CN118314144B - Plant leaf disease identification method and system based on depth intensive residual error module - Google Patents
Plant leaf disease identification method and system based on depth intensive residual error module Download PDFInfo
- Publication number
- CN118314144B CN118314144B CN202410742080.1A CN202410742080A CN118314144B CN 118314144 B CN118314144 B CN 118314144B CN 202410742080 A CN202410742080 A CN 202410742080A CN 118314144 B CN118314144 B CN 118314144B
- Authority
- CN
- China
- Prior art keywords
- feature extraction
- depth
- extraction stage
- attention
- layer
- 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.)
- Active
Links
- 201000010099 disease Diseases 0.000 title claims abstract description 76
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 76
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000000605 extraction Methods 0.000 claims abstract description 74
- 238000012545 processing Methods 0.000 claims abstract description 16
- 239000011159 matrix material Substances 0.000 claims description 33
- 238000010606 normalization Methods 0.000 claims description 32
- 230000004913 activation Effects 0.000 claims description 25
- 239000013598 vector Substances 0.000 claims description 22
- 238000010586 diagram Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 3
- 241000196324 Embryophyta Species 0.000 abstract description 36
- 240000003768 Solanum lycopersicum Species 0.000 abstract description 29
- 235000007688 Lycopersicon esculentum Nutrition 0.000 abstract description 27
- 230000006870 function Effects 0.000 description 21
- 238000012544 monitoring process Methods 0.000 description 5
- 241000607479 Yersinia pestis Species 0.000 description 4
- 241000238631 Hexapoda Species 0.000 description 3
- 239000000575 pesticide Substances 0.000 description 3
- 238000012271 agricultural production Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000002779 inactivation Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 238000003892 spreading Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- 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
- Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
- Y02A40/10—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in agriculture
Landscapes
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of plant disease identification, and discloses a plant leaf disease identification method and a system based on a depth intensive residual error module, wherein the method collects plant leaf images, the plant leaf images are subjected to feature extraction through a first feature extraction stage, a second feature extraction stage, a third feature extraction stage, a fourth feature extraction stage and a fifth feature extraction stage which are sequentially arranged, and finally, a classification result is obtained through the processing of a full connection layer of the fifth feature extraction stage; the invention adopts the depth-intensive residual error module and the multi-head attention module with space offset, the depth-intensive residual error module can fully extract the characteristic information of the tomato leaf diseases from the second characteristic extraction stage to the fifth characteristic extraction stage, and the multi-head attention module with space offset can extract the global disease information of the image and actively shield irrelevant information, thereby maximizing the utilization rate of computing resources.
Description
Technical Field
The invention belongs to the technical field of plant disease identification, and particularly relates to a plant leaf disease identification method and system based on a depth intensive residual error module.
Background
Plant leaf disease identification by image identification technology has been studied and applied in recent years, and has wide application in the fields of crop pest monitoring and control, precision agriculture, agricultural scientific research and the like. Tomato leaf disease identification is a hotspot problem in the fields of computer vision and smart agriculture, which is a very challenging task. On one hand, the disease focus of the tomato leaf disease is small and dense, the disease spot identification difficulty is high, and the similarity of various disease categories is high, so that the identification is difficult. On the other hand, the deep learning model for realizing the identification of the tomato leaf diseases at the present stage is formed by stacking simple convolution blocks, the feature extraction capacity of the model is poor, and the requirements of the tomato leaf disease identification task cannot be met.
The current methods for identifying tomato leaf diseases are mainly divided into two types. One is to build a simple recognition model to realize the recognition of tomato leaf diseases. Another class of methods is to use recognition models of other fields in the field of tomato leaf diseases. The first technical scheme is to realize the identification of tomato leaf diseases by constructing a simple identification model, and the method usually only utilizes a plurality of simple convolution layer stacks, so that focus information of complete disease images cannot be fully extracted, and the problems of low efficiency and low precision are generally existed. The other type of method has a great limitation, firstly, the identification model in other fields is used for identifying the tomato leaf diseases, and due to the fact that large differences exist in the identification objects, the methods and the system adaptation, the method cannot be applied to a certain extent, and secondly, the method uses data sets in other fields during model training, and the problems of low identification precision and low efficiency are generally existed in the field of identifying the tomato leaf diseases.
Disclosure of Invention
The invention aims to provide a plant leaf disease identification method and system based on a depth-dense residual error module, which can be used for identifying tomato leaf diseases, has strong tomato leaf characteristic extraction capability, can specifically identify focuses in the tomato leaf diseases, can accurately identify disease types on the tomato leaf, such as late blight, early blight, spot diseases and the like, and can evaluate the disease incidence degree of the diseases at the same time, so as to realize monitoring and control of the tomato leaf diseases of a plantation.
The invention is realized by the following technical scheme: the plant leaf disease identification method based on the depth intensive residual error module comprises the steps of collecting plant leaf images, carrying out feature extraction on the plant leaf images through a first feature extraction stage, a second feature extraction stage, a third feature extraction stage, a fourth feature extraction stage and a fifth feature extraction stage which are sequentially arranged, and finally, processing by a full-connection layer of the fifth feature extraction stage to obtain a classification result;
in the second feature extraction stage, the feature map passes through two depth intensive residual modules, then passes through an activation function and a batch normalization layer, and then the feature map passes through a convolution layer to finish the downsampling operation; the structure and the processing procedure of the third feature extraction stage are the same as those of the second feature extraction stage;
After the fourth feature extraction stage, the feature map is processed by six depth intensive residual modules, then is processed by an activation function and a batch normalization layer, and reaches a space offset encoder, and the feature map output by the space offset encoder is subjected to a convolution layer to finish the downsampling operation;
In the fifth feature extraction stage, the feature map is firstly processed by two depth intensive residual modules, then processed by an activation function and a batch normalization layer, then processed by a space offset encoder and finally processed by a full-connection layer to obtain a classification result;
The depth intensive residual error module comprises five depth separable convolution layers, wherein the five depth separable convolution layers are in intensive connection, and then the output characteristic images are in residual error connection with the characteristic images input by the depth intensive residual error module through an activation function and a batch normalization layer;
The input characteristic diagram of the space offset encoder is normalized by one layer, then enters the feedforward module, then enters the other layer for normalization, finally reaches the multi-head attention module with space offset, and the processing procedure of the multi-head attention module with space offset is as follows:
Firstly, a two-dimensional matrix of a feature map is projected to a vector space of a query, a key and a value in parallel, then reaches a multi-head attention module, calculates an attention score matrix of the query and the key, adds a space bias limit to the multi-head attention moment matrix, calculates a softening maximum function for the last dimension of the attention score matrix, continuously uses the attention weight obtained in the last step, and performs weighted summation on vectors in the value matrix to obtain an attention weighted value vector;
splicing the attention weighted value vectors of all the heads on the last dimension to form an output tensor;
The stitched output tensor is passed through another learnable linear layer, which is mapped back to the input dimension of the model as the final output of the multi-headed attention layer.
Specifically, in the first feature extraction stage, the plant leaf image is subjected to downsampling operation through two convolution layers, and the downsampled feature map sequentially passes through a nonlinear activation function and batch normalization processing.
The invention provides a plant leaf disease identification system based on a depth-dense residual error module, which comprises an image acquisition device and a plant leaf disease identification device, wherein a plant leaf disease identification model is arranged in the plant leaf disease identification device, and comprises a first feature extraction stage, a second feature extraction stage, a third feature extraction stage, a fourth feature extraction stage and a fifth feature extraction stage which are sequentially arranged;
in the second feature extraction stage, the feature map passes through two depth intensive residual modules, then passes through an activation function and a batch normalization layer, and then the feature map passes through a convolution layer to finish the downsampling operation; the structure and the processing procedure of the third feature extraction stage are the same as those of the second feature extraction stage;
After the fourth feature extraction stage, the feature map is processed by six depth intensive residual modules, then is processed by an activation function and a batch normalization layer, and reaches a space offset encoder, and the feature map output by the space offset encoder is subjected to a convolution layer to finish the downsampling operation;
In the fifth feature extraction stage, the feature map is firstly processed by two depth intensive residual modules, then processed by an activation function and a batch normalization layer, then processed by a space offset encoder and finally processed by a full-connection layer to obtain a classification result;
The depth intensive residual error module comprises five depth separable convolution layers, wherein the five depth separable convolution layers are in intensive connection, and then the output characteristic images are in residual error connection with the characteristic images input by the depth intensive residual error module through an activation function and a batch normalization layer;
The input characteristic diagram of the space offset encoder is normalized by one layer, then enters the feedforward module, then enters the other layer for normalization, finally reaches the multi-head attention module with space offset, and the processing procedure of the multi-head attention module with space offset is as follows:
Firstly, a two-dimensional matrix of a feature map is projected to a vector space of a query, a key and a value in parallel, then reaches a multi-head attention module, calculates an attention score matrix of the query and the key, adds a space bias limit to the multi-head attention moment matrix, calculates a softening maximum function for the last dimension of the attention score matrix, continuously uses the attention weight obtained in the last step, and performs weighted summation on vectors in the value matrix to obtain an attention weighted value vector;
splicing the attention weighted value vectors of all the heads on the last dimension to form an output tensor;
The stitched output tensor is passed through another learnable linear layer, which is mapped back to the input dimension of the model as the final output of the multi-headed attention layer.
The invention also provides a nonvolatile computer storage medium, wherein the computer storage medium stores computer executable instructions, and the computer executable instructions execute the plant leaf disease identification method based on the depth-intensive residual error module.
The invention builds a brand new plant leaf disease recognition model, which has five characteristic extraction stages, wherein the first stage performs downsampling twice for extracting the shallow characteristics of the tomato leaf disease, and the last three stages perform downsampling once for extracting the multi-scale characteristic information of the tomato leaf disease. In addition, the invention provides a depth-intensive residual error module and a multi-head attention module with space offset, wherein the depth-intensive residual error module can fully extract the characteristic information of tomato leaf diseases from the second characteristic extraction stage to the fifth characteristic extraction stage, and the multi-head attention module with space offset can extract the global disease information of an image and actively shield irrelevant information, so that the utilization rate of computing resources is maximized.
The invention is mainly applied to the field of tomato planting, can also be applied to the field of health management of other plants such as forestry plant diseases and insect pests monitoring, landscaping plant diseases and insect pests monitoring and the like, and is mainly applied to agricultural production links in a centralized manner.
Drawings
FIG. 1 is a schematic diagram of a plant leaf disease recognition model according to the present invention.
Fig. 2 is a schematic diagram of a depth-dense residual block.
Fig. 3 is a schematic diagram of a spatially offset encoder.
Detailed Description
The invention is described in further detail below with reference to the examples and the attached drawings.
As shown in fig. 1, in the plant leaf disease identification method based on the depth-dense residual error module, plant leaf images are collected, feature extraction is performed on the plant leaf images through five feature extraction stages which are sequentially arranged, the five feature extraction stages are used for extracting feature information of different scales of the disease images, and finally a classification result is obtained through full-connection layer processing of the fifth feature extraction stage.
As shown in fig. 1, in the first feature extraction stage, the plant leaf image is first subjected to two convolution layers with 3×3 convolution kernels, a step size of 2, and an edge filling of 1, so as to implement a downsampling operation, and reduce the spatial dimension of the input plant leaf image to 1/4 of the original dimension, so as to extract a more abstract and advanced feature representation, while reducing the complexity of subsequent computation. Next, the feature map after downsampling sequentially passes through a nonlinear activation function (ReLU) and a batch normalization process, and after the processes, the feature map enters a second feature extraction stage.
After the feature map enters the second feature extraction stage, the feature map passes through two depth-dense residual modules, then passes through an activation function (ReLU) and a batch normalization layer (Batch Normalization), and then the feature map passes through a convolution layer with a 3×3 convolution kernel and a step length of 2 and an edge filling of 1 to finish the downsampling operation, wherein the size of the output feature map is 1/8 of that of the plant leaf image which is input initially. The structure of the depth dense residual error module is shown in fig. 2, the image is firstly subjected to five depth separable convolution layers (DW convolution layers) with the convolution kernel size of 3, the step length of 1 and the edge filling of 1, the five depth separable convolution layers are in a dense connection mode, and then the output characteristic image is subjected to residual error connection with the characteristic image input by the depth dense residual error module through an activation function (ReLU) and a batch normalization layer, so that the depth dense residual error module has strong characteristic interaction capability, model complexity is effectively reduced, and better performance is provided for the model.
After the feature map enters the third feature extraction stage, the structure and the processing procedure of the third feature extraction stage are the same as those of the second feature extraction stage, and the size of the output feature map is 1/16 of that of the plant leaf image which is input initially.
After the feature map enters a fourth feature extraction stage, the feature map is processed by six depth-intensive residual modules, then is processed by an activation function (ReLU) and a batch normalization layer, and reaches a space offset encoder, the feature map output by the space offset encoder is subjected to downsampling operation by a convolution layer with a convolution kernel of 3×3 and a step length of 2 and an edge filling of 1, and the size of the output feature map is 1/32 of that of the plant leaf image which is input initially.
After the feature map enters a fifth feature extraction stage, the feature map passes through two depth-dense residual modules, is processed by an activation function (ReLU) and a batch normalization layer, passes through a space offset encoder, and finally is processed by a full-connection layer to obtain a classification result.
The structure of the spatial offset encoder is shown in fig. 3, in the spatial offset encoder, the input feature map is normalized by one layer (Layer Normalization), then enters the feedforward module, sequentially passes through the linear layer, the activation function and 2 random inactivation layers in the feedforward module, then enters another layer for normalization (Layer Normalization), finally reaches the multi-head attention module (transducer) with spatial offset, and the detailed processing procedure of the multi-head attention module with spatial offset is as follows:
Firstly, a two-dimensional matrix of a feature map is projected to a vector space of a Query (Query), a Key (Key) and a Value (Value) in parallel, then a multi-head attention module is reached, an attention score matrix of the Query and the Key is calculated, then a space bias limit is added to the multi-head attention moment matrix, a softening maximum (softmax) function is calculated for the last dimension of the attention score matrix, the attention weight obtained in the last step is continuously used, and the vectors in the Value matrix are weighted and summed to obtain an attention weighted Value vector which can be expressed as:
；
Wherein Attention i,j represents the Attention weight matrix between position i and position j, A transpose of the input weight matrix X i representing the position i, X k representing the input weight matrix of the position k, S i-k representing the softmax function operation output result of the position i and the position k; i, j, k are position defining symbols; n is a global spatial representation, θ is an independent spatial representation, and S represents a calculate softening maximum (softmax) function operation.
And repeating the steps for each head, and finally splicing the attention weighted value vectors of all the heads in the last dimension to form an output tensor, wherein the output tensor can be expressed as follows by a formula:
；
Wherein Y i represents the output of the spatial-biased encoder at position i, X j represents the input weight matrix at position j, and S i-j represents the softmax function operation output result at position i and position j;
The stitched output tensor is passed through another learnable linear layer, which is mapped back to the input dimension of the model as the final output of the multi-headed attention layer.
The invention further provides a plant leaf disease identification system based on the depth-dense residual error module, which comprises an image acquisition device and a plant leaf disease identification device, wherein a plant leaf disease identification model is arranged in the plant leaf disease identification device, and comprises five feature extraction stages which are sequentially arranged.
Another embodiment of the present invention provides a non-volatile computer storage medium storing computer-executable instructions for performing the above-described plant leaf disease identification method based on a depth-dense residual module.
According to the invention, in a tomato plantation, farmers shoot leaf images of tomato plants in the field by installing a plurality of high-definition cameras or using unmanned aerial vehicles, then the leaf images are input into a tomato leaf disease identification system based on a depth-intensive residual error module and a space offset transducer structure, and the system is trained by a large number of tomato leaf disease images before application, so that disease types on tomato leaves, such as late blight, early blight, spot disease and the like, can be accurately identified, and meanwhile, the disease degree is evaluated. Once serious diseases of tomato plants are found, the system can timely send out early warning, simultaneously recommend corresponding pesticide varieties and dosage suggestions for farmers, timely implement accurate control on areas affected by the diseases, and avoid further spreading of the diseases. After treatment, the system can record the time, place and environment data of the occurrence of the diseases, and the post-research of the cause of the occurrence of the diseases of the tomato leaves is convenient. The tomato leaf disease identification system based on the depth intensive residual error module and the space offset transducer structure is used for realizing monitoring and control of tomato leaf disease in a plantation, so that damage of plant diseases and insect pests to crops can be reduced, the using amount of pesticides can be reduced, the efficiency of agricultural production can be improved, and precise application and sustainable development of pesticides can be realized.
The preferred embodiments of the invention disclosed above are intended only to assist in the explanation of the invention. The preferred embodiments are not exhaustive or to limit the invention to the precise form disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best understand and utilize the invention. The invention is limited only by the claims and the full scope and equivalents thereof.
Claims (8)
1. The plant leaf disease identification method based on the depth intensive residual error module is characterized by comprising the steps of collecting plant leaf images, carrying out feature extraction on the plant leaf images through a first feature extraction stage, a second feature extraction stage, a third feature extraction stage, a fourth feature extraction stage and a fifth feature extraction stage which are sequentially arranged, and finally obtaining a classification result through the full-connection layer processing of the fifth feature extraction stage;
in the second feature extraction stage, the feature map passes through two depth intensive residual modules, then passes through an activation function and a batch normalization layer, and then the feature map passes through a convolution layer to finish the downsampling operation; the structure and the processing procedure of the third feature extraction stage are the same as those of the second feature extraction stage;
After the fourth feature extraction stage, the feature map is processed by six depth intensive residual modules, then is processed by an activation function and a batch normalization layer, and reaches a space offset encoder, and the feature map output by the space offset encoder is subjected to a convolution layer to finish the downsampling operation;
In the fifth feature extraction stage, the feature map is firstly processed by two depth intensive residual modules, then processed by an activation function and a batch normalization layer, then processed by a space offset encoder and finally processed by a full-connection layer to obtain a classification result;
The depth intensive residual error module comprises five depth separable convolution layers, wherein the five depth separable convolution layers are in intensive connection, and then the output characteristic images are in residual error connection with the characteristic images input by the depth intensive residual error module through an activation function and a batch normalization layer;
The input characteristic diagram of the space offset encoder is normalized by one layer, then enters the feedforward module, then enters the other layer for normalization, finally reaches the multi-head attention module with space offset, and the processing procedure of the multi-head attention module with space offset is as follows:
Firstly, a two-dimensional matrix of a feature map is projected to a vector space of a query, a key and a value in parallel, then reaches a multi-head attention module, calculates an attention score matrix of the query and the key, adds a space bias limit to the multi-head attention moment matrix, calculates a softening maximum function for the last dimension of the attention score matrix, continuously uses the attention weight obtained in the last step, and performs weighted summation on vectors in the value matrix to obtain an attention weighted value vector;
splicing the attention weighted value vectors of all the heads on the last dimension to form an output tensor;
The stitched output tensor is passed through another learnable linear layer, which is mapped back to the input dimension of the model as the final output of the multi-headed attention layer.
2. The plant leaf disease identification method based on the depth-dense residual error module according to claim 1, wherein in the first feature extraction stage, the plant leaf image is subjected to downsampling operation through two convolution layers, and the downsampled feature map sequentially passes through a nonlinear activation function and batch normalization processing.
3. The method for identifying plant leaf disease based on depth-dense residual module of claim 2, wherein the convolution layer has a3 x3 convolution kernel with a step size of 2 and an edge fill of 1.
4. The method for identifying plant leaf disease based on depth-dense residual module of claim 1, wherein the process of deriving the attention-weighted value vector is formulated as:
；
Wherein Attention i,j represents the Attention weight matrix between position i and position j, A transpose of the input weight matrix X i representing the position i, X k representing the input weight matrix of the position k, S i-k representing the softmax function operation output result of the position i and the position k; i, j, k are position defining symbols; n is a global spatial representation, θ is an independent spatial representation, and S represents a calculate softening maximum function operation.
5. The method for identifying plant leaf diseases based on depth-dense residual modules of claim 4 wherein the attention weighted value vectors of all heads are stitched in the last dimension to form an output tensor expressed as:
；
Where Y i represents the output of the spatial-biased encoder at position i, X j represents the input weight matrix at position j, and S i-j represents the softmax function operation output results at positions i and j.
6. The plant leaf disease recognition system based on the depth intensive residual error module comprises an image acquisition device and a plant leaf disease recognition device, wherein a plant leaf disease recognition model is arranged in the plant leaf disease recognition device and comprises a first feature extraction stage, a second feature extraction stage, a third feature extraction stage, a fourth feature extraction stage and a fifth feature extraction stage which are sequentially arranged; the method is characterized in that:
in the second feature extraction stage, the feature map passes through two depth intensive residual modules, then passes through an activation function and a batch normalization layer, and then the feature map passes through a convolution layer to finish the downsampling operation; the structure and the processing procedure of the third feature extraction stage are the same as those of the second feature extraction stage;
After the fourth feature extraction stage, the feature map is processed by six depth intensive residual modules, then is processed by an activation function and a batch normalization layer, and reaches a space offset encoder, and the feature map output by the space offset encoder is subjected to a convolution layer to finish the downsampling operation;
In the fifth feature extraction stage, the feature map is firstly processed by two depth intensive residual modules, then processed by an activation function and a batch normalization layer, then processed by a space offset encoder and finally processed by a full-connection layer to obtain a classification result;
The depth intensive residual error module comprises five depth separable convolution layers, wherein the five depth separable convolution layers are in intensive connection, and then the output characteristic images are in residual error connection with the characteristic images input by the depth intensive residual error module through an activation function and a batch normalization layer;
The input characteristic diagram of the space offset encoder is normalized by one layer, then enters the feedforward module, then enters the other layer for normalization, finally reaches the multi-head attention module with space offset, and the processing procedure of the multi-head attention module with space offset is as follows:
Firstly, a two-dimensional matrix of a feature map is projected to a vector space of a query, a key and a value in parallel, then reaches a multi-head attention module, calculates an attention score matrix of the query and the key, adds a space bias limit to the multi-head attention moment matrix, calculates a softening maximum function for the last dimension of the attention score matrix, continuously uses the attention weight obtained in the last step, and performs weighted summation on vectors in the value matrix to obtain an attention weighted value vector;
Repeating the steps for each head, and finally splicing the attention weighted value vectors of all the heads in the last dimension to form an output tensor;
The stitched output tensor is passed through another learnable linear layer, which is mapped back to the input dimension of the model as the final output of the multi-headed attention layer.
7. The plant leaf disease recognition system based on the depth-dense residual module of claim 6, wherein the first feature extraction stage is to implement a downsampling operation of the plant leaf image by first passing through two convolution layers, and the downsampled feature map is sequentially subjected to a nonlinear activation function and batch normalization.
8. A non-transitory computer storage medium storing computer executable instructions, wherein the computer executable instructions perform the depth-dense residual module-based plant leaf disease identification method of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410742080.1A CN118314144B (en) | 2024-06-11 | 2024-06-11 | Plant leaf disease identification method and system based on depth intensive residual error module |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410742080.1A CN118314144B (en) | 2024-06-11 | 2024-06-11 | Plant leaf disease identification method and system based on depth intensive residual error module |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118314144A CN118314144A (en) | 2024-07-09 |
| CN118314144B true CN118314144B (en) | 2024-08-06 |
Family
ID=91731669
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410742080.1A Active CN118314144B (en) | 2024-06-11 | 2024-06-11 | Plant leaf disease identification method and system based on depth intensive residual error module |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118314144B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116824366A (en) * | 2023-06-14 | 2023-09-29 | 天津商业大学 | Crop disease identification method based on local selection and feature interaction |
| CN116883364A (en) * | 2023-07-17 | 2023-10-13 | 安徽大学 | Apple leaf disease identification method based on CNN and Transformer |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114821238B (en) * | 2022-05-10 | 2022-09-13 | 山东建筑大学 | Image identification method and system of convolutional neural network based on global detail supplementation |
| CN117422961A (en) * | 2023-07-21 | 2024-01-19 | 哲库科技(上海)有限公司 | Image processing method and device, chip, electronic equipment and storage medium |
| CN117456274A (en) * | 2023-11-27 | 2024-01-26 | 湖州师范学院 | Hyperspectral image classification method, hyperspectral image classification device, computer equipment and storage medium |
-
2024
- 2024-06-11 CN CN202410742080.1A patent/CN118314144B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116824366A (en) * | 2023-06-14 | 2023-09-29 | 天津商业大学 | Crop disease identification method based on local selection and feature interaction |
| CN116883364A (en) * | 2023-07-17 | 2023-10-13 | 安徽大学 | Apple leaf disease identification method based on CNN and Transformer |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118314144A (en) | 2024-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| You et al. | A DNN-based semantic segmentation for detecting weed and crop | |
| Razfar et al. | Weed detection in soybean crops using custom lightweight deep learning models | |
| Gayathri et al. | Image analysis and detection of tea leaf disease using deep learning | |
| Chen et al. | Weed detection in sesame fields using a YOLO model with an enhanced attention mechanism and feature fusion | |
| Blok et al. | The effect of data augmentation and network simplification on the image‐based detection of broccoli heads with Mask R‐CNN | |
| Sembiring et al. | Development of concise convolutional neural network for tomato plant disease classification based on leaf images | |
| Ilyas et al. | Multi-scale context aggregation for strawberry fruit recognition and disease phenotyping | |
| Moazzam et al. | A review of application of deep learning for weeds and crops classification in agriculture | |
| Mishra et al. | Weed Species Identification in Different Crops Using Precision Weed Management: A Review. | |
| CN111797760A (en) | Improved crop pest and disease identification method based on Retianet | |
| Farooq et al. | Knowledge transfer via convolution neural networks for multi-resolution lawn weed classification | |
| Haichen et al. | Weeds and crops classification using deep convolutional neural network | |
| CN114972208A (en) | YOLOv 4-based lightweight wheat scab detection method | |
| Sehree et al. | Olive trees cases classification based on deep convolutional neural network from unmanned aerial vehicle imagery | |
| Arif et al. | Weeds detection and classification using convolutional long-short-term memory | |
| Djimeli-Tsajio et al. | Improved detection and identification approach in tomato leaf disease using transformation and combination of transfer learning features | |
| Modi et al. | An automated weed identification framework for sugarcane crop: a deep learning approach | |
| Wang et al. | The identification of straight-curved rice seedling rows for automatic row avoidance and weeding system | |
| Zhang et al. | Automatic counting of lettuce using an improved YOLOv5s with multiple lightweight strategies | |
| CN118314144B (en) | Plant leaf disease identification method and system based on depth intensive residual error module | |
| Vengaiah et al. | CNN model suitability analysis for prediction of tomato leaf diseases | |
| Harvyanti et al. | Application of Convolutional Neural Network for Identifying Cocoa Leaf Disease | |
| Sahu et al. | CNN based disease detection in Apple Leaf via transfer learning | |
| CN116310369A (en) | Multi-image attention feature fusion-based optimal visual angle recognition method for crop insect pests | |
| Fang et al. | Classification system study of soybean leaf disease based on deep learning |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |