CN113344009B - Light and small network self-adaptive tomato disease feature extraction method - Google Patents
Light and small network self-adaptive tomato disease feature extraction method Download PDFInfo
- Publication number
- CN113344009B CN113344009B CN202110621361.8A CN202110621361A CN113344009B CN 113344009 B CN113344009 B CN 113344009B CN 202110621361 A CN202110621361 A CN 202110621361A CN 113344009 B CN113344009 B CN 113344009B
- Authority
- CN
- China
- Prior art keywords
- tomato leaf
- layer
- feature
- foi
- tomato
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a light and small network self-adaptive tomato disease feature extraction method, which comprises the following steps: preprocessing a tomato leaf disease image; the orientation of all the preprocessed tomato leaf disease images is unified through a shaping mechanism; constructing a light and small network model, and extracting global features and interesting features of the tomato leaf disease image; the light and small network model comprises a global feature extraction sub-network GFE-Net and a feature extraction sub-module FOIE-Block of interest. The invention can extract the interested features in the picture to a great extent, realizes high accuracy and high robustness of blade identification, and simultaneously greatly reduces the parameter number of the network.
Description
Technical Field
The invention relates to the field of accurate agriculture and computer vision application, in particular to a light and small network self-adaptive tomato disease feature extraction method.
Background
In recent years, various factors such as climate change, pollinator reduction, plant diseases and the like form serious threats to global grain safety, wherein the plant diseases are important factors which cause serious reduction of quality and quantity of agricultural products, when crops get the diseases, the physiological functions of the crops can be greatly reduced, the plants cannot reach the optimal production state due to the small size, and therefore, the yield is not high and the economic benefit is low. In the tomato planting process, various diseases severely restrict the production of tomatoes, and common diseases include late blight, early blight, leaf mold, mosaic virus disease, spot blight and the like.
In the current research work, the development process of plant disease detection technology can be divided into three stages. The first stage is manual identification, and disease type is judged by experience, and the method is time-consuming, labor-consuming, high in subjectivity and low in accuracy. The second stage is a traditional machine learning identification method, features are extracted by utilizing feature engineering, and then classification identification is carried out through a classifier, but the machine learning-based method still comprises a large number of artificial influence factors, and the feature extraction engineering needs to be carried out under a specific environment, so that the process is complex. The third stage is a deep learning recognition mode, and the black box characteristics of the neural network are utilized, so that the selection of the characteristics does not need to be manually participated, and the end-to-end system engineering is realized. Although the identification accuracy is better for many deep networks, the portability of the model is low due to the characteristics of complex structure, deep layer number, large parameter quantity and the like.
Disclosure of Invention
In order to meet the precision requirements of modern agriculture, the invention provides a light and small network self-adaptive tomato disease feature extraction method, which can extract interesting features in pictures to a great extent, realize high accuracy and high robustness of blade identification, and greatly reduce the parameter quantity of a network.
In order to achieve the above purpose, the technical scheme of the invention is as follows: a light and small network self-adaptive tomato disease feature extraction method comprises the following steps:
preprocessing a tomato leaf disease image;
the orientation of all the preprocessed tomato leaf disease images is unified through a shaping mechanism;
constructing a light and small network model, and extracting global features and interesting features of the tomato leaf disease image; the light and small network model comprises a global feature extraction sub-network GFE-Net and a feature extraction sub-module FOIE-Block of interest.
Furthermore, all the preprocessed tomato leaf disease images are unified in orientation through a orthomorphism mechanism, and the method specifically comprises the following steps: the directions of all the tomato leaf disease images are unified to be the leaf tip upwards and the leaf handle downwards, so that the tomato leaf is positioned in the middle of the image, and the deviation angle between the tomato leaf and the center line of the image is ensured not to exceed +/-5 degrees.
Further, the working process of the orthomorphism mechanism is as follows:
acquiring a tomato leaf disease image;
carrying out gray-scale treatment on the tomato leaf disease image and carrying out Gaussian filtering;
extracting edge characteristics of a tomato leaf disease image through a Canny operator, finding a minimum circumscribed rectangle, and obtaining a rotation angle theta;
and carrying out affine transformation on the original tomato leaf disease image by using the rotation angle theta.
Further, the global feature extraction sub-network GFE-Net is configured to extract global features of a tomato leaf disease image, and comprises 4 Fire modules, 2 convolution layers, 3 maximum pooling layers, 1 global average pooling layer, 1 dropout layer and 1 softmax layer.
Further, each Fire module includes a Squeeze layer and an Expand layer, each layer consisting of only convolution kernels of 1 x 1 or 3 x 3 size.
Furthermore, the interesting feature extraction submodule FOIE-Block is used for extracting interesting features of the tomato leaf disease image and comprises 1 global average pooling layer, 2 full connection layers, 2 activation functions and 1 matrix multiplication operation.
Furthermore, the specific operation of the interesting feature extraction submodule FOIE-Block for extracting the interesting features of the tomato leaf disease image is as follows:
extraction of sub-network GFE-Net convolution from global featuresThe dimension of the tomato leaf characteristic diagram of the layer output is [ W, H, C ]]Inputting the tomato leaf feature map into a feature extraction sub-module FOIE-Block of interest, firstly, summing all values on the tomato leaf feature map with the size of H multiplied by W to average through a global averaging pooling layer (global average pooling), wherein the output value of each channel is g C The calculation formula is as follows:
wherein g C For the output value of each channel, u c (i, j) is the value in the feature map, and the total output of the C channels is G, then G= [ G ] 1 ,g 2 ,…,g C ]The dimension is [1, C]；
And carrying out operation processing on the total output G through a full connection layer and a nonlinear function ReLU, wherein the full connection layer has the following formula:
wherein the dimension of R isWeight parameter of full connection layer +.>This layer reduces the number of channels from C to +.>Meanwhile, the nonlinear relation among the channels is obtained by utilizing an activation function ReLU, namely, the characteristic weight of each channel is learned;
the variable R is subjected to operation processing through a full-connection layer and a gate function Sigmoid, wherein the weight parameter of the full-connection layer is as followsThe full-connection layer operation of the layer restores the number of channels to the original number, and after the gate function sigmoid further captures the nonlinear relation among the channels, the interesting feature Score FOI Score is output, the value of the interesting feature Score is recorded as FOI_S, and the operation formula is as follows:
wherein FOI_S is a scalar comprising C values, and [ FOI_S ] 1 ，FOI_S 2 ，...，FOI_S C ]The magnitude of the value of (c) represents the magnitude of the feature of interest score for the corresponding channel.
Weighting the obtained feature scores of interest channel by channel onto the tomato leaf profile, i.e. each tomato leaf profile [ W, H,1]，[W,H,2]，...，[W,H,C]Respectively with FOI_S 1 ，FOI_S 2 ，...，FOI_S C The operation formula of multiplication is as follows:
X C ＝FOI_S C ·u C
the output result of the layer is C H multiplied by W tomato leaf feature graphs which are consistent with the input dimension, but the values on the tomato leaf feature graphs are recalibrated at the moment, namely the feature values of the interesting features become larger, and the irrelevant feature values become smaller or even are restrained, so that the interesting features are extracted.
By adopting the technical scheme, the invention can obtain the following technical effects:
1. the network accuracy designed by the method is high and can reach 97.89%.2. The network model is small, the model size can be 2.64MB, and the portability is stronger. 3. The recognition speed is high and reaches 101 ms/sheet, and the method is easy to use in production. 4. The network has good stability and robustness, the accuracy rate of identifying the disease picture containing Gaussian noise reaches 83.32%, and the tomato leaf diseases and insect pests can be accurately identified in a complex environment.
Drawings
FIG. 1 is a graph of disease classification for a new data set PV1 obtained by resampling;
FIG. 2 is a diagram of the operation of the orthographic mechanism;
FIG. 3 is a diagram of a Fire module configuration;
FIG. 4 is a diagram of the feature of interest extraction submodule FOIE-Block;
FIG. 5 is an overall block diagram of a lightweight small network;
FIG. 6 is a graph of accuracy versus loss for a network training set and validation set;
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings and detailed description. The following examples will provide those skilled in the art with a more complete understanding of the present invention and are not intended to limit the invention to the embodiments described.
Along with the proposal of accurate agriculture concepts, accuracy, rapidness and intelligence become the latest targets of agricultural development, so that in order to meet the accurate requirements of modern agriculture, an efficient disease identification method is proposed according to the current research work progress and in combination with the actual environment background of agriculture, and the designed light and small-sized neural network model simultaneously meets the requirements of high accuracy, high speed, high robustness and low parameter quantity.
The light and small network of the invention enhances the representation of the interesting features (FOI, feature of interest) by learning the feature weights on the basis of obtaining the global features, and simultaneously suppresses the irrelevant features to realize the self-adaptive extraction of the features. The network comprises a global feature extraction (GFE, global feature extraction) sub-network GFE-Net and a interested feature extraction (FOIE, feature of interest extraction) sub-module FOIE-Block, wherein the GFE-Net extracts global features such as color, texture, shape and the like of a picture through a series of convolution pooling operations to obtain the integral attribute of the picture; the FOIE-Block uses a channel attention mechanism to perform the squeize operation on the feature map containing the global features to obtain the information descriptors of the channel dimension, generates a weight for each feature channel by capturing the nonlinear interaction relation between the channel descriptors, and finally acts the weight on the input feature map, thereby improving the expression of the interesting features and inhibiting the irrelevant features. In summary, the light and small network designed by the invention can extract the interested features in the picture to a great extent through the self-adaptive extraction of the features, thereby realizing high accuracy and high robustness of blade identification and greatly reducing the parameter number of the network.
Example 1
The embodiment provides a light and small network self-adaptive tomato disease feature extraction method, which comprises the following specific implementation steps:
s1, preprocessing a tomato leaf disease image
Specifically, tomato leaf disease images were downloaded from the large public Plant disease dataset forum Plant Village, and the screened dataset was 11250 pieces in total of 8 categories. The sample resampling process calculates the average value of 8 categories, then randomly oversamples the category below the average value, and finally generates a new data set PV1 with balanced data comparison, as shown in FIG. 1.
S2, unifying and concrete all the preprocessed tomato leaf disease images through a orthomorphism mechanism, wherein the adopted orthomorphism mechanism aims at unifying the directions of all the tomato leaf disease images into the directions of upward leaf tips and downward leaf handles, so that the tomato leaf is positioned in the middle position of the image as much as possible, and the deviation angle between the tomato leaf and the central line of the image is ensured not to exceed +/-5 degrees. The orientation of all the blade images is unified through the orthographic mechanism, so that the convolutional neural network is facilitated to extract more and more detailed disease features, and the learning efficiency of the network is greatly improved. The workflow of the orthographic mechanism is shown in fig. 2: (1) acquiring a tomato leaf disease image; (2) graying treatment is carried out on the tomato leaf disease image; (3) carrying out Gaussian filtering on the tomato leaf disease image after the graying treatment; (4) extracting edge characteristics of a tomato leaf disease image through a Canny operator; (5) finding the minimum circumscribed rectangle and obtaining a rotation angle theta; (6) and carrying out affine transformation on the original tomato leaf disease image by using the rotation angle theta.
S3, constructing a light and small network model, and extracting global features and interesting features of the tomato leaf disease image as shown in FIG. 5 to realize high accuracy and strong robustness of leaf identification; the light and small network model comprises a global feature extraction sub-network GFE-Net and a feature extraction sub-module FOIE-Block of interest.
Specifically, the global feature extraction sub-network GFE-Net is used as an infrastructure for introducing channel attention and is mainly used for extracting global features of tomato leaf diseases. The base network is a network comprising 4 Fire modules, 2 convolution layers, 3 max pooling layers, 1 global average pooling layer, 1 dropout layer, and 1 softmax layer. As shown in fig. 3, each Fire module includes a Squeeze layer and an Expand layer, each consisting of only convolution kernels of 1×1 or 3×3 size, which reduces the amount of parameters of the network by a factor of 9. The 2 convolution layers are respectively positioned in front of and behind the basic network, so that basic characteristics and high-level semantic information of the picture can be effectively extracted. The 3 largest pooling layers can extract feature images with low resolution and strong semantic information. The global average pooling layer is used for replacing the full connection layer, so that the calculation complexity and parameter quantity of the network can be effectively reduced, and the network training speed is improved. The introduction of the Dropout layer can effectively prevent overfitting. The Softmax function calculates the probability of each category and outputs the prediction result.
The interesting feature extraction submodule FOIE-Block realizes the extraction of interesting Features (FOI) by means of the working mechanism of channel attention. Channel attention may explicitly model the interdependencies between channels, re-adaptively calibrating the characteristic responses of the channels. The importance degree of each characteristic channel is automatically acquired through learning, and then the useful characteristics are selectively promoted according to the importance degree, the characteristics with weak expression capacity are restrained, and the expression capacity of the whole network is enhanced.
As shown in fig. 4, the interesting feature extraction submodule FOIE-Block includes 1 global average pooling layer, 2 full connection layers, 2 activation functions and 1 matrix multiplication operation, and the process of extracting interesting features by the module is as follows:
the first step: a channel descriptor is obtained. Tomato leaf feature map dimension output from sub-network GFE-Net convolution layer is [ W, H, C ]]Inputting the values into FOIE-Block, and obtaining all values on a tomato leaf characteristic diagram with the size of H×W through a global average pooling layer (global average pooling)And averaging the output value of each channel g C The operation formula is as follows:
wherein g C For the output value of each channel, if the total output of the C channels is G, g= [ G ] 1 ,g 2 ,…,g C ]The dimension is [1, C]. Obviously g C The larger the channel, the more characteristic information that the channel contains.
And a second step of: and (5) reducing the dimension and obtaining the nonlinear relation among the channels. The output G obtained above is subjected to operation processing through a full connection layer and a nonlinear function ReLU, and the full connection layer has the following formula:
wherein the dimension of R isWeight parameter of full connection layer +.>This layer reduces the number of channels from C to +.>Thereby reducing the amount of computation. And meanwhile, the nonlinear relation among the channels is acquired by utilizing an activation function ReLU, namely, the characteristic weight of each channel is learned.
And a third step of: and (5) up-scaling to obtain the score of the interesting feature. The output R obtained in the last step is subjected to operation processing through a full-connection layer and a gate function Sigmoid, wherein the weight parameter of the full-connection layer is as followsThe full connection layer operation of the layer restores the channel number to the original numberAfter further capturing the nonlinear relation between channels, the gate function Sigmoid outputs a feature Score of interest, the value of which is recorded as foi_s, and the operation formula is as follows:
wherein FOI_S is a scalar comprising C values, and [ FOI_S ] 1 ，FOI_S 2 ，...，FOI_S C ]The magnitude of the value of (2) represents the height of the interesting characteristic score of the corresponding channel;
fourth step: and extracting the interesting characteristic. The obtained feature scores of interest are weighted channel by channel onto the previous tomato leaf feature map, i.e. each feature map [ W, H,1]，[W,H,2]，...，[W,H,C]Respectively with FOI_S 1 ，FOI_S 2 ，...，FOI_S C The operation formula of multiplication is as follows:
X C ＝FOI_S C ·u C
the output result of the layer is C H multiplied by W tomato leaf feature graphs, and the input dimension is the same, but the values on the tomato leaf feature graphs are recalibrated at the moment, namely the feature values of the interesting features become larger, the irrelevant feature values become smaller and even are inhibited, so that the interesting features are extracted.
The specific implementation method of the light and small network model comprises the following steps:
the preprocessed and shaped tomato leaf disease image is input into a sub-network GFE-Net, simple features of the image are extracted through a 7×7 convolution layer, then a large number of 1×1 convolutions and 3×3 convolutions in 4 Fire modules are utilized to carry out convolution operation on the simple features for multiple times, and further richer global features are extracted. Because the obtained global feature map has higher dimension, 3 largest pooling layers are adopted in the network to reduce dimension of the feature map, and the parameter quantity and the calculation amount of the network are effectively reduced under the condition of hardly influencing the network performance. After global feature extraction is completed, the feature graphs are input into a sub-module FOIE-Block, and the sub-module compresses the feature graphs and learns feature weights from channel dimensions, so that the interesting Features (FOI) are found, the expression of the FOI is enhanced, useless features are restrained, and the network learning efficiency is improved. And (3) the new tomato leaf feature map after feature recalibration is subjected to a convolution layer of 1 multiplied by 1 and a global average pooling layer to obtain high-level semantic information output of features, and finally, disease categories are output through softmax. The light and small network model provided by the invention can carry out self-adaptive extraction on the picture characteristics, not only can extract global characteristics, but also can strengthen the expression of interesting characteristics, and greatly improves the comprehensive performance of the network.
The change curves of the accuracy and the loss of the light and small network training set and the verification set are shown in fig. 6, the accuracy of the verification set is gradually increased, the loss is gradually decreased, the training is carried out until 100 rounds of basic convergence, and the network training is completed. The confusion matrix on the test set is shown in table 1, and the identification accuracy of each disease category on the test set is 96.58% of early blight, 99.34% of health, 99.45% of late blight, 98.65% of leaf mold, 99.34% of mosaic disease, 99.44% of spot blight, 98.98% of bacterial spot disease and 99.38% of leaf mite disease.
TABLE 1 confusion matrix on test set
The embodiments of the present invention are preferred embodiments and are not intended to be limiting in any way. The technical features or combinations of technical features described in the embodiments of the present invention should not be regarded as isolated, and they may be combined with each other to achieve a better technical effect. Additional implementations are also included within the scope of the preferred embodiments of the present invention and should be understood by those skilled in the art to which the inventive examples pertain.
Claims (1)
1. The light and small network self-adaptive tomato disease feature extraction method is characterized by comprising the following steps of:
preprocessing a tomato leaf disease image;
the orientation of all the preprocessed tomato leaf disease images is unified through a shaping mechanism;
constructing a light and small network model, and extracting global features and interesting features of the tomato leaf disease image; the light and small network model comprises a global feature extraction sub-network GFE-Net and an interesting feature extraction sub-module FOIE-Block;
the orientation of all the preprocessed tomato leaf disease images is unified through a shaping mechanism, and the method specifically comprises the following steps: the directions of all the tomato leaf disease images are unified to be the leaf tip upwards and the leaf handle downwards, so that the tomato leaf is positioned in the middle of the image, and the deviation angle between the tomato leaf and the center line of the image is not more than +/-5 degrees;
the working process of the shaping mechanism is as follows:
acquiring a tomato leaf disease image;
carrying out gray-scale treatment on the tomato leaf disease image and carrying out Gaussian filtering;
extracting edge characteristics of a tomato leaf disease image through a Canny operator, finding a minimum circumscribed rectangle, and obtaining a rotation angle theta;
affine transformation is carried out on the original tomato leaf disease image by utilizing the rotation angle theta;
the global feature extraction sub-network GFE-Net is used for extracting global features of tomato leaf disease images and comprises 4 Fire modules, 2 convolution layers, 3 maximum pooling layers, 1 global average pooling layer, 1 dropout layer and 1 softmax layer;
each Fire module comprises a Squeeze layer and an expansion layer, and each layer is composed of convolution kernels with the size of 1 multiplied by 1 or 3 multiplied by 3;
the interesting feature extraction submodule FOIE-Block is used for extracting interesting features of the tomato leaf disease image and comprises 1 global average pooling layer, 2 full connection layers, 2 activation functions and 1 matrix multiplication operation;
the specific operation of extracting the interesting features of the tomato leaf disease image by the interesting feature extraction submodule FOIE-Block is as follows:
tomato leaf feature map dimension output by global feature extraction sub-network GFE-Net convolution layer is [ W, H, C ]]Inputting the characteristic map of the tomato leafIn the interesting feature extraction submodule FOIE-Block, all values on the tomato leaf feature map with the size of H multiplied by W are summed and averaged through a global averaging pooling layer, and the output value of each channel is g C The calculation formula is as follows:
wherein g C For the output value of each channel, u c (i, j) is the value in the feature map, and the total output of the C channels is G, then G= [ G ] 1 ,g 2 ,…,g C ]The dimension is [1, C]；
And carrying out operation processing on the total output G through a full connection layer and a nonlinear function ReLU, wherein the full connection layer has the following formula:
wherein the dimension of R isWeight parameter of full connection layer +.>This layer reduces the number of channels from C to +.>Meanwhile, the nonlinear relation among the channels is obtained by utilizing an activation function ReLU, namely, the characteristic weight of each channel is learned;
the variable R is subjected to operation processing through a full-connection layer and a gate function Sigmoid, wherein the weight parameter of the full-connection layer is as followsFull connection layer operation of the layer willAfter the number of channels is restored to the original number and the nonlinear relation among the channels is further captured by the gate function sigmoid, the interesting feature Score FOI Score is output, the value of the interesting feature Score is recorded as FOI_S, and the operation formula is as follows:
wherein FOI_S is a scalar comprising C values, and [ FOI_S ] 1 ，FOI_S 2 ，...，FOI_S C ]The magnitude of the value of (2) represents the height of the interesting characteristic score of the corresponding channel;
weighting the obtained feature scores of interest channel by channel onto the tomato leaf profile, i.e. each tomato leaf profile [ W, H,1]，[W,H,2]，...，[W,H,C]Respectively with FOI_S 1 ，FOI_S 2 ，...，FOI_S C The operation formula of multiplication is as follows:
X C ＝FOI_S C ·u C
the output result of the layer is C H multiplied by W tomato leaf feature graphs which are consistent with the input dimension, but the values on the tomato leaf feature graphs are recalibrated at the moment, namely the feature values of the interesting features become larger, and the irrelevant feature values become smaller or even are restrained, so that the interesting features are extracted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110621361.8A CN113344009B (en) | 2021-06-03 | 2021-06-03 | Light and small network self-adaptive tomato disease feature extraction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110621361.8A CN113344009B (en) | 2021-06-03 | 2021-06-03 | Light and small network self-adaptive tomato disease feature extraction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113344009A CN113344009A (en) | 2021-09-03 |
| CN113344009B true CN113344009B (en) | 2023-08-01 |
Family
ID=77473445
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110621361.8A Active CN113344009B (en) | 2021-06-03 | 2021-06-03 | Light and small network self-adaptive tomato disease feature extraction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113344009B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114550164B (en) * | 2021-12-28 | 2024-05-31 | 吉林农业大学 | Deep learning-based tomato leaf disease identification research method |
| CN115050014A (en) * | 2022-06-15 | 2022-09-13 | 河北农业大学 | Small sample tomato disease identification system and method based on image text learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111696101A (en) * | 2020-06-18 | 2020-09-22 | 中国农业大学 | Light-weight solanaceae disease identification method based on SE-Inception |
| CN112183635A (en) * | 2020-09-29 | 2021-01-05 | 南京农业大学 | Method for realizing segmentation and identification of plant leaf lesions by multi-scale deconvolution network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101879207B1 (en) * | 2016-11-22 | 2018-07-17 | 주식회사 루닛 | Method and Apparatus for Recognizing Objects in a Weakly Supervised Learning Manner |
-
2021
- 2021-06-03 CN CN202110621361.8A patent/CN113344009B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111696101A (en) * | 2020-06-18 | 2020-09-22 | 中国农业大学 | Light-weight solanaceae disease identification method based on SE-Inception |
| CN112183635A (en) * | 2020-09-29 | 2021-01-05 | 南京农业大学 | Method for realizing segmentation and identification of plant leaf lesions by multi-scale deconvolution network |
Non-Patent Citations (1)
| Title |
|---|
| 基于无人机遥感图像的松材线虫病监测技术研究;李浩;徐航煌;郑恒宇;陈学永;;中国农机化学报(第09期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113344009A (en) | 2021-09-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110148120B (en) | Intelligent disease identification method and system based on CNN and transfer learning | |
| CN107977932B (en) | Face image super-resolution reconstruction method based on discriminable attribute constraint generation countermeasure network | |
| Faisal et al. | IHDS: Intelligent harvesting decision system for date fruit based on maturity stage using deep learning and computer vision | |
| CN111860330B (en) | Apple leaf disease identification method based on multi-feature fusion and convolutional neural network | |
| CN113344009B (en) | Light and small network self-adaptive tomato disease feature extraction method | |
| Islam et al. | Rice leaf disease recognition using local threshold based segmentation and deep CNN | |
| CN109740483A (en) | A kind of rice growing season detection method based on deep-neural-network | |
| CN107563389A (en) | A kind of corps diseases recognition methods based on deep learning | |
| CN111400536A (en) | Low-cost tomato leaf disease identification method based on lightweight deep neural network | |
| CN111797760A (en) | Improved crop pest and disease identification method based on Retianet | |
| CN110827273A (en) | Tea disease detection method based on regional convolution neural network | |
| CN111127423A (en) | Rice pest and disease identification method based on CNN-BP neural network algorithm | |
| CN113221913A (en) | Agriculture and forestry disease and pest fine-grained identification method and device based on Gaussian probability decision-level fusion | |
| CN115331104A (en) | Crop planting information extraction method based on convolutional neural network | |
| CN114972208A (en) | YOLOv 4-based lightweight wheat scab detection method | |
| Hu et al. | Self-adversarial training and attention for multi-task wheat phenotyping | |
| Rai et al. | Classification of diseased cotton leaves and plants using improved deep convolutional neural network | |
| CN113344077A (en) | Anti-noise solanaceae disease identification method based on convolution capsule network structure | |
| CN110766082B (en) | Plant leaf disease and insect pest degree classification method based on transfer learning | |
| Monigari et al. | Plant leaf disease prediction | |
| CN115330833A (en) | Fruit yield estimation method with improved multi-target tracking | |
| CN113989689A (en) | Crop pest and disease identification method and system based on unmanned aerial vehicle | |
| Kiran et al. | Plant Leaf Disease Detection Using Efficient Image Processing and Machine Learning Algorithms | |
| Zheng et al. | Fruit tree disease recognition based on convolutional neural networks | |
| CN111695560A (en) | Method for actively positioning and focusing crop diseases and insect pests based on convolutional neural network |
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 |