CN107016405B - A kind of pest image classification method based on classification prediction convolutional neural networks - Google Patents
A kind of pest image classification method based on classification prediction convolutional neural networks Download PDFInfo
- Publication number
- CN107016405B CN107016405B CN201710103514.3A CN201710103514A CN107016405B CN 107016405 B CN107016405 B CN 107016405B CN 201710103514 A CN201710103514 A CN 201710103514A CN 107016405 B CN107016405 B CN 107016405B
- Authority
- CN
- China
- Prior art keywords
- image
- classification
- pest
- training
- network
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of pest image classification methods based on classification prediction convolutional neural networks, solve the low defect of pest image classification accuracy compared with prior art.The present invention is the following steps are included: being collected training image and pre-processing；Image sample data is labeled；Disaggregated model of the training based on classification prediction convolutional neural networks；Testing image is pre-processed；Pest image classification is carried out automatically based on hierarchy model.Present invention employs classification prediction frame, the segmentation result of first forecast image carries out final classification prediction then in conjunction with general image jointly.
Description
Technical field
The present invention relates to prediction sorting technique field, specifically a kind of evils based on classification prediction convolutional neural networks
Worm image classification method.
Background technique
Pest is the formidable enemy in crop growth, has generation within crops entire growth period, crops can be caused big
Measure the underproduction.Existing pest identifies, sorter makees to be mainly to complete by a small number of plant protection experts and agriculture technical staff.But pest
It is many kinds of, each plant protection expert it is poor its can also can only identification division pest.More and more signs show to pest point
Increasing for class demand has increasingly sharpened with the relatively small number of contradiction of pest systematicalian.Now in area of pattern recognition, based on deep
Degree study learning algorithm become numerous scholars research hot spot, be widely used in computer vision field, as recognition of face,
Image classification, image segmentation etc., and achieve preferable effect.However, applying in pest classification of images method and system
Then there is the problem that discrimination is not high, robustness is poor, this is also due to caused by the diversity of pest sample, feature complexity
's.
Therefore, it how to be directed to the sample label of pest, has realized pest identification using the nerual network technique for having supervision
As technical problem urgently to be solved.
Summary of the invention
The purpose of the present invention is to solve the low defects of pest image classification accuracy in the prior art, provide a kind of base
It solves the above problems in the pest image classification methods of classification prediction convolutional neural networks.
To achieve the goals above, technical scheme is as follows:
A kind of pest image classification method based on classification prediction convolutional neural networks, comprising the following steps:
Training image is collected and is pre-processed, collects several width images as training image, all training images are equal
Size normalization processing is carried out, 256 × 256 pixels is processed into, obtains several training samples；
Image sample data is labeled, samples pictures content is manually marked, mark out image segmentation boundary,
Classification and pest species divide the image into pest, crop, background three classes, and combined training sample is as training sample data collection
It closes；
Disaggregated model of the training based on classification prediction convolutional neural networks, training is based on classification prediction convolutional neural networks
Disaggregated model completes the training of neural network classification model using training sample as input；
Testing image is pre-processed, pest image to be measured is normalized by 256 × 256 pixels, is obtained
To new testing image；
Pest image classification is carried out automatically based on hierarchy model, and by treated, testing image inputs trained classification
It predicts in convolutional neural networks model, carries out the automatic identification of pest image type.
Disaggregated model of the training based on classification prediction convolutional neural networks the following steps are included:
Image segmentation network model of the training based on FCNN；
The full convolutional network layer that the number of plies is 7 layers is set, inputs training sample, uses disease pest, crop, the background manually marked
Segmentation result carries out the training of network model using deep learning frame Caffe as study mark, exports as can be to pest
The full convolutional neural networks model that image is split；
Image classification model of the training based on CNN,
It is 8 layers of convolution sorter network that the number of plies, which is arranged, and the pest that obtains after being divided according to trained parted pattern makees
Object, three width image of background, in addition undivided general image, four width images as input, classify by the pest manually marked in total
As a result as study mark, the training of network model is carried out using deep learning frame Caffe, exports image classification model.
It is described pest image classification is carried out automatically based on hierarchy model the following steps are included:
Testing image is input to and is trained in FCNN parted pattern, it is pre- to obtain pest, crop, the segmentation of background in image
Survey result；
Original image is split according to segmentation prediction result, obtains three width new images: pest image, crop image,
Background image；
Totally four width images are defeated as input for pest image, crop image, the background image that testing image and segmentation are obtained
Enter in the disaggregated model based on CNN completed to training, prediction obtains the maximum classification of possibility and exports as classification results.
Image segmentation network model of the training based on FCNN the following steps are included:
7 layers of full convolutional network structure are constructed, the size of every layer of convolutional network is as follows:
First layer convolution kernel size is 11x11, and characteristic pattern number is 96；Second layer convolution kernel is 5x5, and characteristic pattern number is
256；Third and fourth, five layers of convolution kernel be 3x3, characteristic pattern number is respectively 384,384,256；Six, the seven layers of convolution kernel are
1x1, characteristic pattern number are respectively 512,3；
The output size of the last one convolutional layer is 27x27x3, and wherein 27x27 is two-dimensional image space size, and 3 be segmentation
The number of class, 3 target values pest, crop according to belonging to each of which position receptive field on any position 27X27 still carry on the back
Jing Laiding, if including more than one classification in receptive field, using accounting for that classification of most number of pixels as mark；
For each position of 27X27, softmax operation is carried out in 3 classifications, final loss function is in 27X27
The sum of softmax loss on position L, calculation formula are as follows:
Wherein Divide classification for the mark on position (h, w), M is segmentation
The number of classification is equal to 3；
Definition based on network and objective function inputs training sample data, carries out mould based on stochastic gradient descent algorithm
Type training optimization.
Image classification model of the training based on CNN the following steps are included:
The full convolution feature extraction network of 45 layers of building, respectively to full images, pest image, crop image, Background
Feature is extracted as carrying out convolution；
Before the configuration of its convolutional layer four layers it is consistent, convolution kernel size is respectively 11,5,3,3, and characteristic pattern number is respectively
96,256,384,384；In the last layer convolution, convolution kernel size is 3, and full images characteristic pattern number is 256, pest figure
As being 96, crop image 48, background image 12；
The full link sort network of three layers of building uses the convolutional network of feature extraction to export as input, 3 layers of building
Sorter network, first layer and the second layer are dimensioned to 4096, and the size of third layer classification layer is the number 82 of pest species；
For the output of the last layer, the softmax operation of 82 values is carried out, learning objective loss function is defined as follows:
WhereinF (k) is the value on optimal one layer of kth position,For input picture mark evil
The not corresponding position of insects, K are the number for dividing classification, are 82 here；
Definition based on network and objective function inputs training sample data, carries out mould based on stochastic gradient descent algorithm
Type training optimization.
Beneficial effect
A kind of pest image classification method based on classification prediction convolutional neural networks of the invention, compared with prior art
Using classification prediction frame, the segmentation result of first forecast image carries out final classification then in conjunction with general image jointly
Prediction.Method of the present invention possesses feature representation more abundant than the prior art and noise removal capability, according to segmented image
Level forecasts segmentation is carried out, pest information is paid close attention to, reduces noise jamming, provide more abundant for prediction and more has needle
To the feature of property.
Detailed description of the invention
Fig. 1 is method precedence diagram of the invention.
Specific embodiment
The effect of to make to structure feature of the invention and being reached, has a better understanding and awareness, to preferable
Examples and drawings cooperation detailed description, is described as follows:
As shown in Figure 1, a kind of pest image classification method based on classification prediction convolutional neural networks of the present invention,
The following steps are included:
The first step is collected training image and pre-processes.Several width images are collected as training image, all training
Image carries out size normalization processing, is processed into 256 × 256 pixels, obtains several training samples.
Second step is labeled image sample data.Samples pictures content is manually marked, image point is marked out
Cut edge circle, classification and pest species divide the image into pest, crop, background three classes, and combined training sample is as training sample
Data acquisition system.Here, image is divided into pest, crop, background three classes, and combine subsequent mark collectively as model training
Data.
Third step, disaggregated model of the training based on classification prediction convolutional neural networks.Training is based on classification prediction convolution mind
Disaggregated model through network completes the training of neural network classification model using training sample as input.It includes following step
It is rapid:
(1) image segmentation network model of the training based on FCNN.Image Segmentation Model is in entire classification prediction network
The first order after being split prediction to input picture, could pass through prediction result knot so needing first to complete the training of the model
It closes original input picture and obtains each input in the sorter network of the second level, such as pest image, crop image, background image.
The setting number of plies is 7 layers of full convolutional network layer, inputs training sample, uses disease pest, crop, the background point manually marked
Result is cut as study mark, the training of network model is carried out using deep learning frame Caffe, is exported as that can scheme to pest
As the full convolutional neural networks model being split.
A, 7 layers of full convolutional network structure are constructed, the size of every layer of convolutional network is as follows:
First layer convolution kernel size is 11x11, and characteristic pattern number is 96；Second layer convolution kernel is 5x5, and characteristic pattern number is
256；Third and fourth, five layers of convolution kernel be 3x3, characteristic pattern number is respectively 384,384,256；Six, the seven layers of convolution kernel are
1x1, characteristic pattern number are respectively 512,3.
B, the output size of the last one convolutional layer is 27x27x3, and wherein 27x27 is two-dimensional image space size, and 3 be point
The number of class is cut, it is to be divided into 3 classes that the reason of the last layer is set as 3, which is pest image segmentation, it is of course possible to be set according to demand
It is set to more classes.3 target values pest, crop according to belonging to each of which position receptive field on any position 27X27 is also
It is background to determine, if including more than one classification in receptive field, using accounting for that classification of most number of pixels as mark
Note.
C, in order to finally obtain classification belonging to each position, that is, classification prediction result is obtained, for the every of 27X27
A position, carries out softmax operation in 3 classifications, and final loss function is that the softmax on the position 27X27 loses it
And L, calculation formula are as follows:
Wherein Divide classification for the mark on position (h, w), M is segmentation
The number of classification is equal to 3.
D, the definition based on network and objective function inputs training sample data, by traditional stochastic gradient descent algorithm
Carry out model training optimization.
(2) image classification model of the training based on CNN.Currently, the highest level of image classification is the depth based on CNN
Algorithm is practised, therefore also uses CNN network here, and suitably optimized according to current task.Specific network design and training
Process is as follows:
It is 8 layers of convolution sorter network that the number of plies, which is arranged, and the pest that obtains after being divided according to trained parted pattern makees
Object, three width image of background, in addition undivided general image, four width images as input, classify by the pest manually marked in total
As a result as study mark, the training of network model is carried out using deep learning frame Caffe, exports image classification model.
A, the full convolution feature extraction network for constructing 45 layers, respectively to full images, pest image, crop image, background
Image carries out convolution and extracts feature.Here independent characteristic extraction operation is carried out to four width images respectively with 4 convolutional networks, be arranged
5 layers of convolution is consistent with the sorter network of current optimal level, and from the experimental results, and certain 5 layers of convolution is in efficiency
It can be optimal in the balance of effect；
Before the configuration of its convolutional layer four layers it is consistent, convolution kernel size is respectively 11,5,3,3, and characteristic pattern number is respectively
96,256,384,384；In the last layer convolution, convolution kernel size is 3, and full images characteristic pattern number is 256, pest figure
As being 96, crop image 48, background image 12.
B, the full link sort network for constructing three layers uses the convolutional network of feature extraction to export as input, constructs 3 layers
Sorter network, first layer and the second layer be dimensioned to 4096, and the size of third layer classification layer is the number of pest species
82。
C, for the output of the last layer, the softmax operation of 82 values is carried out, learning objective loss function defines such as
Under:
WhereinF (k) is the value on optimal one layer of kth position,For input picture mark evil
The not corresponding position of insects, K are the number for dividing classification, are 82 here.
D, the definition based on network and objective function inputs training sample data, is carried out based on stochastic gradient descent algorithm
Model training optimization.
4th step, pre-processes testing image.Pest image to be measured is normalized by 256 × 256 pixels
Processing, obtains new testing image.
5th step carries out pest image classification based on hierarchy model automatically.Using classification prediction frame will treated to
Altimetric image inputs in trained classification prediction convolutional neural networks model, carries out the automatic identification of pest image type.Its
Specific step is as follows:
(1) segmentation result of first forecast image, testing image is input to and is trained in FCNN parted pattern, obtains image
Middle pest, crop, background segmentation prediction result.
(2) original image is split according to segmentation prediction result, obtains three width new images: pest image, crop map
Picture, background image.
(3) general image is combined to carry out final classification prediction jointly, the pest image that testing image and segmentation are obtained,
Totally four width images are input in the disaggregated model based on CNN of training completion as input for crop image, background image, prediction
The maximum classification of possibility is obtained to export as classification results.
The basic principles, main features and advantages of the present invention have been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and what is described in the above embodiment and the description is only the present invention
Principle, various changes and improvements may be made to the invention without departing from the spirit and scope of the present invention, these variation and
Improvement is both fallen in the range of claimed invention.The present invention claims protection scope by appended claims and its
Equivalent defines.
Claims (3)
1. a kind of pest image classification method based on classification prediction convolutional neural networks, which comprises the following steps:
11) training image is collected and is pre-processed, collect several width images as training image, all training images into
Row size normalization processing, is processed into 256 × 256 pixels, obtains several training samples；
12) image sample data is labeled, samples pictures content is manually marked, mark out image segmentation boundary,
Classification and pest species divide the image into pest, crop, background three classes, and combined training sample is as training sample data collection
It closes；
13) disaggregated model of the training based on classification prediction convolutional neural networks, training is based on classification prediction convolutional neural networks
Disaggregated model completes the training of neural network classification model using training sample as input；The training is based on classification prediction
The disaggregated models of convolutional neural networks the following steps are included:
131) image segmentation network model of the training based on FCNN；
The full convolutional network layer that the number of plies is 7 layers is set, inputs training sample, uses disease pest, crop, the background segment manually marked
As a result as study mark, the training of network model is carried out using deep learning frame Caffe, is exported as can be to pest image
The full convolutional neural networks model being split；
Image segmentation network model of the training based on FCNN the following steps are included:
1311) 7 layers of full convolutional network structure are constructed, the size of every layer of convolutional network is as follows:
First layer convolution kernel size is 11 × 11, and characteristic pattern number is 96；Second layer convolution kernel is 5x5, and characteristic pattern number is
256；Third and fourth, five layers of convolution kernel be 3 × 3, characteristic pattern number is respectively 384,384,256；Six, the seven layers of convolution kernel are equal
It is 1 × 1, characteristic pattern number is respectively 512,3；
1312) output size of the last one convolutional layer is 27 × 27 × 3, wherein 27 × 27 be two-dimensional image space size, 3 are
Divide the number of class, 3 target values pest, crop according to belonging to each of which position receptive field on any 27 × 27 position is also
It is background to determine, if including more than one classification in receptive field, using accounting for that classification of most number of pixels as mark
Note；
1313) be directed to 27 × 27 each position, carry out softmax operation in 3 classifications, final loss function be 27 ×
The sum of softmax loss on 27 positions L, calculation formula are as follows:
Wherein Divide classification for the mark on position (h, w), M is segmentation classification
Number, be equal to 3；
1314) definition based on network and objective function inputs training sample data, carries out mould based on stochastic gradient descent algorithm
Type training optimization；
132) image classification model of the training based on CNN,
It is 8 layers of convolution sorter network that the number of plies, which is arranged, the pest that is obtained after being divided according to trained parted pattern, crop,
Three width image of background, in addition undivided general image, four width images are as input in total, the pest classification results manually marked
It is marked as study, the training of network model is carried out using deep learning frame Caffe, export image classification model；
14) testing image is pre-processed, pest image to be measured is normalized by 256 × 256 pixels, is obtained
New testing image；
15) pest image classification is carried out based on hierarchy model automatically, testing image inputs trained classification by treated
It predicts in convolutional neural networks model, carries out the automatic identification of pest image type.
2. a kind of pest image classification method based on classification prediction convolutional neural networks according to claim 1, special
Sign is, it is described pest image classification carried out automatically based on hierarchy model the following steps are included:
21) testing image is input to and is trained in FCNN parted pattern, it is pre- to obtain pest, crop, the segmentation of background in image
Survey result；
22) original image is split according to segmentation prediction result, obtains three width new images: pest image, crop image, back
Scape image；
23) totally four width images are defeated as input for pest image, crop image, the background image obtained testing image and segmentation
Enter in the disaggregated model based on CNN completed to training, prediction obtains the maximum classification of possibility and exports as classification results.
3. a kind of pest image classification method based on classification prediction convolutional neural networks according to claim 1, special
Sign is, image classification model of the training based on CNN the following steps are included:
31) the full convolution feature extraction network for constructing 45 layers, respectively to full images, pest image, crop image, background image
It carries out convolution and extracts feature；
Before the configuration of its convolutional layer four layers it is consistent, convolution kernel size is respectively 11,5,3,3, characteristic pattern number is respectively 96,
256,384,384；In the last layer convolution, convolution kernel size is 3, and full images characteristic pattern number is 256, and pest image is
96, crop image 48, background image 12；
32) the full link sort network for constructing three layers uses the convolutional network of feature extraction to export as input, 3 layers of building
Sorter network, first layer and the second layer are dimensioned to 4096, and the size of third layer classification layer is the number 82 of pest species；
33) it is directed to the output of the last layer, carries out the softmax operation of 82 values, learning objective loss function is defined as follows:
WhereinF (k) is the value on optimal one layer of kth position,Pest type is marked for input picture
Not corresponding position, K are the number for dividing classification, are 82 here；
34) definition based on network and objective function inputs training sample data, carries out model based on stochastic gradient descent algorithm
Training optimization.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710103514.3A CN107016405B (en) | 2017-02-24 | 2017-02-24 | A kind of pest image classification method based on classification prediction convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710103514.3A CN107016405B (en) | 2017-02-24 | 2017-02-24 | A kind of pest image classification method based on classification prediction convolutional neural networks |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107016405A CN107016405A (en) | 2017-08-04 |
| CN107016405B true CN107016405B (en) | 2019-08-30 |
Family
ID=59440466
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710103514.3A Active CN107016405B (en) | 2017-02-24 | 2017-02-24 | A kind of pest image classification method based on classification prediction convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107016405B (en) |
Families Citing this family (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107665355B (en) * | 2017-09-27 | 2020-09-29 | 重庆邮电大学 | Agricultural pest detection method based on regional convolutional neural network |
| CN107808116B (en) * | 2017-09-28 | 2020-05-05 | 中国科学院合肥物质科学研究院 | Wheat and wheat spider detection method based on deep multilayer feature fusion learning |
| CN107578300A (en) * | 2017-10-24 | 2018-01-12 | 济南浪潮高新科技投资发展有限公司 | A kind of elevator card launches the method and device that work is audited automatically |
| CN107666612A (en) * | 2017-10-31 | 2018-02-06 | 中国科学技术大学 | Block content categorizing method based on convolutional neural networks |
| CN107966447B (en) * | 2017-11-14 | 2019-12-17 | 浙江大学 | workpiece surface defect detection method based on convolutional neural network |
| CN108229543A (en) * | 2017-12-22 | 2018-06-29 | 中国科学院深圳先进技术研究院 | Image classification design methods and device |
| CN108376399A (en) * | 2018-02-11 | 2018-08-07 | 甘肃省电力公司风电技术中心 | The sand and dust influence degree analysis method monitored in real time based on photovoltaic panel |
| CN108537108A (en) * | 2018-02-12 | 2018-09-14 | 中国农业大学 | A kind of Asiatic migrotory locust age recognition methods and device |
| CN108541683A (en) * | 2018-04-18 | 2018-09-18 | 济南浪潮高新科技投资发展有限公司 | A kind of unmanned plane pesticide spraying system based on convolutional neural networks chip |
| CN108765449B (en) * | 2018-05-16 | 2022-04-26 | 南京信息工程大学 | Image background segmentation and identification method based on convolutional neural network |
| CN109087281A (en) * | 2018-07-02 | 2018-12-25 | 北京百度网讯科技有限公司 | Display screen peripheral circuit detection method, device, electronic equipment and storage medium |
| CN109145770B (en) * | 2018-08-01 | 2022-07-15 | 中国科学院合肥物质科学研究院 | Automatic wheat spider counting method based on combination of multi-scale feature fusion network and positioning model |
| CN109558787A (en) * | 2018-09-28 | 2019-04-02 | 浙江农林大学 | A kind of Bamboo insect pests recognition methods based on convolutional neural networks model |
| CN110309841A (en) * | 2018-09-28 | 2019-10-08 | 浙江农林大学 | A kind of hickory nut common insect pests recognition methods based on deep learning |
| CN109389607A (en) * | 2018-10-12 | 2019-02-26 | 上海鹰觉科技有限公司 | Ship Target dividing method, system and medium based on full convolutional neural networks |
| CN111178120B (en) * | 2018-12-25 | 2023-04-21 | 中国科学院合肥物质科学研究院 | Pest image detection method based on crop identification cascading technology |
| CN111178121B (en) * | 2018-12-25 | 2023-04-07 | 中国科学院合肥物质科学研究院 | Pest image positioning and identifying method based on spatial feature and depth feature enhancement technology |
| CN109754011B (en) * | 2018-12-29 | 2019-11-12 | 北京中科寒武纪科技有限公司 | Data processing method, device and Related product based on Caffe |
| CN109949283B (en) * | 2019-03-12 | 2023-05-26 | 天津瑟威兰斯科技有限公司 | Method and system for identifying insect species and activity based on convolutional neural network |
| CN110009702B (en) * | 2019-04-16 | 2023-08-04 | 聊城大学 | Fall webworm larva screen image positioning method for intelligent spraying robot |
| CN110969166A (en) * | 2019-12-04 | 2020-04-07 | 国网智能科技股份有限公司 | Small target identification method and system in inspection scene |
| CN111368873A (en) * | 2020-01-15 | 2020-07-03 | 江西理工大学 | Convolutional neural network model based on depth fusion and histological image classification method thereof |
| CN111709932B (en) * | 2020-06-17 | 2023-04-07 | 广州麦仑信息科技有限公司 | Method for detecting paper defects from coarse to fine |
| CN112488244B (en) * | 2020-12-22 | 2022-09-09 | 中国科学院合肥物质科学研究院 | Dense distribution small target pest automatic counting method using thermodynamic diagram to carry out point marking mode |
| CN113239947B (en) * | 2021-03-10 | 2022-09-23 | 安徽省农业科学院农业经济与信息研究所 | Pest image classification method based on fine-grained classification technology |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104067314A (en) * | 2014-05-23 | 2014-09-24 | 中国科学院自动化研究所 | Human-shaped image segmentation method |
| CN104850836A (en) * | 2015-05-15 | 2015-08-19 | 浙江大学 | Automatic insect image identification method based on depth convolutional neural network |
| CN104899255A (en) * | 2015-05-15 | 2015-09-09 | 浙江大学 | Image database establishing method suitable for training deep convolution neural network |
| CN106295139A (en) * | 2016-07-29 | 2017-01-04 | 姹ゅ钩 | A kind of tongue body autodiagnosis health cloud service system based on degree of depth convolutional neural networks |
| CN106372390A (en) * | 2016-08-25 | 2017-02-01 | 姹ゅ钩 | Deep convolutional neural network-based lung cancer preventing self-service health cloud service system |
-
2017
- 2017-02-24 CN CN201710103514.3A patent/CN107016405B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104067314A (en) * | 2014-05-23 | 2014-09-24 | 中国科学院自动化研究所 | Human-shaped image segmentation method |
| CN104850836A (en) * | 2015-05-15 | 2015-08-19 | 浙江大学 | Automatic insect image identification method based on depth convolutional neural network |
| CN104899255A (en) * | 2015-05-15 | 2015-09-09 | 浙江大学 | Image database establishing method suitable for training deep convolution neural network |
| CN106295139A (en) * | 2016-07-29 | 2017-01-04 | 姹ゅ钩 | A kind of tongue body autodiagnosis health cloud service system based on degree of depth convolutional neural networks |
| CN106372390A (en) * | 2016-08-25 | 2017-02-01 | 姹ゅ钩 | Deep convolutional neural network-based lung cancer preventing self-service health cloud service system |
Non-Patent Citations (1)
| Title |
|---|
| Intelligent alerting for fruit-melon lesion image based on momentum deep learning;Wenxue Tan，et al.;《Multimedia Tools and Applications》;20161231;第75卷(第24期);page16741-16761 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107016405A (en) | 2017-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107016405B (en) | A kind of pest image classification method based on classification prediction convolutional neural networks | |
| Song et al. | Kiwifruit detection in field images using Faster R-CNN with VGG16 | |
| CN110263705A (en) | Towards two phase of remote sensing technology field high-resolution remote sensing image change detecting method | |
| CN106951836B (en) | crop coverage extraction method based on prior threshold optimization convolutional neural network | |
| CN109255334A (en) | Remote sensing image terrain classification method based on deep learning semantic segmentation network | |
| CN113392775B (en) | Sugarcane seedling automatic identification and counting method based on deep neural network | |
| CN110717903A (en) | Method for detecting crop diseases by using computer vision technology | |
| CN108921201B (en) | Dam defect identification and classification method based on feature combination with CNN | |
| CN111179216B (en) | Crop disease identification method based on image processing and convolutional neural network | |
| CN107392130A (en) | Classification of Multispectral Images method based on threshold adaptive and convolutional neural networks | |
| CN107832797B (en) | Multispectral image classification method based on depth fusion residual error network | |
| CN111400536B (en) | Low-cost tomato leaf disease identification method based on lightweight deep neural network | |
| CN112686902B (en) | Two-stage calculation method for brain glioma identification and segmentation in nuclear magnetic resonance image | |
| CN108154195A (en) | Tealeaves recognition methods and the tealeaves sorting equipment using this method | |
| CN105913090B (en) | SAR image objective classification method based on SDAE-SVM | |
| Lv et al. | A visual identification method for the apple growth forms in the orchard | |
| CN111178177A (en) | Cucumber disease identification method based on convolutional neural network | |
| CN110503140A (en) | Classification method based on depth migration study and neighborhood noise reduction | |
| CN109242826A (en) | Mobile device end label shape object radical method of counting and system based on target detection | |
| CN110147725A (en) | A kind of high spectrum image feature extracting method for protecting projection based on orthogonal index office | |
| CN114140665A (en) | Dense small target detection method based on improved YOLOv5 | |
| Yang et al. | Instance segmentation and classification method for plant leaf images based on ISC-MRCNN and APS-DCCNN | |
| CN114170511A (en) | Pavement crack disease identification method based on Cascade RCNN | |
| CN114693616A (en) | Rice disease detection method, equipment and medium based on improved target detection model and convolutional neural network | |
| CN114627411A (en) | Crop growth period identification method based on parallel detection under computer vision |
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 |