CN109472774A - A kind of tongue picture picture quality detection method based on deep learning - Google Patents

A kind of tongue picture picture quality detection method based on deep learning Download PDF

Info

Publication number
CN109472774A
CN109472774A CN201811185076.0A CN201811185076A CN109472774A CN 109472774 A CN109472774 A CN 109472774A CN 201811185076 A CN201811185076 A CN 201811185076A CN 109472774 A CN109472774 A CN 109472774A
Authority
CN
China
Prior art keywords
tongue picture
picture image
tongue
layer
image
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
Application number
CN201811185076.0A
Other languages
Chinese (zh)
Inventor
许家佗
罗志宇
崔骥
屠立平
黄景斌
胡晓娟
江涛
崔龙涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai University of Traditional Chinese Medicine
Original Assignee
Shanghai University of Traditional Chinese Medicine
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai University of Traditional Chinese Medicine filed Critical Shanghai University of Traditional Chinese Medicine
Priority to CN201811185076.0A priority Critical patent/CN109472774A/en
Publication of CN109472774A publication Critical patent/CN109472774A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield 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
    • G06N3/04—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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30168—Image quality inspection

Abstract

The tongue picture picture quality detection method based on deep learning that the present invention relates to a kind of, comprising the following steps: step S1, construct tongue picture image data set, the tongue picture image data set includes training dataset and validation data set；Step S2, the tongue picture image concentrated to the tongue picture image data in step S1 pre-processes；Step S3, convolutional neural networks are constructed, is handled using the convolutional neural networks through the pretreated tongue picture image of step S2, obtains tongue picture image detection model；Step S4, tongue picture image to be detected is detected using the tongue picture image detection model that step S3 is obtained.It the advantage is that, by deep learning, improve the recognition accuracy of tongue picture image detection model；Feature extraction need not be carried out to tongue picture image, only the original photo of tongue picture picture is handled, quickly can be identified and be classified；To tongue picture image detection model iteration optimization, accuracy rate height and not the tongue picture image detection model of over-fitting are obtained.

Description

A kind of tongue picture picture quality detection method based on deep learning
Technical field
The present invention relates to technical field of image processing more particularly to a kind of tongue picture picture quality detections based on deep learning Method.
Background technique
Traditional Chinese Medicine depends on the subjective judgement of doctor to the understanding of tongue picture, human factor tongue picture is differentiated influence compared with Greatly, judge mainly by the experience of doctor, different doctors is caused to have different diagnostic results, shadow to the tongue picture of same patient Ring the successive treatment of patient.
To solve the above-mentioned problems, judged with carrying out objective and quantization to tongue picture, judged using computer technology.It is logical Normal way is to carry out feature extraction to tongue picture using computer technology, carry out with the method for machine learning to tongue picture image Detection.But this method process is cumbersome, accuracy rate is low, ineffective.
Therefore, need one kind can easy detection process, improve Detection accuracy, effective evaluation carried out to tongue picture image Quality determining method.
Summary of the invention
The purpose of the present invention is aiming at the shortcomings in the prior art, provide a kind of tongue picture picture quality based on deep learning Detection method.
To achieve the above object, the technical solution adopted by the present invention is that:
A kind of tongue picture picture quality detection method based on deep learning, which comprises the following steps:
Step S1, tongue picture image data set is constructed, the tongue picture image data set includes training dataset and verify data Collection；
Step S2, the tongue picture image concentrated to the tongue picture image data in step S1 pre-processes；
Step S3, convolutional neural networks are constructed, using the convolutional neural networks to pretreated described through step S2 Tongue picture image is handled, and tongue picture image detection model is obtained；
Step S4, tongue picture image to be detected is examined using the tongue picture image detection model that step S3 is obtained It surveys.
Preferably, in the step S2, the pretreatment includes weight compression processing, rotation processing, reversion processing, albefaction Processing and intercepting process.
Preferably, the step S2 includes:
Step S21, weight compression processing is carried out to the tongue picture image；
Step S22, weight contracting is carried out treated the tongue picture image carries out respectively rotation processing, anti-to through step S21 Turn processing, whitening processing and intercepting process.
Preferably, the heavy compression processing is to carry out weight to the tongue picture image to be reduced to 28 × 28 pixels.
Preferably, in the step S1, the image data collection includes qualified tongue picture image and unqualified tongue picture Image, wherein the quantity of the qualification tongue picture image is equal with the quantity of the unqualified tongue picture image.
Preferably, the training dataset includes the qualified tongue picture image and the unqualified tongue picture image, described Training data is concentrated, and the quantity of the qualification tongue picture image is equal with the quantity of the unqualified tongue picture image；
The validation data set includes the qualified tongue picture image and the unqualified tongue picture image, in the verify data It concentrates, the quantity of the qualification tongue picture image is equal with the quantity of the unqualified tongue picture image；
The quantity for the qualified tongue picture image that the verify data is concentrated is less than the conjunction that the training data is concentrated The quantity of lattice tongue picture image.
Preferably, the convolutional neural networks include the first input layer, the second input layer, the first convolutional layer, the first pond Layer, the second convolutional layer, the second pond layer, full articulamentum, multilayer perceptron and output layer；
First input layer is connect, for inputting the training dataset with first convolutional layer；
Second input layer is connect, for inputting the validation data set with first convolutional layer；
First convolutional layer, second convolutional layer, second pond layer, described connects first pond layer entirely Layer, the multilayer perceptron and the output layer is connect to be sequentially connected with.
Preferably, first convolutional layer is weighted processing to the tongue picture image using Gauss weight, and described First convolutional layer uses unsaturation activation primitive；
Second convolutional layer uses unsaturation activation primitive.
Preferably, the unsaturation activation primitive is ReLU function.
Preferably, the number for the convolution kernel that first convolutional layer uses is 20, and the size of the convolution kernel is 5 × 5.
Preferably, the number for the convolution kernel that second convolutional layer uses is 50, and the size of the convolution kernel is 5 × 5.
Preferably, first pond layer carries out the processing of maximum value pondization to the result that first convolutional layer exports；
Second pond layer carries out the processing of maximum value pondization to the result that second convolutional layer exports.
Preferably, the step S3 includes:
Step S31, the training dataset is trained by the convolutional neural networks, obtains the tongue picture image Detection model；
Step S32, the tongue picture image detection model that step S31 is obtained is carried out using the validation data set excellent Change.
Preferably, in the step S32, the tongue picture image detection model is iterated using error minimum method Optimization.
Preferably, the AUC value of the tongue picture image detection model is 0.98, the accuracy rate of the tongue picture image detection model It is 0.98.
The invention adopts the above technical scheme, compared with prior art, has the following technical effect that
A kind of tongue picture picture quality detection method based on deep learning of the invention improves tongue picture by deep learning The recognition accuracy of image detection model；Need not to tongue picture image carry out feature extraction, only to the original photo of tongue picture picture into Row processing, quickly can be identified and be classified；By being iterated optimization to tongue picture image detection model, one can be obtained Accuracy rate is high and the tongue picture image detection model of over-fitting, accuracy rate are not up to 0.98.
Detailed description of the invention
Fig. 1 is the work step figure of an illustrative examples of the invention.
Fig. 2 is the work step figure of the step S2 of an illustrative examples of the invention.
Fig. 3 is the work step figure of the step S3 of an illustrative examples of the invention.
Fig. 4 is the structure chart of the convolutional neural networks of an illustrative examples of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, those of ordinary skill in the art without creative labor it is obtained it is all its His embodiment, shall fall within the protection scope of the present invention.
It should be noted that in the absence of conflict, the feature in embodiment and embodiment in the present invention can phase Mutually combination.
The present invention will be further explained below with reference to the attached drawings and specific examples, but not as the limitation of the invention.
Embodiment 1
An illustrative examples of the invention, as shown in Figure 1, a kind of tongue picture picture quality detection based on deep learning Method, comprising the following steps:
Step S1, tongue picture image data set is constructed
For tongue picture image data set by a large amount of tongue picture image constructions, tongue picture image data set includes training dataset and verifying number According to collection.
Step S2, the tongue picture image concentrated to the tongue picture image data in step S1 pre-processes
All tongue picture images of tongue picture image data set are pre-processed, to reduce interference of the noise to tongue picture image.
Step S3, convolutional neural networks are constructed, using convolutional neural networks to through the pretreated tongue picture image of step S2 It is handled, obtains tongue picture image detection model
By convolutional neural networks, tongue picture image detection model is trained and is optimized using tongue picture image data set.
Step S4, the tongue picture image detection model obtained using step S3 detects tongue picture image to be detected
It is detected using the tongue picture iconic model being iterated after optimizing that step S3 is obtained, to tongue picture figure to be detected As carrying out quickly identification classification.
Further, in step sl:
Tongue picture image includes qualified tongue picture image and unqualified tongue picture image, wherein the quantity of qualified tongue picture image is n, The quantity of unqualified tongue picture image is n, i.e., the quantity of qualified tongue picture image is equal with the quantity of unqualified tongue picture image.
Qualified tongue picture image be without atomization, without it is under-exposed, without overexposure, without incomplete tongue picture image.
Unqualified tongue picture image is the tongue picture at least with any defect in atomization, under-exposure, overexposure, incompleteness Image.
Training dataset is concentrated by qualified tongue picture image and unqualified tongue picture image construction, and in training data, qualified tongue As the quantity of image is equal with the quantity of unqualified tongue picture image.
Validation data set is concentrated equally by qualified tongue picture image and unqualified tongue picture image construction in verify data, is closed The quantity of lattice tongue picture image is equal with the quantity of unqualified tongue picture image.
In order to guarantee training effectiveness and accuracy rate, training data concentrate, the quantity of qualified tongue picture image with it is unqualified The quantity of tongue picture image is n, is concentrated in verify data, the quantity of the quantity of qualified tongue picture image and unqualified tongue picture image It is m, and m is less than n.
Further, in step s 2:
It include weight compression processing, rotation processing, reversion processing, whitening processing and interception to the pretreatment that tongue picture image carries out Processing.
Specifically, as shown in Fig. 2, step S2 includes:
Step S21, weight compression processing is carried out to tongue picture image
Weight contracting is carried out to tongue picture image, the size of tongue picture image is made to be compressed to 28 × 28 pixels.
Step S22, weight contracting is carried out treated tongue picture image carries out respectively rotation processing, at reversion to through step S21 Reason, whitening processing and intercepting process
Tongue picture image compressed for weight, carries out rotation processing, reversion processing, whitening processing and intercepting process respectively, Further increase the quantity of tongue picture image.
Further, in step s3:
As shown in figure 3, step S3 includes:
Step S31, training dataset is trained by convolutional neural networks, obtains tongue picture image detection model；
Step S32, optimization is iterated to the tongue picture image detection model that step S31 is obtained using validation data set
The tongue picture image that verify data is concentrated is detected using tongue picture image detection model, will test result and true As a result it is compared, calculates the error between testing result and legitimate reading, error is carried out reversely according to error minimum principle It propagates, in back-propagation process, updates the weight in convolutional neural networks, continuous iteration.
Further, as shown in figure 4, convolutional neural networks include the first input layer, the second input layer, the first convolutional layer, First pond layer, the second convolutional layer, the second pond layer, full articulamentum, multilayer perceptron and output layer, the first input layer and second Input layer is connect with the first convolutional layer respectively, the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, full connection Layer, multilayer perceptron and output layer are sequentially connected.
First input layer is used to input the tongue picture image of training data concentration, and the second discrepancy layer is for inputting validation data set In tongue picture image.
In the first convolutional layer, using a 5 × 5 convolution kernel to the tongue picture of the first input layer or the input of the second input layer Image carries out first time process of convolution, and wherein a is the positive integer more than or equal to 1.When carrying out process of convolution, weighed using Gauss Value is weighted processing, and uses unsaturation activation primitive.Unsaturation activation primitive is that ReLU activates letter in the present embodiment Number.
The characteristic pattern for a tongue picture image that first pond layer exports the first convolutional layer carries out first time maximum value pond Hua Chu Reason.
In the second convolutional layer, second is carried out to the tongue picture image that the first pond layer exports using b 5 × 5 convolution kernel Secondary process of convolution, wherein b is the positive integer more than or equal to 1.When carrying out process of convolution, using unsaturation activation primitive, i.e., ReLU activation primitive.
The characteristic pattern for the b tongue picture image that second pond layer exports the second convolutional layer carries out second of maximum value pond Hua Chu Reason.
After the tongue picture image that full articulamentum exports the second pond layer is handled, according to the output vector dimension of its setting C, to c tongue picture images of multilayer perceptron output, wherein c is the positive integer more than or equal to 1.
After multilayer perceptron carries out Classification and Identification to the tongue picture image that full articulamentum exports, identified to output layer output category As a result.
Further, the quantity of full articulamentum is at least one.
If full articulamentum is 2, the output vector dimension of the first full articulamentum is c1, the output of the second full articulamentum to Amount dimension is c2, c2< c1。
If full articulamentum is 3, the output vector dimension of the first full articulamentum is c1, the output of the second full articulamentum to Amount dimension is c2, the output vector dimension of the full articulamentum of third is c3, and c3< c2< c1。
And so on, if full articulamentum is n, the output vector dimension of the first full articulamentum is c1... ..., n-th is complete The output vector dimension of articulamentum is cn, cn< cn-1< ... < c2< c1。
Further, tongue picture image detection model is evaluated using AUC value and accuracy rate, to show that tongue picture image is examined Survey the accuracy rate and stability of model.
It is an advantage of the current invention that obtaining the identification of a quick tongue picture quality picture classification by the optimization that iterates Device carrys out the optimized parameter of decision model training, finally obtains one by calculating the AUC value and accuracy rate of tongue picture image used The high tongue picture picture quality detection model for being unlikely to over-fitting again of accuracy rate.
Embodiment 2
The present embodiment is an Application Example of the invention.
A kind of tongue picture picture quality detection method based on deep learning, comprising the following steps:
Step S1, tongue picture image data set is constructed
Tongue picture image data set is by 200 opening and closing lattice tongue picture images and 200 unqualified tongue picture image constructions, wherein training Data set includes 200 opening and closing lattice tongue picture images and 200 unqualified tongue picture images, and validation data set includes 100 opening and closing lattice tongue pictures Image and 100 unqualified tongue picture images.
Step S2, the tongue picture image concentrated to the tongue picture image data in step S1 pre-processes
Step S21, weight compression processing, every tongue are carried out to 200 opening and closing lattice tongue picture images and 200 unqualified tongue picture images As image is compressed into 28 × 28 pixel sizes.
Step S22, the compressed 200 opening and closing lattice tongue picture image of counterweight and 200 unqualified tongue picture images are revolved respectively Turn processing, reversion processing, whitening processing and intercepting process, available 1600 tongue picture images.
Step S3, convolutional neural networks are constructed, using convolutional neural networks to through the pretreated tongue picture image of step S2 It is handled, obtains tongue picture image detection model
Convolutional neural networks include the first input layer, the second input layer, the first convolutional layer, the first pond layer, the second convolution Layer, the second pond layer, full articulamentum, multilayer perceptron and output layer, the first input layer and the second input layer respectively with the first volume Lamination connection, the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, full articulamentum, multilayer perceptron and defeated Layer is sequentially connected out.
Wherein, the first convolutional layer uses 20 5 × 5 convolution kernels, and the step-length of convolution kernel is 1；Second convolutional layer uses 50 A 5 × 5 convolution kernel, the step-length of convolution kernel are 1；In first pond layer, the core in pond is 2 × 2, and the step-length in pond is 2；? In two pond layers, the core in pond is 2 × 2, and the step-length in pond is 2.
Full articulamentum includes the first full articulamentum and the second full articulamentum, wherein the output vector of the first full articulamentum is tieed up Degree is 500, and the output vector dimension of the second full articulamentum is 2.
Step S31, training dataset is trained by convolutional neural networks, obtains tongue picture image detection model
Step S32, optimization is iterated to the tongue picture image detection model that step S31 is obtained using validation data set.
Step S4, the tongue picture image detection model obtained using step S3 detects tongue picture image to be detected.
The present embodiment and control group 1 and control group 2 are compared, compare the accuracy rate of three kinds of methods, as shown in the table. Wherein, referring to periodical literature, (" the Chinese medicine tongue picture quality evaluation study based on support vector machines ", Wang Yazhen waits to control group 1 Beijing biomedical engineering, 2015,34 (6): 551-557), control group 2 referring to periodical literature, (" comment by the quality of Chinese medicine tongue picture picture Valence research ", Zhang Xiang waits the world combination of Chinese tradiational and Western medicine impurity, 2017,12 (11): 1607-1611).
Accuracy rate The present embodiment Control group 1 Control group 2
Training dataset 0.99 0.915 0.953
Validation data set 0.98 0.875 0.935
As seen from the above table, the tongue picture picture quality detection model accuracy rate that the present embodiment uses is higher than control group 1 and control The accuracy rate of group 2.
The foregoing is merely preferred embodiments of the present invention, are not intended to limit embodiments of the present invention and protection model It encloses, to those skilled in the art, should can appreciate that all with made by description of the invention and diagramatic content Equivalent replacement and obviously change obtained scheme, should all be included within the scope of the present invention.

Claims (10)

1. a kind of tongue picture picture quality detection method based on deep learning, which comprises the following steps:
Step S1, tongue picture image data set is constructed, the tongue picture image data set includes training dataset and validation data set；
Step S2, the tongue picture image concentrated to the tongue picture image data in step S1 pre-processes；
Step S3, convolutional neural networks are constructed, using the convolutional neural networks to through the pretreated tongue picture of step S2 Image is handled, and tongue picture image detection model is obtained；
Step S4, tongue picture image to be detected is detected using the tongue picture image detection model that step S3 is obtained.
2. the tongue picture picture quality detection method according to claim 1 based on deep learning, which is characterized in that the step Suddenly S2 includes:
Step S21, weight compression processing is carried out to the tongue picture image；
Step S22, weight contracting is carried out treated the tongue picture image carries out respectively rotation processing, at reversion to through step S21 Reason, whitening processing and intercepting process.
3. the tongue picture picture quality detection method according to claim 2 based on deep learning, which is characterized in that described heavy Compression processing is to carry out weight to the tongue picture image to be reduced to 28 × 28 pixels.
4. the tongue picture picture quality detection method according to claim 1 based on deep learning, which is characterized in that described In step S1, the image data collection includes qualified tongue picture image and unqualified tongue picture image, wherein the qualification tongue picture The quantity of image is equal with the quantity of the unqualified tongue picture image.
5. the tongue picture picture quality detection method according to claim 4 based on deep learning, which is characterized in that the instruction Practicing data set includes the qualified tongue picture image and the unqualified tongue picture image, is concentrated in the training data, the qualification The quantity of tongue picture image is equal with the quantity of the unqualified tongue picture image；
The validation data set includes the qualified tongue picture image and the unqualified tongue picture image, in the validation data set In, the quantity of the qualification tongue picture image is equal with the quantity of the unqualified tongue picture image；
The quantity for the qualified tongue picture image that the verify data is concentrated is less than the qualified tongue that the training data is concentrated As the quantity of image.
6. the tongue picture picture quality detection method according to claim 1 based on deep learning, which is characterized in that the volume Product neural network includes the first input layer, the second input layer, the first convolutional layer, the first pond layer, the second convolutional layer, the second pond Layer, full articulamentum, multilayer perceptron and output layer；
First input layer is connect, for inputting the training dataset with first convolutional layer；
Second input layer is connect, for inputting the validation data set with first convolutional layer；
First convolutional layer, first pond layer, second convolutional layer, second pond layer, the full articulamentum, The multilayer perceptron and the output layer are sequentially connected with.
7. the tongue picture picture quality detection method according to claim 6 based on deep learning, which is characterized in that described One convolutional layer is weighted processing to the tongue picture image using Gauss weight, and first convolutional layer is swashed using unsaturation Function living；
Second convolutional layer uses unsaturation activation primitive.
8. the tongue picture picture quality detection method according to claim 6 based on deep learning, which is characterized in that described One pond layer carries out the processing of maximum value pondization to the result that first convolutional layer exports；
Second pond layer carries out the processing of maximum value pondization to the result that second convolutional layer exports.
9. the tongue picture picture quality detection method according to claim 1 based on deep learning, which is characterized in that the step Suddenly S3 includes:
Step S31, the training dataset is trained by the convolutional neural networks, obtains the tongue picture image detection Model；
Step S32, the tongue picture image detection model that step S31 is obtained is optimized using the validation data set.
10. the tongue picture picture quality detection method according to claim 9 based on deep learning, which is characterized in that in institute It states in step S32, optimization is iterated to the tongue picture image detection model using error minimum method.
CN201811185076.0A 2018-10-11 2018-10-11 A kind of tongue picture picture quality detection method based on deep learning Pending CN109472774A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811185076.0A CN109472774A (en) 2018-10-11 2018-10-11 A kind of tongue picture picture quality detection method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811185076.0A CN109472774A (en) 2018-10-11 2018-10-11 A kind of tongue picture picture quality detection method based on deep learning

Publications (1)

Publication Number Publication Date
CN109472774A true CN109472774A (en) 2019-03-15

Family

ID=65663943

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811185076.0A Pending CN109472774A (en) 2018-10-11 2018-10-11 A kind of tongue picture picture quality detection method based on deep learning

Country Status (1)

Country Link
CN (1) CN109472774A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110827275A (en) * 2019-11-22 2020-02-21 吉林大学第一医院 Liver nuclear magnetic artery phase image quality grading method based on raspberry group and deep learning
CN111899846A (en) * 2020-07-14 2020-11-06 苏州睿达科技有限公司 Tongue picture detector based on end cloud cooperation and detection method thereof
CN112349427A (en) * 2020-10-21 2021-02-09 上海中医药大学 Diabetes prediction method based on tongue picture and depth residual convolutional neural network
CN113362334A (en) * 2020-03-04 2021-09-07 北京悦熙兴中科技有限公司 Tongue picture processing method and device
CN113553991A (en) * 2021-08-11 2021-10-26 上海中医药大学附属龙华医院 Tongue picture recognition method based on deep learning

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017073823A1 (en) * 2015-10-29 2017-05-04 한국 한의학 연구원 Device and method for deriving adaptive threshold value and distinguishing between tongue fur, tongue texture, and mixed area thereof
CN107977671A (en) * 2017-10-27 2018-05-01 浙江工业大学 A kind of tongue picture sorting technique based on multitask convolutional neural networks

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017073823A1 (en) * 2015-10-29 2017-05-04 한국 한의학 연구원 Device and method for deriving adaptive threshold value and distinguishing between tongue fur, tongue texture, and mixed area thereof
CN107977671A (en) * 2017-10-27 2018-05-01 浙江工业大学 A kind of tongue picture sorting technique based on multitask convolutional neural networks

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
张翔等: "中医舌图像的质量评价研究", 《世界中西医结合杂志》 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110827275A (en) * 2019-11-22 2020-02-21 吉林大学第一医院 Liver nuclear magnetic artery phase image quality grading method based on raspberry group and deep learning
CN110827275B (en) * 2019-11-22 2023-12-22 吉林大学第一医院 Liver nuclear magnetic artery image quality grading method based on raspberry pie and deep learning
CN113362334A (en) * 2020-03-04 2021-09-07 北京悦熙兴中科技有限公司 Tongue picture processing method and device
CN111899846A (en) * 2020-07-14 2020-11-06 苏州睿达科技有限公司 Tongue picture detector based on end cloud cooperation and detection method thereof
CN111899846B (en) * 2020-07-14 2023-07-25 苏州睿达科技有限公司 Tongue image detector based on end cloud cooperation and detection method thereof
CN112349427A (en) * 2020-10-21 2021-02-09 上海中医药大学 Diabetes prediction method based on tongue picture and depth residual convolutional neural network
CN113553991A (en) * 2021-08-11 2021-10-26 上海中医药大学附属龙华医院 Tongue picture recognition method based on deep learning

Similar Documents

Publication Publication Date Title
CN109472774A (en) A kind of tongue picture picture quality detection method based on deep learning
Sharma et al. Performance analysis of deep learning CNN models for disease detection in plants using image segmentation
CN109754017B (en) Hyperspectral image classification method based on separable three-dimensional residual error network and transfer learning
CN104866868B (en) Metal coins recognition methods based on deep neural network and device
CN110148162A (en) A kind of heterologous image matching method based on composition operators
CN108334899A (en) Quantify the bone age assessment method of information integration based on hand bone X-ray bone and joint
CN110188824A (en) A kind of small sample plant disease recognition methods and system
CN108052881A (en) The method and apparatus of multiclass entity object in a kind of real-time detection construction site image
CN108648191A (en) Pest image-recognizing method based on Bayes&#39;s width residual error neural network
CN107832802A (en) Quality of human face image evaluation method and device based on face alignment
CN109726751A (en) Method based on depth convolutional neural networks identification brain Electrical imaging figure
Wang et al. Face recognition system based on CNN
CN108549910A (en) A kind of corn seed producing fruit ear image preliminary classification method based on convolutional neural networks
CN110135505A (en) Image classification method, device, computer equipment and computer readable storage medium
Sharma et al. KrishiMitr (Farmer’s Friend): using machine learning to identify diseases in plants
CN107909588A (en) Partition system under MRI cortex based on three-dimensional full convolutional neural networks
CN109389171A (en) Medical image classification method based on more granularity convolution noise reduction autocoder technologies
Vuong et al. Multi-task deep learning for colon cancer grading
CN109740734A (en) A kind of method of neuron spatial arrangement in optimization convolutional neural networks
CN106056161B (en) A kind of visible detection method for Plane Rotation target
Wagle et al. Effect of Data Augmentation in the Classification and Validation of Tomato Plant Disease with Deep Learning Methods.
Pinto et al. The real-time mobile application for identification of diseases in coffee leaves using the CNN model
CN108960413A (en) A kind of depth convolutional neural networks method applied to screw surface defects detection
Agrawal et al. Enhanced optic disk and cup segmentation with glaucoma screening from fundus images using position encoded CNNs
Younis et al. Robust optimization of mobilenet for plant disease classification with fine tuned parameters

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

Application publication date: 20190315

RJ01 Rejection of invention patent application after publication