by taking the parameters and the weights of the pre-training model as the initialization parameters of the network, and freezing the structure before the global average pooling layer to obtain the pre-training model to be improved; adding a full-connection layer in front of the softmax classification layer of the pre-training model to be improved to form a new classification layer, and modifying network parameters corresponding to the softmax classification layer to be suitable for a solanaceae disease degree classification task to obtain a target pre-training model; training the new classification layer in the target pre-training model to obtain a trained model; and fine-tuning network parameters corresponding to the trained model to obtain a disease grade identification model for preset transfer learning.
It should be noted that, for the tomato disease degree classification recognition model based on transfer learning, features learned by the pre-training model have positive effects on tomato disease degree classification, and fine adjustment of the whole network is beneficial to network training, so that the transfer learning is performed in a form of combining feature extraction and fine adjustment. The migration training process is shown in a migration training schematic diagram in fig. 7, wherein the migration training process can be divided into the following steps:
step one: training the ImageNet source data by using the improved MSS-ResNet101 model, ending training after convergence to derive a pre-training model and saving the parameter weight of the pre-training model to serve as the initialization parameter of the network.
Step two: and initializing the model by using parameters and weights of the pre-training model in the step one, and freezing the structure before the global average pooling layer so as not to participate in updating the model training. And a full-connection layer is added before the softmax classifying layer to form a new classifying layer for training from the beginning, and the softmax classifying layer is modified to 16 so as to be suitable for the disease degree classifying task of solanaceous plants (such as tomatoes).
Step three: and fine tuning is carried out on the trained network parameters, so that the phenomenon of over-fitting caused by small data size is avoided, and meanwhile, the high-level semantic extraction capacity of the model is improved.
It can be understood that in the process of training the model, under the condition that the number of data samples of tomato diseases of various types is relatively balanced, the larger the scale of the data samples is, the better the performance of the model obtained after training can be. Particularly, in a fine-grained image classification task, the data samples have immeasurable influence, and the rich sample images can greatly improve the recognition performance of the model. The number of samples per disease level class of tomato disease should be as large as possible in order to ensure the performance of the model. However, the method is limited by factors such as uncontrollable disease extent, large seasonal variation, high labor cost, long span time and the like, and it is difficult to collect a large number of sample images of the disease extent of tomato diseases. Thus in the solanaceae affected degree grading identification study of the present protocol, the target dataset may be composed of Plant Village tomato disease samples and Dataset of Tomato Leaves dataset, with bacterial blotch, late blight, spot blight and healthy leaves selected as the final sample data. The finally determined data set does not distinguish the disease degree of the disease, so the scheme firstly carries out preliminary division on the data set according to the area occupation ratio of the disease area on the basis of the disease degree grading standard, and then reclassifies the citrus disease samples according to the color depth of the disease area by a manual reclassification classification method so as to ensure the accuracy of classification of the data set.
It should be understood that, before outputting the trained model, experiments are performed respectively in order to verify the influence of the migration learning method and fine tuning provided by the scheme on the recognition performance of the MSS-ResNet101 model, and the accuracy rate comparison results under the training set and the verification set are shown in Table 2. Table 2 shows that the recognition effect of the non-migration learning method on the training set and the verification set is improved compared with the migration learning method. The recognition effect of the model can be further improved by fine adjustment on the basis of transfer learning, and the recognition rate of 87.05% is finally obtained on the verification set. Meanwhile, according to the schematic diagrams of the recognition rate comparison results of the verification set before and after the transfer learning shown in fig. 8, the recognition starting point of the model can be greatly improved based on the transfer learning method, so that the effectiveness and the superiority of the method proposed by the scheme can be verified. In the process of optimizing the model, in order to verify the influence of different optimization algorithms on the model identification performance, under the condition of the same test conditions, three different optimization algorithms are adopted to train the model respectively, the accuracy rate results of the model under a training set and a verification set are shown in table 2, and the drawn loss value change curves are shown in the schematic diagram of the influence of different optimization algorithms on the loss value change shown in fig. 9. As can be seen from table 3, the SGDM algorithm is improved by 1.79% and 0.85% compared with SGD and Adam, respectively, in terms of training accuracy; in terms of verification accuracy, the SGDM algorithm is improved by 1.1% and 0.64% compared with SGD and Adam respectively; meanwhile, as can be seen from fig. 9, the loss value of the SGDM algorithm is slightly reduced in the whole, and the loss value of the final convergence is lower. Therefore, the SGDM optimization algorithm can achieve better effects on both accuracy and convergence values. Therefore, before the target model is output, the model is optimized through an SGDM optimization algorithm so as to obtain an optimal model.
Table 2 comparison of recognition performance before and after transfer learning
TABLE 3 influence of different optimization algorithms on recognition performance
In a specific implementation, after the optimal model is obtained through the method, the model is output, the recognition accuracy and the misjudgment rate of the model on each disease degree category are verified, the trained optimal model is used for recognizing the test set of the tomato sample, and the confusion matrix diagram for hierarchical recognition of the tomato disease degree is shown in fig. 10. As can be clearly seen from fig. 10, the false recognition rate of different diseases with similar classification is relatively high, for example: the probability of the 7-level late blight being mistakenly identified as the 5-level late blight is 0.07, the probability of the 7-level late blight being mistakenly identified as the 9-level late blight is 0.08, and the reason for analysis is probably that the training classification sample label required by the scheme is manually calibrated and can have errors. Aiming at the problem that the classification and identification of the disease degree of the solanaceae plant has higher requirement on the key feature differentiation of a sample image, firstly, an unsupervised K-maens segmentation method is used for enhancing the disease feature in a tomato disease image sample, an improved MSS-ResNet101 model is used for carrying out pre-training on an ImageNet data set in combination with transfer learning to obtain corresponding parameters and weights, and then a local fine adjustment classification layer is used for realizing the classification and identification task of the disease degree of the solanaceae plant. The model experiment results show that the final recognition rate of the model migration learning method provided by the scheme is 88.15% on the verification set, and the SGDM optimization algorithm is verified to be more suitable for the model in the scheme than SGD and Adam.
Further, the step S20 includes:
step S201: and weakening the background in the processed image based on a disease grade identification model of preset transfer learning to obtain a target image with prominent disease characteristics.
It should be noted that, based on a disease level recognition model of preset transfer learning, the background in the processed image is weakened, and a target image highlighting the disease feature is obtained.
Step S202: and extracting disease features with different scales from the target image through a multi-scale feature structure in a disease grade identification model of preset transfer learning.
It should be noted that, the multiscale structure acceptance module in the disease level recognition model based on preset transfer learning performs multiscale feature extraction on the processed plant image to obtain a multiscale structure feature map.
It can be understood that the acceptance module expands convolution operations between different network layers, uses convolution kernels with different sizes to perform feature extraction to obtain different receptive fields, and the network structure is as shown in fig. 3, firstly, performs parallel convolution operations on input, namely, 1×1 convolution, series connection of 1×1 convolution and 3×3 convolution, series connection of 1×1 convolution and 5×5 convolution, series connection of 3×3 maximum pooling layer and 1×1 convolution, and finally splices the extracted features of the four parts in the channel dimension direction to obtain a multi-scale structural feature map. The convolution kernels with different sizes can obtain various local features, the features obtained under different receptive fields are fused, and feature information extracted by the model can be greatly enriched, so that the recognition performance is improved.
In the specific implementation, the multi-scale structural feature map is subjected to multi-path convolution according to Split operation corresponding to a convolution kernel attention mechanism SKNet in a residual structure in the disease level identification model of the preset transfer learning, so that a convolved feature map is obtained; summing the convolved feature images element by element according to the Fuse operation to obtain a new feature image; and carrying out feature fusion on the new feature images according to the Select operation to obtain disease feature images with different scales.
Step S203: and identifying the disease characteristics based on the disease grade identification model of the preset transfer learning, and determining a Solanaceae disease image according to the characteristic identification result.
The disease feature images with different scales obtained through the fusion are input into the global average pooling layer, the full connecting layer and the Softmax layer to perform disease feature recognition, and the solanaceae disease image is determined according to the feature recognition result.
In the embodiment, the acquired solanaceae image is preprocessed based on a K-means algorithm to obtain a processed image; weakening the background in the processed image based on a disease grade identification model of preset transfer learning to obtain a target image with outstanding disease characteristics; extracting disease features with different scales from the target image through a multi-scale feature structure in a disease grade identification model of preset transfer learning; and identifying the disease characteristics based on the disease grade identification model of the preset transfer learning, and determining a solanaceae disease image according to a characteristic identification result, wherein the disease grade identification model of the preset transfer learning is a model obtained by transferring model parameters and parameter weights of an improved MSS-ResNet101 model to a pre-training model to obtain a transferred model and training a classification layer in the transferred model, and the classification layer is a new classification layer formed by adding a full-connection layer in front of a softmax classification layer. And counting the area of the disease spots in the solanaceae disease image, and determining the solanaceae disease degree according to the counting result. According to the invention, the K-means algorithm is used for preprocessing the obtained solanaceae image, disease feature recognition is carried out on the solanaceae image according to the migrated model, and the solanaceae disease image is determined according to the feature recognition result, so that the disease degree is further determined.
In addition, in order to achieve the above object, the present invention also proposes a solanaceae disease grade identification device based on transfer learning, which includes a memory, a processor, and a solanaceae disease grade identification program based on transfer learning stored on the memory and executable on the processor, the solanaceae disease grade identification program based on transfer learning being configured to implement the step of solanaceae disease grade identification based on transfer learning as described above.
In addition, in order to achieve the above object, the present invention also proposes a storage medium having stored thereon a solanaceae disease level recognition program based on transfer learning, which when executed by a processor, implements the steps of the solanaceae disease level recognition method based on transfer learning as described above.
Referring to fig. 11, fig. 11 is a block diagram showing the construction of a first embodiment of the disease rank recognition device of the solanaceae based on transfer learning of the present invention.
As shown in fig. 11, the device for identifying the grade of the solanaceae disease based on the transfer learning according to the embodiment of the invention comprises:
the image preprocessing module 10 is used for preprocessing the acquired solanaceae image based on a K-means algorithm to obtain a processed image;
The disease image recognition module 20 is configured to perform disease feature recognition on the processed image based on a disease level recognition model of preset transfer learning, and determine a solanaceae disease image according to a feature recognition result, where the disease level recognition model of preset transfer learning is a model obtained by transferring model parameters and parameter weights of an improved MSS-res net101 model to a pre-training model to obtain a transferred model, and training a classification layer in the transferred model, where the classification layer is a model obtained by adding a full-connection layer before a softmax classification layer to form a new classification layer;
and the disease degree determining module 30 is configured to count the area of the disease spots in the image of the disease of the solanaceae, and determine the disease degree of the solanaceae according to the statistical result.
In the embodiment, the acquired solanaceae image is preprocessed based on a K-means algorithm to obtain a processed image; disease characteristic recognition is carried out on the processed image based on a disease level recognition model of preset transfer learning, a solanaceae disease image is determined according to a characteristic recognition result, the disease level recognition model of the preset transfer learning is a model obtained by transferring model parameters and parameter weights of an improved MSS-ResNet101 model to a pre-training model to obtain a transferred model and training a classification layer in the transferred model, and the classification layer is a new classification layer formed by adding a full-connection layer in front of a softmax classification layer. And counting the area of the disease spots in the solanaceae disease image, and determining the solanaceae disease degree according to the counting result. According to the invention, the K-means algorithm is used for preprocessing the obtained solanaceae image, disease feature recognition is carried out on the solanaceae image according to the migrated model, and the solanaceae disease image is determined according to the feature recognition result, so that the disease degree is further determined.
Further, the image preprocessing module 10 is further configured to perform pixel point segmentation on the obtained solanaceae image based on a K-means algorithm, so as to obtain a pixel point dataset; and dividing the Solanaceae image according to the pixel similarity corresponding to the pixel point data set, and extracting an image containing key disease features from the divided image.
Further, the solanaceae disease grade recognition device based on transfer learning further comprises a model training module, wherein the model training module is used for adopting an ImageNet image data set as a source data set of transfer training; training the improved MSS-ResNet101 model based on the ImageNet source data set to obtain a pre-training model; and constructing a disease grade identification model for preset transfer learning based on the parameters and weights corresponding to the pre-training model and the pre-training model.
Further, the model training module is further configured to initialize the pre-training model by using parameters and weights of the pre-training model as initialization parameters of a network, and freeze a structure in front of a global average pooling layer to obtain a pre-training model to be improved; adding a full-connection layer in front of the softmax classification layer of the pre-training model to be improved to form a new classification layer, and modifying network parameters corresponding to the softmax classification layer to be suitable for a solanaceae disease degree classification task to obtain a target pre-training model; training the new classification layer in the target pre-training model to obtain a trained model; and fine-tuning network parameters corresponding to the trained model to obtain a disease grade identification model for preset transfer learning.
Further, the disease image recognition module 20 is further configured to attenuate a background in the processed image based on a disease level recognition model of preset transfer learning, so as to obtain a target image with prominent disease features; extracting disease features with different scales from the target image through a multi-scale feature structure in a disease grade identification model of preset transfer learning; and identifying the disease characteristics based on the disease grade identification model of the preset transfer learning, and determining a Solanaceae disease image according to the characteristic identification result.
Further, the disease degree determining module 30 is further configured to perform disease type recognition on the solanaceae disease image based on the preset transfer learning disease level recognition model, so as to obtain a disease type; classifying the solanaceae disease images according to the disease categories to obtain solanaceae disease image sets of all disease types; and counting the disease spot areas in the image collection of the solanaceae diseases of each disease type, and determining the degree of the solanaceae diseases according to the counting result.
Further, the disease degree determining module 30 is further configured to obtain a total pixel amount of the disease area region; determining the disease degree of a single blade according to the total pixel quantity and the total pixel quantity of the whole blade area; and counting the single-leaf disease degree in the solanaceae disease image set of each disease type, and determining the solanaceae disease degree grade according to the disease degree counting result.
It should be understood that the foregoing is illustrative only and is not limiting, and that in specific applications, those skilled in the art may set the invention as desired, and the invention is not limited thereto.
It should be noted that the above-described working procedure is merely illustrative, and does not limit the scope of the present invention, and in practical application, a person skilled in the art may select part or all of them according to actual needs to achieve the purpose of the embodiment, which is not limited herein.
In addition, technical details which are not described in detail in the embodiment can be referred to the method for identifying the grade of the solanaceae disease based on the transfer learning provided in any embodiment of the present invention, and are not described here again.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the terms first, second, third, etc. do not denote any order, but rather the terms first, second, third, etc. are used to interpret the terms as names.
From the above description of embodiments, it will be clear to a person skilled in the art that the above embodiment method may be implemented by means of software plus a necessary general hardware platform, but may of course also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. read only memory mirror (Read Only Memory image, ROM)/random access memory (Random Access Memory, RAM), magnetic disk, optical disk), comprising instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method according to the embodiments of the present invention.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.
Claims (7)
1. The method for identifying the grade of the solanaceae disease based on the transfer learning is characterized by comprising the following steps of:
preprocessing the acquired solanaceae image based on a K-means algorithm to obtain a processed image;
performing disease feature recognition on the processed image based on a disease level recognition model of preset transfer learning, and determining a solanaceae disease image according to a feature recognition result, wherein the disease level recognition model of the preset transfer learning is a model obtained by transferring model parameters and parameter weights of an improved MSS-ResNet101 model to a pre-training model to obtain a transferred model and training a classification layer in the transferred model, and the classification layer is a new classification layer formed by adding a full-connection layer in front of a softmax classification layer;
and counting the area of the disease spots in the solanaceae disease image, and determining the solanaceae disease degree according to the counting result.
2. The method for identifying the grade of the solanaceae disease based on the transfer learning as claimed in claim 1, wherein the step of preprocessing the acquired solanaceae image based on the K-means algorithm to obtain a processed image comprises the steps of:
performing pixel point segmentation on the obtained Solanaceae image based on a K-means algorithm to obtain a pixel point data set;
and dividing the Solanaceae image according to the pixel similarity corresponding to the pixel point data set, and extracting an image containing key disease features from the divided image.
3. The method for identifying the grade of the disease of the solanaceae based on the transfer learning according to claim 1, wherein the disease grade identification model based on the preset transfer learning carries out disease feature identification on the processed image, and before the step of determining the image of the disease of the solanaceae according to the feature identification result, the method further comprises:
adopting an ImageNet image data set as a source data set of migration training;
training the improved MSS-ResNet101 model based on the ImageNet source data set to obtain a pre-training model;
and constructing a disease grade identification model for preset transfer learning based on the parameters and weights corresponding to the pre-training model and the pre-training model.
4. The method for identifying the grade of the disease of the solanaceae based on the transfer learning as claimed in claim 3, wherein the step of constructing a disease grade identification model of the preset transfer learning based on the parameters and the weights corresponding to the pre-training model and the pre-training model comprises the following steps:
initializing the pre-training model by taking the parameters and the weights of the pre-training model as the initialization parameters of the network, and freezing the structure before the global average pooling layer to obtain the pre-training model to be improved;
adding a full-connection layer in front of the softmax classification layer of the pre-training model to be improved to form a new classification layer, and modifying network parameters corresponding to the softmax classification layer to be suitable for a solanaceae disease degree classification task to obtain a target pre-training model;
training the new classification layer in the target pre-training model to obtain a trained model;
and fine-tuning network parameters corresponding to the trained model to obtain a disease grade identification model for preset transfer learning.
5. The method for identifying a disease level of the solanaceae based on transfer learning according to any one of claims 1 to 4, wherein said disease level identification model based on preset transfer learning performs disease feature identification on the processed image, and the step of determining the disease image of the solanaceae based on the feature identification result comprises:
Weakening the background in the processed image based on a disease grade identification model of preset transfer learning to obtain a target image with outstanding disease characteristics;
extracting disease features with different scales from the target image through a multi-scale feature structure in a disease grade identification model of preset transfer learning;
and identifying the disease characteristics based on the disease grade identification model of the preset transfer learning, and determining a Solanaceae disease image according to the characteristic identification result.
6. The method for identifying the grade of the solanaceae disease based on the transfer learning of claim 1, wherein the step of counting the area of the disease spots in the solanaceae disease image and determining the degree of the solanaceae disease according to the counting result comprises the following steps:
performing disease identification on the solanaceae disease image based on the disease grade identification model of the preset transfer learning to obtain disease types;
classifying the solanaceae disease images according to the disease categories to obtain solanaceae disease image sets of all disease types;
and counting the disease spot areas in the image collection of the solanaceae diseases of each disease type, and determining the degree of the solanaceae diseases according to the counting result.
7. The method for identifying the grade of the solanaceae disease based on the transfer learning of claim 6, wherein the step of counting the area of the disease spots in the image set of the solanaceae disease of each disease type and determining the degree of the solanaceae disease according to the counted result further comprises:
Acquiring the total pixel quantity of a disease spot area region;
determining the disease degree of a single blade according to the total pixel quantity and the total pixel quantity of the whole blade area;
and counting the single-leaf disease degree in the solanaceae disease image set of each disease type, and determining the solanaceae disease degree grade according to the disease degree counting result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310131684.8A CN116612307A (en) | 2023-02-15 | 2023-02-15 | Solanaceae disease grade identification method based on transfer learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310131684.8A CN116612307A (en) | 2023-02-15 | 2023-02-15 | Solanaceae disease grade identification method based on transfer learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116612307A true CN116612307A (en) | 2023-08-18 |
Family
ID=87677039
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310131684.8A Pending CN116612307A (en) | 2023-02-15 | 2023-02-15 | Solanaceae disease grade identification method based on transfer learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116612307A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117058492A (en) * | 2023-10-13 | 2023-11-14 | 之江实验室 | Two-stage training disease identification method and system based on learning decoupling |
-
2023
- 2023-02-15 CN CN202310131684.8A patent/CN116612307A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117058492A (en) * | 2023-10-13 | 2023-11-14 | 之江实验室 | Two-stage training disease identification method and system based on learning decoupling |
| CN117058492B (en) * | 2023-10-13 | 2024-02-27 | 之江实验室 | Two-stage training disease identification method and system based on learning decoupling |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111695467B (en) | Spatial spectrum full convolution hyperspectral image classification method based on super-pixel sample expansion | |
| CN111553127B (en) | Multi-label text data feature selection method and device | |
| CN114787876A (en) | System and method for image pre-processing | |
| CN106803090A (en) | A kind of image-recognizing method and device | |
| CN110942091B (en) | Semi-supervised few-sample image classification method for searching reliable abnormal data center | |
| CN107683469A (en) | A kind of product classification method and device based on deep learning | |
| JP6897749B2 (en) | Learning methods, learning systems, and learning programs | |
| Beikmohammadi et al. | SWP-LeafNET: A novel multistage approach for plant leaf identification based on deep CNN | |
| CN110751027B (en) | Pedestrian re-identification method based on deep multi-instance learning | |
| CN109492093A (en) | File classification method and electronic device based on gauss hybrid models and EM algorithm | |
| CN109993201A (en) | A kind of image processing method, device and readable storage medium storing program for executing | |
| CN111861103A (en) | Fresh tea leaf classification method based on multiple features and multiple classifiers | |
| CN108629373A (en) | A kind of image classification method, system, equipment and computer readable storage medium | |
| Bonaccorso | Hands-On Unsupervised Learning with Python: Implement machine learning and deep learning models using Scikit-Learn, TensorFlow, and more | |
| CN116612307A (en) | Solanaceae disease grade identification method based on transfer learning | |
| CN115457332A (en) | Image multi-label classification method based on graph convolution neural network and class activation mapping | |
| CN110188864B (en) | Small sample learning method based on distribution representation and distribution measurement | |
| CN111797267A (en) | Medical image retrieval method and system, electronic device and storage medium | |
| CN113764034B (en) | Method, device, equipment and medium for predicting potential BGC in genome sequence | |
| Nga et al. | Combining binary particle swarm optimization with support vector machine for enhancing rice varieties classification accuracy | |
| CN108596118A (en) | A kind of Remote Image Classification and system based on artificial bee colony algorithm | |
| CN104732246B (en) | A kind of semi-supervised coorinated training hyperspectral image classification method | |
| CN116188855A (en) | Multi-scale plant disease identification method, device, storage medium and apparatus | |
| CN111144453A (en) | Method and equipment for constructing multi-model fusion calculation model and method and equipment for identifying website data | |
| CN110837853A (en) | Rapid classification model construction method |
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 |