CN107392224A - A kind of crop disease recognizer based on triple channel convolutional neural networks - Google Patents
A kind of crop disease recognizer based on triple channel convolutional neural networks Download PDFInfo
- Publication number
- CN107392224A CN107392224A CN201710438282.7A CN201710438282A CN107392224A CN 107392224 A CN107392224 A CN 107392224A CN 201710438282 A CN201710438282 A CN 201710438282A CN 107392224 A CN107392224 A CN 107392224A
- Authority
- CN
- China
- Prior art keywords
- layer
- neural networks
- image
- convolutional neural
- pond
- 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
Classifications
-
- 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
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of crop disease recognizer based on triple channel convolutional neural networks, its technical characteristics is：Carry out image preprocessing；RGB color disease leaf image is decomposed into tri- passages of R, G and B, the input as convolutional neural networks；Build triple channel convolutional neural networks model；Initialize weight and offset parameter；Retraining is carried out to the convolutional neural networks trained using tranining database, so that the model parameter trained can be fitted current database, the more abstract feature of extraction image, carries out image recognition.The present invention is reasonable in design, has carried out experimental verification on apple and cucumber disease leaf image database, can describe cucumber disease leaf image feature exactly, has made the characteristic value of resulting object pixel more representative, improve discrimination；The characteristic value that the present invention obtains describes more comprehensive to the characteristic information of cucumber disease leaf image, therefore can carry out effective cucumber disease leaf image identification.
Description
Technical field
The invention belongs to image identification technical field, especially a kind of crop disease based on triple channel convolutional neural networks
Recognizer.
Background technology
Crop disease is a kind of serious natural calamity in agricultural production, and restricts crop yield, quality and wisdom
One of principal element of agricultural sustainable development.China is a large agricultural country, and crop disease generation is than more serious state
Family, crop disease Study of recognition are particularly important.Crop disease is only identified in time, could effectively be carried out disease control, be entered
And mitigate infringement of the disease to crop, reduce Pesticide use amount.The generation of most of crop disease often shows blade first
On, disease often causes blade scab occur, and different types of disease causes the different colours of blade, shape and texture
Scab, as shown in Fig. 2 a to Fig. 2 d.Therefore, based on disease leaf image crop disease recognition methods research be always plant protection,
One important research direction of the various fields such as image procossing, computer visual angle and pattern-recognition, occur much being based on leaf
The crop disease recognition methods of picture.Wherein, feature extraction is a committed step of these methods, the feature extracted
Quality directly affects the accuracy of identification of disease recognition algorithm.Due to the complicated variety of disease leaf image, and the face of scab
Color, shape and texture are constantly changing over time so that people have devised hundreds of various characteristic of division, be used for
Improve the discrimination and robustness of crop disease recognition methods.
Although using existing crop disease recognition methods, kind more than 100 can be extracted from a width disease leaf image
Different types of feature, but be difficult to determine that contribution of which feature to disease recognition is maximum.Particularly, some features are to some
Crop disease identification is optimal, but the identification of other crop diseases is not necessarily optimal.Therefore, existing many crop diseases
Evil recognition methods is not met by the needs of actual crop disease automatic recognition system.
Convolutional neural networks (CNNs) are a modifications of multi-layer perception (MLP) (MLP), it be from biological concept develop,
And developed from traditional neural network (NN).Traditional NN includes three layers：Input layer, hidden layer and output layer；And CNNs
Include multiple hidden layers (including multiple convolutional layers and pond layer).Existing CNNs is by successively improving the feature of image so that special
Sign space constantly changes, thus, it is possible to analyze more complicated image classification and identification problem.CNNs local sensing, weights are common
Enjoy, pond and the features such as sandwich construction, can not only reduce memory data output and training parameter number, and can greatly carry
The performance of high algorithm.The basic structure of CNNs based on LeNet frameworks is as shown in Figure 3.Therefore, how CNNs is applied to crop
Disease recognition with meet the needs of actual crop disease automatic recognition system be at present it is in the urgent need to address the problem of.
The content of the invention
It is overcome the deficiencies in the prior art the mesh of the present invention, proposes that one kind is reasonable in design, discrimination is high and identification speed
The fast crop disease recognizer based on triple channel convolutional neural networks of degree.
The present invention solves its technical problem and takes following technical scheme to realize：
A kind of crop disease recognizer based on triple channel convolutional neural networks, comprises the following steps：
Step 1：Carry out image preprocessing：The each image that training disease leaf image of different sizes is concentrated first is adjusted
Whole is unified dimension, then carries out contrast normalized to image using contrast self-adapting histogram equilibrium method so that
Displacement, swing and the change of scale of image are uniformly distributed in particular range；
Step 2：RGB color disease leaf image is decomposed into tri- passages of R, G and B, as the defeated of convolutional neural networks
Enter；
Step 3：Build triple channel convolutional neural networks model；
Step 4：Initialize weight and offset parameter；
Step 5：Retraining is carried out to the convolutional neural networks trained using tranining database, so that the mould trained
Shape parameter can be fitted current database, the more abstract feature of extraction image, carry out image recognition.
Further, the concrete methods of realizing of the step 3 is：
A pond layer is added between the first two convolutional layer of general convolutional neural networks, its basic structure is input
Layer I → convolutional layer C1 → pond layer P1 → convolutional layer C2 → pond layer P2 → convolutional layer C3 → pond layer P3 → full articulamentum F →
Output layer；
If convolution kernel size is k × k, the pond window size of maximum pond method is p × p, full articulamentum neuron number
For 1000, output layer has C neuron output C kind crop disease species；
In convolutional layer C1, k × k receptive field that each neuron is specified with input disease leaf image is rolled up
Product, obtain multiple different characteristic patterns and export to give pond layer P1；
In the layer P1 of pond, maximum pond down-sampling is carried out with p × p region to convolutional layer C1 characteristic patterns, does not change feature
The number of figure；
In convolutional layer C2, convolution is carried out again to the characteristic pattern obtained after Chi HuacengP1Chiization；
In the layer P2 of pond, maximum pond down-sampling is carried out in the characteristic pattern that p × p region obtains to convolutional layer C2；
In convolutional layer C3, convolution is further carried out to the obtained characteristic patterns of pond layer P2；
In the layer P3 of pond, then the characteristic pattern obtained in p × p region to convolutional layer C3 carries out maximum pond down-sampling；
Two full articulamentum F are connected entirely comprising 1000 neurons with pond layer P3；
A grader or a monolayer neural networks are selected in output layer, every one kind is assigned to by calculating input sample
Other probability carries out image recognition；Pass through adjusting parameter in the training process so that the correct label of disease leaf image is corresponding
Maximum probability；Softmax graders are selected in output layer, when there is C kind diseases leaf image to need to be classified, in output layer
There is C neuron, output C is less than 1 positive, represents the probability of each image generic to be tested.
Further, the implementation method of the step 4 is：
In terms of the following two using different small random numbers to convolutional neural networks in all parameter carry out it is random just
Beginningization：(1) small random number is taken, ensures that convolutional neural networks will not enter saturation state because weights are excessive, and causes to train
Failure；(2) different random numbers is taken, convolutional neural networks normally learning training can be ensured.
Further, it is [- 0.5,0.5] or [- 1,1] to initialize weights and the scope of offset parameter.
The advantages and positive effects of the present invention are：
1st, the present invention carries out crop disease identification by triple channel convolutional neural networks (CNNs), efficiently solves crop
Scab segmentation and the feature extraction problem of disease blade, can be extracted optimal, more abstract from disease leaf image automatically
Substantive characteristics and carry out disease recognition.Compared with traditional crop disease recognition methods, its maximum difference is, the party
Method is the automatic learning classification feature from disease leaf image, rather than the feature by priori extraction hand-designed, institute
So that this method can overcome the blindness of traditional characteristic extracting method and take the deficiencies of big.Particularly, in disease recognition process
In, this method need not carry out complicated scab segmentation and feature extraction to disease leaf image, but directly input colored disease
Evil leaf image carries out disease recognition, therefore, has the characteristics of discrimination height and fast recognition speed.
2nd, the present invention is reasonable in design, and experimental verification has been carried out on apple and cucumber disease leaf image database, can
Cucumber disease leaf image feature is described exactly, is made the characteristic value of resulting object pixel more representative, is improved knowledge
Not rate；The characteristic value that the present invention obtains describes more comprehensive to the characteristic information of cucumber disease leaf image, therefore can enter
The effective cucumber disease leaf image identification of row.
Brief description of the drawings
Fig. 1 is the triple channel CNNs of present invention basic structure；
Fig. 2 a to 2d are the Apple Leaves image used in the present invention, wherein, Fig. 2 a are normal disease-free blade, and Fig. 2 b are rust
Sick blade, Fig. 2 c are head blight blade, and Fig. 2 d are mosaic disease blade；
Fig. 3 is CNNs structural representation.
Embodiment
The embodiment of the present invention is further described below in conjunction with accompanying drawing.
One kind is based on the crop disease recognizer of triple channel convolutional neural networks (CNNs), comprises the following steps：
Step 1：Simple image pre-processes：The each image that training disease leaf image of different sizes is concentrated is adjusted to
Unified dimension, contrast normalized is then carried out to image using contrast self-adapting histogram equilibrium method (CLAHE), made
Displacement, swing and the change of scale for obtaining image are uniformly distributed in particular range.
Step 2：RGB color disease leaf image is decomposed into tri- passages of R, G and B, the input as CNNs.
Step 3：Build triple channel convolutional neural networks model.
The present invention designs triple channel CNNs models as shown in Figure 1, and wherein CNNs framework and Fig. 2 is essentially identical.In order to
The time consumption for training of network is reduced, we add a pond layer between the first two convolutional layer.Basic structure be input layer I →
Convolutional layer C1 → pond layer P1 → convolutional layer C2 → pond layer P2 → convolutional layer C3 → pond layer P3 → full articulamentum F → output layer
O。
It is assumed that convolution kernel size is k × k, the pond window size of maximum pond method is p × p, and full articulamentum neuron is individual
Number is 1000 or so, and output layer has C neuron output C kind crop disease species.
In C1, k × k receptive field that each neuron is specified with input disease leaf image carries out convolution, obtains
Multiple different characteristic patterns are exported to P1；
In P1, maximum pond down-sampling is carried out with p × p region to C1 characteristic patterns, does not change the number of characteristic pattern；
In C2, convolution is carried out again to the characteristic pattern obtained behind P1 ponds；
In P2, maximum pond down-sampling is carried out in the characteristic pattern that p × p region obtains to C2；
In C3, the characteristic pattern obtained to P2 further carries out convolution；
In P3, then the characteristic pattern obtained in p × p region to C3 carries out maximum pond down-sampling；
Two full articulamentum F are connected entirely comprising 1000 neurons with P3；
A grader or a monolayer neural networks are selected in output layer, every one kind is assigned to by calculating input sample
Other probability carries out image recognition.Pass through adjusting parameter in the training process so that the correct label of disease leaf image is corresponding
Maximum probability.Because the accuracy rate of Softmax graders is high, Serial Distribution Processing ability is strong, can effectively handle non-linear ask
Topic, and amount of calculation is smaller, training speed is very fast, so the output layer selection Softmax graders of model of the present invention.There are C kinds sick
Evil leaf image needs to be classified, so having C neuron in output layer, output C is less than 1 positive, represented each
The probability of image generic to be tested.
In MATLAB Deep learning Toolbox, a characteristic pattern of convolutional layer and all features on upper strata
Figure all associates, and to different convolution kernels after all characteristic patterns of preceding layer carry out convolution and corresponding element adds up, then adds one
Biasing, then asks sigmod functions to be worth to a characteristic pattern.Moreover, the number of the characteristic pattern of convolutional layer is initialized in CNNs
Shi Zhiding, its size are determined by the size of convolution kernel and last layer input feature vector figure.Pond layer is the feature to a upper convolutional layer
One down-sampling of figure, the sample mode of use are to carry out aggregate statistics to the neighbor cell domain of last layer characteristic pattern, are typically taken
The maximum or average value of zonule.If the size for assuming characteristic pattern in a upper convolutional layer is n × n, the size of convolution kernel is k
× k, then the size of the characteristic pattern of the convolutional layer is (n-k+1) × (n-k+1), and pond layer does not change the size of characteristic pattern.Thus
The size of each convolutional layer for the CNNs models that the present invention designs and the characteristic pattern of pond layer can be calculated.
Step 4：Initialize weight and offset parameter.Specific method is as follows：
Start to train before CNNs, it is necessary to initialize weight and offset parameter.Typically examined in terms of following two
Consider and random initializtion is carried out to parameter all in CNNs using some different small random numbers：(a) small random number is taken, can
Ensure that CNNs will not enter saturation state because weights are excessive, and cause failure to train；(b) different random numbers is taken, can
Ensure CNNs normally learning trainings.If initializing weights using identical, big numerical value, CNNs may not have study energy
Power.The weights of random initializtion and the scope of biasing are typically taken as [- 0.5,0.5] or [- 1,1].
Step 5：Retraining (fine setting) is carried out to the CNNs trained using tranining database, so that the model trained
Parameter can be fitted current database, the more abstract feature of extraction image, carry out image recognition.
The effect of the present invention is verified below by experiment.Experimental result is as shown in table 1 and table 2：
Table 1
Table 2
Table 1 is the present invention and the knowledge of TDDS, SVM, TSVM, DNN in Plantvillage Apple Leaves image data bases
Not rate comparing result.Table 2 is the present invention and the discrimination of TDDS, SVM, TSVM, DNN on cucumber disease leaf image database
Comparing result.From table 1 and table 2 as can be seen that the recognition effect of the present invention is best, DNN recognition effect takes second place.Its reason is
CNNs and DNN can obtain the more abstract characteristic of division in the high-rise expression of disease leaf image automatically.Although other three kinds
The discrimination of method is all bigger, but they are the recognition results on the scab image being partitioned into, and the inventive method and DNN
The recognition result of method is in the original image only recognition result on the leaf image after simple normalization.By further real
Test and learn, the discrimination of first three methods TDSS, SVM and TSVM on the original image of no segmentation is both less than 60%.
It is emphasized that embodiment of the present invention is illustrative, rather than it is limited, therefore present invention bag
Include and be not limited to embodiment described in embodiment, it is every by those skilled in the art's technique according to the invention scheme
The other embodiment drawn, also belongs to the scope of protection of the invention.
Claims (4)
1. a kind of crop disease recognizer based on triple channel convolutional neural networks, it is characterised in that comprise the following steps：
Step 1：Carry out image preprocessing：The each image that training disease leaf image of different sizes is concentrated is adjusted to first
Unified dimension, contrast normalized is then carried out to image using contrast self-adapting histogram equilibrium method so that image
Displacement, swing and change of scale be uniformly distributed in particular range；
Step 2：RGB color disease leaf image is decomposed into tri- passages of R, G and B, the input as convolutional neural networks；
Step 3：Build triple channel convolutional neural networks model；
Step 4：Initialize weight and offset parameter；
Step 5：Retraining is carried out to the convolutional neural networks trained using tranining database, so that the model ginseng trained
Number can be fitted current database, the more abstract feature of extraction image, carry out image recognition.
2. a kind of crop disease recognizer based on triple channel convolutional neural networks according to claim 1, its feature
It is：The concrete methods of realizing of the step 3 is：
A pond layer is added between the first two convolutional layer of general convolutional neural networks, its basic structure is input layer I
→ convolutional layer C1 → pond layer P1 → convolutional layer C2 → pond layer P2 → convolutional layer C3 → pond layer P3 → full articulamentum F → output
Layer；
If convolution kernel size is k × k, the pond window size of maximum pond method is p × p, and full articulamentum neuron number is
1000, output layer has C neuron output C kind crop disease species；
In convolutional layer C1, k × k receptive field that each neuron is specified with input disease leaf image carries out convolution, obtains
Exported to multiple different characteristic patterns and give pond layer P1；
In the layer P1 of pond, maximum pond down-sampling is carried out with p × p region to convolutional layer C1 characteristic patterns, does not change characteristic pattern
Number；
In convolutional layer C2, convolution is carried out again to the characteristic pattern obtained after Chi HuacengP1Chiization；
In the layer P2 of pond, maximum pond down-sampling is carried out in the characteristic pattern that p × p region obtains to convolutional layer C2；
In convolutional layer C3, convolution is further carried out to the obtained characteristic patterns of pond layer P2；
In the layer P3 of pond, then the characteristic pattern obtained in p × p region to convolutional layer C3 carries out maximum pond down-sampling；
Two full articulamentum F are connected entirely comprising 1000 neurons with pond layer P3；
A grader or a monolayer neural networks are selected in output layer, each classification is assigned to by calculating input sample
Probability carries out image recognition；Pass through adjusting parameter in the training process so that general corresponding to the correct label of disease leaf image
Rate is maximum；Softmax graders are selected in output layer, when there is C kind diseases leaf image to need to be classified, have C in output layer
Individual neuron, output C are less than 1 positive, represent the probability of each image generic to be tested.
3. a kind of crop disease recognizer based on triple channel convolutional neural networks according to claim 1, its feature
It is：The implementation method of the step 4 is：
In terms of the following two using different small random numbers to convolutional neural networks in all parameter carry out random initializtion：
(1) small random number is taken, ensures that convolutional neural networks will not enter saturation state because weights are excessive, and causes failure to train；
(2) different random numbers is taken, convolutional neural networks normally learning training can be ensured.
4. a kind of crop disease recognizer based on triple channel convolutional neural networks according to claim 1 or 3, it is special
Sign is：The scope for initializing weights and offset parameter is [- 0.5,0.5] or [- 1,1].
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710438282.7A CN107392224A (en) | 2017-06-12 | 2017-06-12 | A kind of crop disease recognizer based on triple channel convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710438282.7A CN107392224A (en) | 2017-06-12 | 2017-06-12 | A kind of crop disease recognizer based on triple channel convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107392224A true CN107392224A (en) | 2017-11-24 |
Family
ID=60332212
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710438282.7A Pending CN107392224A (en) | 2017-06-12 | 2017-06-12 | A kind of crop disease recognizer based on triple channel convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107392224A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108052894A (en) * | 2017-12-11 | 2018-05-18 | 北京飞搜科技有限公司 | More attribute recognition approaches, equipment, medium and the neutral net of a kind of target object |
| CN108304812A (en) * | 2018-02-07 | 2018-07-20 | 郑州大学西亚斯国际学院 | A kind of crop disease recognition methods based on convolutional neural networks and more video images |
| CN108509920A (en) * | 2018-04-04 | 2018-09-07 | 南京信息工程大学 | The face identification method of the multichannel combined feature selecting study of more patch based on CNN |
| CN109063319A (en) * | 2018-07-27 | 2018-12-21 | 天津大学 | A kind of analogy method of bioecosystem neural network based |
| CN109063826A (en) * | 2018-03-19 | 2018-12-21 | 重庆大学 | A kind of convolutional neural networks implementation method based on memristor |
| CN109344699A (en) * | 2018-08-22 | 2019-02-15 | 天津科技大学 | Winter jujube disease recognition method based on depth of seam division convolutional neural networks |
| CN109615010A (en) * | 2018-12-13 | 2019-04-12 | 济南大学 | Chinese traditional medicinal materials recognition method and system based on double scale convolutional neural networks |
| CN109919917A (en) * | 2019-02-21 | 2019-06-21 | 国网山东省电力公司临沂供电公司 | The foreign matter detecting method of overhead transmission line based on image procossing |
| CN110059751A (en) * | 2019-04-19 | 2019-07-26 | 南京链和科技有限公司 | A kind of tire code and tire condition recognition methods based on machine learning |
| CN110163061A (en) * | 2018-11-14 | 2019-08-23 | 腾讯科技（深圳）有限公司 | For extracting the method, apparatus, equipment and computer-readable medium of video finger print |
| CN110222696A (en) * | 2019-04-16 | 2019-09-10 | 天津科技大学 | A method of plant leaf blade disease identification is carried out using GPCNNs and ELM |
| CN110245747A (en) * | 2019-06-21 | 2019-09-17 | 华中师范大学 | Image processing method and device based on full convolutional neural networks |
| CN111832581A (en) * | 2020-09-21 | 2020-10-27 | 平安科技（深圳）有限公司 | Lung feature recognition method and device, computer equipment and storage medium |
| CN116956174A (en) * | 2019-05-13 | 2023-10-27 | 北京绪水互联科技有限公司 | Classification model for cold head state classification detection and life prediction and generation method of prediction model |
| CN116975246A (en) * | 2023-08-03 | 2023-10-31 | 深圳市博锐高科科技有限公司 | Data acquisition method, device, chip and terminal |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104850836A (en) * | 2015-05-15 | 2015-08-19 | 浙江大学 | Automatic insect image identification method based on depth convolutional neural network |
| CN106203621A (en) * | 2016-07-11 | 2016-12-07 | 姚颂 | The processor calculated for convolutional neural networks |
| CN106682704A (en) * | 2017-01-20 | 2017-05-17 | 中国科学院合肥物质科学研究院 | Method of disease image identification based on hybrid convolutional neural network fused with context information |
| CN106778785A (en) * | 2016-12-23 | 2017-05-31 | 东软集团股份有限公司 | Build the method for image characteristics extraction model and method, the device of image recognition |
-
2017
- 2017-06-12 CN CN201710438282.7A patent/CN107392224A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104850836A (en) * | 2015-05-15 | 2015-08-19 | 浙江大学 | Automatic insect image identification method based on depth convolutional neural network |
| CN106203621A (en) * | 2016-07-11 | 2016-12-07 | 姚颂 | The processor calculated for convolutional neural networks |
| CN106778785A (en) * | 2016-12-23 | 2017-05-31 | 东软集团股份有限公司 | Build the method for image characteristics extraction model and method, the device of image recognition |
| CN106682704A (en) * | 2017-01-20 | 2017-05-17 | 中国科学院合肥物质科学研究院 | Method of disease image identification based on hybrid convolutional neural network fused with context information |
Non-Patent Citations (1)
| Title |
|---|
| 张帅等: "基于分层卷积深度学习***的植物叶片识别研究", 《北京林业大学学报》 * |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108052894A (en) * | 2017-12-11 | 2018-05-18 | 北京飞搜科技有限公司 | More attribute recognition approaches, equipment, medium and the neutral net of a kind of target object |
| CN108304812A (en) * | 2018-02-07 | 2018-07-20 | 郑州大学西亚斯国际学院 | A kind of crop disease recognition methods based on convolutional neural networks and more video images |
| CN109063826A (en) * | 2018-03-19 | 2018-12-21 | 重庆大学 | A kind of convolutional neural networks implementation method based on memristor |
| CN108509920A (en) * | 2018-04-04 | 2018-09-07 | 南京信息工程大学 | The face identification method of the multichannel combined feature selecting study of more patch based on CNN |
| CN108509920B (en) * | 2018-04-04 | 2022-04-26 | 南京信息工程大学 | CNN-based face recognition method for multi-patch multi-channel joint feature selection learning |
| CN109063319B (en) * | 2018-07-27 | 2023-04-07 | 天津大学 | Simulation method of biological ecosystem based on neural network |
| CN109063319A (en) * | 2018-07-27 | 2018-12-21 | 天津大学 | A kind of analogy method of bioecosystem neural network based |
| CN109344699A (en) * | 2018-08-22 | 2019-02-15 | 天津科技大学 | Winter jujube disease recognition method based on depth of seam division convolutional neural networks |
| CN110163061A (en) * | 2018-11-14 | 2019-08-23 | 腾讯科技（深圳）有限公司 | For extracting the method, apparatus, equipment and computer-readable medium of video finger print |
| CN109615010B (en) * | 2018-12-13 | 2020-11-10 | 济南大学 | Traditional Chinese medicine material identification method and system based on double-scale convolutional neural network |
| CN109615010A (en) * | 2018-12-13 | 2019-04-12 | 济南大学 | Chinese traditional medicinal materials recognition method and system based on double scale convolutional neural networks |
| CN109919917A (en) * | 2019-02-21 | 2019-06-21 | 国网山东省电力公司临沂供电公司 | The foreign matter detecting method of overhead transmission line based on image procossing |
| CN109919917B (en) * | 2019-02-21 | 2020-10-20 | 国网山东省电力公司临沂供电公司 | Image processing-based foreign matter detection method for overhead transmission line |
| CN110222696A (en) * | 2019-04-16 | 2019-09-10 | 天津科技大学 | A method of plant leaf blade disease identification is carried out using GPCNNs and ELM |
| CN110059751A (en) * | 2019-04-19 | 2019-07-26 | 南京链和科技有限公司 | A kind of tire code and tire condition recognition methods based on machine learning |
| CN116956174A (en) * | 2019-05-13 | 2023-10-27 | 北京绪水互联科技有限公司 | Classification model for cold head state classification detection and life prediction and generation method of prediction model |
| CN110245747A (en) * | 2019-06-21 | 2019-09-17 | 华中师范大学 | Image processing method and device based on full convolutional neural networks |
| CN111832581A (en) * | 2020-09-21 | 2020-10-27 | 平安科技（深圳）有限公司 | Lung feature recognition method and device, computer equipment and storage medium |
| CN111832581B (en) * | 2020-09-21 | 2021-01-29 | 平安科技（深圳）有限公司 | Lung feature recognition method and device, computer equipment and storage medium |
| WO2022057309A1 (en) * | 2020-09-21 | 2022-03-24 | 平安科技（深圳）有限公司 | Lung feature recognition method and apparatus, computer device, and storage medium |
| CN116975246A (en) * | 2023-08-03 | 2023-10-31 | 深圳市博锐高科科技有限公司 | Data acquisition method, device, chip and terminal |
| CN116975246B (en) * | 2023-08-03 | 2024-04-26 | 深圳市博锐高科科技有限公司 | Data acquisition method, device, chip and terminal |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107392224A (en) | A kind of crop disease recognizer based on triple channel convolutional neural networks | |
| CN105069413B (en) | A kind of human posture's recognition methods based on depth convolutional neural networks | |
| CN110348376A (en) | A kind of pedestrian's real-time detection method neural network based | |
| CN108304826A (en) | Facial expression recognizing method based on convolutional neural networks | |
| CN108647742A (en) | Fast target detection method based on lightweight neural network | |
| CN104408435A (en) | Face identification method based on random pooling convolutional neural network | |
| CN107563422A (en) | A kind of polarization SAR sorting technique based on semi-supervised convolutional neural networks | |
| CN107798356A (en) | Crop leaf disease recognition method based on depth convolutional neural networks | |
| CN106203363A (en) | Human skeleton motion sequence Activity recognition method | |
| CN109489576B (en) | Contour detection method based on primary visual path calculation model | |
| CN106023065A (en) | Tensor hyperspectral image spectrum-space dimensionality reduction method based on deep convolutional neural network | |
| CN107480649A (en) | Fingerprint sweat pore extraction method based on full convolution neural network | |
| CN111696101A (en) | Light-weight solanaceae disease identification method based on SE-Inception | |
| CN109241995B (en) | Image identification method based on improved ArcFace loss function | |
| CN109785344A (en) | The remote sensing image segmentation method of binary channel residual error network based on feature recalibration | |
| CN107145893A (en) | A kind of image recognition algorithm and system based on convolution depth network | |
| CN103593652B (en) | A kind of cucumber disease recognition methods processed based on cucumber leaves symptomatic picture | |
| CN110163102A (en) | A kind of cervical cell image classification recognition methods based on convolutional neural networks | |
| CN101004791A (en) | Method for recognizing facial expression based on 2D partial least square method | |
| CN107657204A (en) | The construction method and facial expression recognizing method and system of deep layer network model | |
| CN108537777A (en) | A kind of crop disease recognition methods based on neural network | |
| CN108664994A (en) | A kind of remote sensing image processing model construction system and method | |
| Hong et al. | Automatic recognition of coal and gangue based on convolution neural network | |
| CN107516103A (en) | A kind of image classification method and system | |
| CN110378208A (en) | A kind of Activity recognition method based on depth residual error 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171124 |