CN114463651A - Crop pest and disease identification method based on ultra-lightweight efficient convolutional neural network - Google Patents
Crop pest and disease identification method based on ultra-lightweight efficient convolutional neural network Download PDFInfo
- Publication number
- CN114463651A CN114463651A CN202210012912.5A CN202210012912A CN114463651A CN 114463651 A CN114463651 A CN 114463651A CN 202210012912 A CN202210012912 A CN 202210012912A CN 114463651 A CN114463651 A CN 114463651A
- Authority
- CN
- China
- Prior art keywords
- module
- layer
- neural network
- features
- ultra
- 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
- 238000000034 method Methods 0.000 title claims abstract description 42
- 241000607479 Yersinia pestis Species 0.000 title claims abstract description 31
- 201000010099 disease Diseases 0.000 title claims abstract description 31
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 31
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 29
- 238000011176 pooling Methods 0.000 claims abstract description 12
- 239000010410 layer Substances 0.000 claims description 84
- 238000010606 normalization Methods 0.000 claims description 31
- 230000004913 activation Effects 0.000 claims description 16
- 241000238631 Hexapoda Species 0.000 claims description 13
- 230000006798 recombination Effects 0.000 claims description 12
- 238000005215 recombination Methods 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 claims description 3
- 230000036541 health Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 2
- 239000002356 single layer Substances 0.000 claims description 2
- 238000012545 processing Methods 0.000 claims 2
- 238000013528 artificial neural network Methods 0.000 abstract description 6
- 238000000605 extraction Methods 0.000 abstract description 2
- 238000012423 maintenance Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 10
- 230000009466 transformation Effects 0.000 description 6
- 230000003993 interaction Effects 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 230000012010 growth Effects 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 208000015181 infectious disease Diseases 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008635 plant growth Effects 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Images
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
- G06N3/045—Combinations of 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
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 crop pest and disease identification method based on an ultra-lightweight efficient convolutional neural network. According to the method, a depth separable convolution module is used for carrying out efficient picture high-dimensional feature extraction, a spatial pyramid pooling layer is combined for carrying out local and global feature maintenance, and then the method is put into a full-connection classifier for classification training. Compared with the existing method, the method has less network parameters, the network training speed is higher, and the classification precision similar to that of a complex neural network can be achieved on a small sample set. The method has lower requirements on hardware equipment in an actual application scene, and is more suitable for deployment and application on a low-calculation-force mobile end platform.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to a crop pest and disease identification method based on an ultra-lightweight efficient convolutional neural network.
Background
Crop pest infection is a main factor influencing the healthy growth of plants, and the grain safety is greatly threatened. The real-time, rapid and accurate monitoring of the whole life cycle of crop diseases and insect pests has important significance for protecting the growth of crops, ensuring the yield of crops and maintaining the safety of grains. However, due to the fact that the diseases and pests are various, the method relying on manual visual inspection and based on experience interpretation is very low in efficiency, misdiagnosis is easy to occur, and the accuracy and timeliness of crop disease and pest monitoring are poor. Therefore, the automatic detection of crop diseases and insect pests based on the computer vision technology provides an efficient and low-cost method for real-time crop growth monitoring and management. In recent years, the explosive development of deep learning techniques in various fields has verified their excellent performance in image interpretation and understanding. Most of the existing plant disease classification methods based on deep learning adopt a convolutional neural network originally developed for a general image classification task. Although the existing neural network architectures can also be directly applied to pest and disease picture recognition of crops, the particularity of the crop pictures is not considered, the network architectures generally have huge network training parameters and have higher requirements on the computing capability of hardware equipment, and the training and reasoning time of the model takes longer time, so that the rapid and flexible deployment of the neural network architectures on platforms with limited computing power (such as mobile end platforms: mobile phones) is severely restricted.
Disclosure of Invention
The invention aims to solve the technical problems that aiming at the defects of the method, an ultra-lightweight high-efficiency convolutional neural network is provided for realizing the crop disease and insect pest identification task based on the image, and the problems that the parameter quantity of the existing model is huge, the training reasoning time is long, and the platform computing power requirement is high are mainly solved.
The proposed network consists of two parts, a depth feature extraction module that employs residual depth convolution and a classification module that receives multi-scale features enhanced by a spatial pyramid pooling layer. The network is built in a very compact design with only about 10 ten thousand parameters, which greatly facilitates the need for lightweight models in real demand. Publicly available plant data sets were used in the experiments. The proposed network shows superior advantages compared to the most advanced architectures, with the lowest computational complexity and competitive classification performance.
In order to solve the problems, the invention provides a crop disease and insect pest identification method based on an ultra-lightweight efficient convolutional neural network, which mainly comprises the following steps:
step 1, collecting health, disease and insect pest image data of different crop types;
step 2, preprocessing the collected crop image data set, and dividing the data set into a training set, a testing set and a verification set according to a certain proportion;
step 3, inputting the training set data into the ultra-lightweight high-efficiency convolutional neural network for training;
the ultra-lightweight high-efficiency convolutional neural network structure consists of 5 basic modules, wherein the first module sequentially comprises a convolutional layer, a batch normalization layer and an activation function layer, the second module comprises a plurality of pooling layers, the third module comprises two residual error depth separable convolutional modules with the step length of 1 and the step length of 2, the fourth module comprises a convolutional layer, a batch normalization layer and an activation function layer, the characteristics output by the fourth module are input into the fifth module through space pyramid pooling, and the five modules are single-layer fully-connected layers;
step 4, continuously inputting the verification set data into the network in the training process to check the result and evaluate the performance;
step 5, repeating the step 3 and the step 4, and only keeping the model with the best performance on the verification set until the training is finished;
and 6, acquiring a finally trained network model, and inputting data on the test set into the crop disease and insect pest identification model to obtain a final disease and insect pest detection result.
Further, the step 2 specifically includes: removing blurred images, out-of-focus images and images lost by a shooting subject from the collected images, and then sequentially selecting images from the data set according to the proportion of 70%, 20% and 10% as a training set, a verification set and a test set respectively.
Further, the ultra-lightweight high-efficiency convolutional neural network structure used in step 3 is composed of 5 basic modules, the first module is composed of 1 convolutional layer with a convolutional kernel of 3 × 3 and a step size of 1, 1 batch normalization layer and a ReLU activation function layer in sequence, the second module is composed of two pooling layers with a maximal convolutional kernel of 3 × 3 and a step size of 2, the third module is composed of two residual depth separable convolutional modules with a step size of 1 and a step size of 2, the fourth module is composed of 1 convolutional layer with a convolutional kernel of 1 × 1 and a step size of 1, 1 batch normalization layer and a ReLU activation function layer in sequence, the features output by the fourth module are subjected to spatial pyramid pooling operation and finally input into the fifth module, it consists of a single fully connected layer with 2016 an input dimension and 38 an output dimension.
Further, the residual depth separable convolution module with step size of 1 copies the input features into two identical features, and passes one group of features through 1 submodule, which is composed of 1 convolution kernel of 3 × 3, a depth separable convolution module with step size of 1, 1 batch normalization layer, 1 convolution kernel of 1 × 1, a convolution layer with step size of 1, 1 batch normalization layer, and 1 ReLU activation layer, and the features processed by the submodule are spliced with the original features, and then a channel recombination module is used to obtain the final output features.
Further, the residual depth separable convolution module with step size of 2 copies the input features into two identical features, and passes one group of the features through a first sub-module, which is composed of 1 convolution kernel of 3 × 3, a depth separable convolution module with step size of 2, 1 batch normalization layer, 1 convolution kernel of 1 × 1 and step size of 1, 1 batch normalization layer and 1 ReLU activation layer; another set of features is passed through a second sub-module consisting of 1 convolution kernel with 1 x 1 and a convolution layer with step size 1, 1 batch normalization layer, one ReLU active layer, 1 convolution kernel with 3 x 3 and a depth separable convolution module with step size 2, one batch normalization layer, 1 convolution kernel with 1 x 1 and a step size 1, 1 batch normalization layer, and one ReLU active layer. Finally, the features processed by the two sub-modules are spliced, and then the final output feature is obtained through a channel recombination module.
Furthermore, the depth separable convolution module takes the thought of a residual error network as reference, the input features are copied into two identical feature groups, one of the two identical feature groups is subjected to feature transformation through a depth convolution layer and then is spliced with the original features, and because the depth separable convolution is not subjected to information transformation of channel dimensions, 1 channel recombination module is designed after the features, and the information interaction of the features in the channel dimensions is enhanced.
Furthermore, the channel recombination module divides the multi-channel features into N groups, then transposes the feature groups according to the group transformation dimensionality, and groups and splices the transposed feature groups again to form a final new feature map.
Further, the loss function used in the network training in step 3 is a cross-entropy loss function.
Further, the performance evaluation in step 4 includes accuracy, recall and F1 index.
Compared with the prior art, the invention provides an ultra-lightweight high-efficiency neural network classifier for identifying agricultural pest and disease pictures, introduces deep separable convolution into a feature extractor, greatly reduces the number of training parameters required by a common feature extractor, introduces a channel recombination module for enhancing the interaction of multi-channel features in order to increase the information interaction of the features in multi-channel dimensions, and provides a spatial pyramid pooling layer on the feature classifier for maintaining the information of local and global multi-scale features of the feature map. Compared with the existing convolutional neural network classification framework, the method provided by the invention has fewer network parameters, the network training speed is higher, and the classification precision similar to that of a complex neural network can be achieved on a small sample set. Therefore, the method has lower requirements on hardware equipment in practical application scenes, and is very suitable for deployment and application on low-computing-power mobile end platforms (such as mobile phones and the like).
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention.
Fig. 2 is a structural diagram of an ultra-lightweight high-efficiency convolutional neural network constructed in an embodiment of the present invention.
Fig. 3 is a block diagram of a depth separable convolution module.
Detailed Description
The present invention will be further described with reference to the accompanying drawings.
As shown in figure 1, the invention provides a crop pest and disease identification method based on an ultra-lightweight efficient convolutional neural network, which comprises the following steps:
step 1, collecting health, disease and insect pest image data of different crop types;
step 2, preprocessing the collected crop image data set, and dividing the data set into a training set, a testing set and a verification set according to a certain proportion;
step 3, inputting training set data into the ultra-lightweight high-efficiency convolutional neural network for training, wherein a loss function used for network training is a cross entropy loss function;
step 4, continuously inputting the verification set data into the network in the training process to check the result and evaluate the performance;
step 5, repeating the step 3 and the step 4, and only keeping the model with the best performance on the verification set until the training is finished;
and 6, acquiring a finally trained network model, and inputting data on the test set into the crop disease and insect pest identification model to obtain a final disease and insect pest detection result.
Further, the step 2 specifically includes: removing blurred images, out-of-focus images and images lost by a shooting subject from the collected images, and then sequentially selecting images from the data set according to the proportion of 70%, 20% and 10% as a training set, a verification set and a test set respectively.
Further, as shown in fig. 2, the ultra-lightweight high-efficiency convolutional neural network structure used in step 3 is composed of 5 basic modules, the first module is composed of 1 convolutional layer with convolution kernel of 3 × 3 and step size of 1, 1 batch normalization layer and a ReLU activation function layer in sequence, the second module is composed of two pooling layers with maximum convolution kernel of 3 × 3 and step size of 2, the third module is composed of two residual depth separable convolution modules with step size of 1 and step size of 2, the fourth module is composed of 1 convolutional layer with convolution kernel of 1 × 1 and step size of 1, 1 batch normalization layer and a ReLU activation function layer in sequence, the features output by the fourth module are subjected to spatial pyramid pooling operation and finally input into the fifth module, it consists of a single fully connected layer with 2016 an input dimension and 38 an output dimension.
Further, as shown in fig. 3a, the residual depth separable convolution module with step size of 1 copies the input features into two identical features, and passes one set of the features through 1 sub-module, which is composed of 1 convolution kernel of 3 × 3, a depth separable convolution module with step size of 1, 1 batch normalization layer, 1 convolution kernel of 1 × 1, a convolution layer with step size of 1, 1 batch normalization layer, and 1 ReLU activation layer, and the features processed by the sub-module are spliced with the original features, and then pass through a channel recombination module to obtain the final output features.
Further, as shown in fig. 3b, the residual depth separable convolution module with step size 2 first copies the input features into two identical features, and passes one set of the features through a first sub-module, which is composed of 1 convolution kernel of 3 × 3, a depth separable convolution module with step size 2, 1 batch normalization layer, 1 convolution kernel of 1 × 1, a convolution layer with step size 1, 1 batch normalization layer, and 1 ReLU activation layer; another set of features is passed through a second sub-module consisting of 1 convolution kernel with 1 x 1 and a convolution layer with step size 1, 1 batch normalization layer, one ReLU active layer, 1 convolution kernel with 3 x 3 and a depth separable convolution module with step size 2, one batch normalization layer, 1 convolution kernel with 1 x 1 and a step size 1, 1 batch normalization layer, and one ReLU active layer. Finally, the features processed by the two sub-modules are spliced, and then the final output feature is obtained through a channel recombination module.
Furthermore, the depth separable convolution module takes the thought of a residual error network as reference, the input features are copied into two identical feature groups, one of the two identical feature groups is subjected to feature transformation through a depth convolution layer and then is spliced with the original features, and because the depth separable convolution is not subjected to information transformation of channel dimensions, 1 channel recombination module is designed after the features, and the information interaction of the features in the channel dimensions is enhanced.
Furthermore, the channel recombination module divides the multi-channel features into N groups, then transposes the feature groups according to the group transformation dimensionality, and groups and splices the transposed feature groups again to form a final new feature map.
The method and the existing method are respectively tested on the PLANT-VILLAGE open data set for classification precision and classification efficiency, and the test results are as follows:
TABLE 1 Classification accuracy comparison
TABLE 2 comparison of Classification efficiencies
VGG16, MobileNet _ v2, Xception, ShuffleNet _ v2 in Table 1 are the existing methods, compared with the existing methods, the method of the invention achieves 98.13% accuracy and 97.49 recall rate, and the F1 index of 0.9776 is very close to the optimal Xception method in the overall classification accuracy. As can be seen from Table 2, compared with other methods, the method of the present invention has very few model parameters, memory occupation and floating point operands, and in terms of operation speed, the method of the present invention is 2.6 times faster than ShuffleNet _ v2 and 380 times faster than Xception. In conclusion, compared with the existing neural network classification framework, the method has fewer network parameters and higher network training speed, can achieve the classification precision performance similar to the optimal network framework on a small sample set, achieves the optimization of hundreds of times on the operation efficiency, and only occupies 6Mb in memory space. Thus, the inventive method is well suited for operation on low-computing platforms.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (8)
1. A crop pest and disease identification method based on an ultra-lightweight efficient convolutional neural network is characterized by comprising the following steps:
step 1, collecting health, disease and insect pest image data of different crop types;
step 2, preprocessing the collected crop image data set, and dividing the data set into a training set, a testing set and a verification set according to a certain proportion;
step 3, inputting the training set data into the ultra-lightweight high-efficiency convolutional neural network for training;
the ultra-lightweight high-efficiency convolutional neural network structure consists of 5 basic modules, wherein the first module sequentially comprises a convolutional layer, a batch normalization layer and an activation function layer, the second module comprises a plurality of pooling layers, the third module comprises two residual error depth separable convolutional modules with the step length of 1 and the step length of 2, the fourth module comprises a convolutional layer, a batch normalization layer and an activation function layer, the characteristics output by the fourth module are input into the fifth module through space pyramid pooling, and the five modules are single-layer fully-connected layers;
step 4, continuously inputting the verification set data into the network in the training process to check the result and evaluate the performance;
step 5, repeating the step 3 and the step 4, and only keeping the model with the best performance on the verification set until the training is finished;
and 6, acquiring a finally trained network model, and inputting data on the test set into the crop disease and insect pest identification model to obtain a final disease and insect pest detection result.
2. The crop pest and disease identification method based on the ultra-lightweight efficient convolutional neural network as claimed in claim 1, characterized in that: and 2, removing blurred images, defocused images and images lost by a shooting subject from the acquired images, and then sequentially selecting images from the data set according to the proportion of 70%, 20% and 10% as a training set, a verification set and a test set respectively.
3. The crop pest and disease identification method based on the ultra-lightweight efficient convolutional neural network as claimed in claim 1, characterized in that: in step 3, the first module is formed by stacking 1 convolution layer with convolution kernel of 3 × 3 and step length of 1, 1 batch normalization layer and a ReLU activation function layer in sequence;
the second module is formed by stacking two pooling layers with the maximum convolution kernel of 3 x 3 and the step length of 2;
the fourth module is formed by overlapping 1 convolution layer with convolution kernel of 1 × 1 and step length of 1, 1 batch normalization layer and a ReLU activation function layer in sequence;
the fifth module consists of a single fully connected layer with 2016 an input dimension and 38 an output dimension.
4. The crop pest and disease identification method based on the ultra-lightweight efficient convolutional neural network as claimed in claim 1, characterized in that: the specific processing process of the residual depth separable convolution module with the step length of 1 is as follows;
firstly, copying input features into two identical features, enabling one group of features to pass through 1 submodule, wherein the submodule is composed of 1 depth separable convolution module with convolution kernel of 3 x 3 and step length of 1, 1 batch normalization layer, 1 convolution kernel of 1 x 1 and step length of 1, 1 batch normalization layer and 1 ReLU activation layer, splicing the features processed by the submodule and original features, and then obtaining final output features through a channel recombination module.
5. The crop pest and disease identification method based on the ultra-lightweight efficient convolutional neural network as claimed in claim 1, characterized in that: the specific processing procedure of the residual depth separable convolution module with the step length of 2 is as follows;
firstly, copying input features into two identical features, and enabling one group of features to pass through a first sub-module, wherein the sub-module consists of 1 depth separable convolution module with convolution kernel of 3 × 3 and step length of 2, 1 batch normalization layer, 1 convolution kernel of 1 × 1 and step length of 1, 1 batch normalization layer and 1 ReLU activation layer; and the other group of features passes through a second sub-module, the sub-module consists of 1 convolution kernel with 1 × 1 and the step size of 1 convolution layer, 1 batch normalization layer, a ReLU active layer, 1 convolution kernel with 3 × 3 and the step size of 2 depth separable convolution modules, one batch normalization layer, 1 convolution kernel with 1 × 1 and the step size of 1 convolution layer, 1 batch normalization layer and one ReLU active layer, and finally the processed features of the two sub-modules are spliced and the final output features are obtained through a channel recombination module.
6. The crop pest and disease identification method based on the ultra-lightweight high-efficiency convolutional neural network as claimed in claim 4 or 5, characterized in that: the channel recombination module divides the multichannel characteristics into N groups, then transforms the characteristics according to the group dimension, transposes the characteristic groups, and groups and splices the transposed characteristic groups again to form a final new characteristic diagram.
7. The crop pest and disease identification method based on the ultra-lightweight efficient convolutional neural network as claimed in claim 1, characterized in that: and 3, the loss function used for network training in the step 3 is a cross entropy loss function.
8. The crop pest and disease identification method based on the ultra-lightweight efficient convolutional neural network as claimed in claim 1, characterized in that: the performance evaluation in step 4 included accuracy, recall, and F1 index.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210012912.5A CN114463651A (en) | 2022-01-07 | 2022-01-07 | Crop pest and disease identification method based on ultra-lightweight efficient convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210012912.5A CN114463651A (en) | 2022-01-07 | 2022-01-07 | Crop pest and disease identification method based on ultra-lightweight efficient convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114463651A true CN114463651A (en) | 2022-05-10 |
Family
ID=81409281
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210012912.5A Pending CN114463651A (en) | 2022-01-07 | 2022-01-07 | Crop pest and disease identification method based on ultra-lightweight efficient convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114463651A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115116054A (en) * | 2022-07-13 | 2022-09-27 | 江苏科技大学 | Insect pest identification method based on multi-scale lightweight network |
| CN117122308A (en) * | 2023-07-24 | 2023-11-28 | 苏州大学 | Electrocardiogram measurement method and system based on mobile phone built-in acceleration sensor |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111126127A (en) * | 2019-10-23 | 2020-05-08 | 武汉大学 | High-resolution remote sensing image classification method guided by multi-level spatial context characteristics |
| CN111814622A (en) * | 2020-06-29 | 2020-10-23 | 华南农业大学 | Crop pest type identification method, system, equipment and medium |
| CN113627281A (en) * | 2021-07-23 | 2021-11-09 | 中南民族大学 | SK-EfficientNet-based lightweight crop disease identification method |
-
2022
- 2022-01-07 CN CN202210012912.5A patent/CN114463651A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111126127A (en) * | 2019-10-23 | 2020-05-08 | 武汉大学 | High-resolution remote sensing image classification method guided by multi-level spatial context characteristics |
| CN111814622A (en) * | 2020-06-29 | 2020-10-23 | 华南农业大学 | Crop pest type identification method, system, equipment and medium |
| CN113627281A (en) * | 2021-07-23 | 2021-11-09 | 中南民族大学 | SK-EfficientNet-based lightweight crop disease identification method |
Non-Patent Citations (1)
| Title |
|---|
| 程越;刘志刚;: "基于轻量型卷积神经网络的交通标志识别方法", 计算机系统应用, no. 02, 15 February 2020 (2020-02-15) * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115116054A (en) * | 2022-07-13 | 2022-09-27 | 江苏科技大学 | Insect pest identification method based on multi-scale lightweight network |
| CN115116054B (en) * | 2022-07-13 | 2024-05-24 | 江苏科技大学 | Multi-scale lightweight network-based pest and disease damage identification method |
| CN117122308A (en) * | 2023-07-24 | 2023-11-28 | 苏州大学 | Electrocardiogram measurement method and system based on mobile phone built-in acceleration sensor |
| CN117122308B (en) * | 2023-07-24 | 2024-04-12 | 苏州大学 | Electrocardiogram measurement method and system based on mobile phone built-in acceleration sensor |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Identification of maize leaf diseases using improved deep convolutional neural networks | |
| CN114463651A (en) | Crop pest and disease identification method based on ultra-lightweight efficient convolutional neural network | |
| CN110110596B (en) | Hyperspectral image feature extraction, classification model construction and classification method | |
| CN111696101A (en) | Light-weight solanaceae disease identification method based on SE-Inception | |
| CN108090447A (en) | Hyperspectral image classification method and device under double-branch deep structure | |
| CN115116054B (en) | Multi-scale lightweight network-based pest and disease damage identification method | |
| CN115331104A (en) | Crop planting information extraction method based on convolutional neural network | |
| CN114972208B (en) | YOLOv 4-based lightweight wheat scab detection method | |
| CN112308825A (en) | SqueezeNet-based crop leaf disease identification method | |
| CN110414338B (en) | Pedestrian re-identification method based on sparse attention network | |
| CN115410087A (en) | Transmission line foreign matter detection method based on improved YOLOv4 | |
| CN114972264A (en) | Method and device for identifying mung bean leaf spot based on MS-PLNet model | |
| CN117372881A (en) | Intelligent identification method, medium and system for tobacco plant diseases and insect pests | |
| CN115965864A (en) | Lightweight attention mechanism network for crop disease identification | |
| CN116152198A (en) | Tomato leaf spot recognition method based on Wave-SubNet lightweight model | |
| CN116258914A (en) | Remote sensing image classification method based on machine learning and local and global feature fusion | |
| CN117744745B (en) | Image optimization method and optimization system based on YOLOv network model | |
| CN110852398B (en) | Aphis gossypii glover recognition method based on convolutional neural network | |
| CN114120046B (en) | Lightweight engineering structure crack identification method and system based on phantom convolution | |
| Rajeswarappa et al. | Crop Pests Identification based on Fusion CNN Model: A Deep Learning | |
| CN115063700A (en) | Detection method based on small-scale pine wood nematode disease tree | |
| CN114299091A (en) | Automatic weed segmentation method based on DA-Net | |
| CN118429329B (en) | Road disease identification method and device based on RD-YOLO network | |
| CN117876843B (en) | Efficient crop disease identification method capable of dynamically reducing image redundancy | |
| CN113436200B (en) | RGB image classification method based on lightweight segmentation convolutional 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 |