CN110084804A - A kind of underwater works defect inspection method based on Weakly supervised deep learning - Google Patents
A kind of underwater works defect inspection method based on Weakly supervised deep learning Download PDFInfo
- Publication number
- CN110084804A CN110084804A CN201910361732.6A CN201910361732A CN110084804A CN 110084804 A CN110084804 A CN 110084804A CN 201910361732 A CN201910361732 A CN 201910361732A CN 110084804 A CN110084804 A CN 110084804A
- Authority
- CN
- China
- Prior art keywords
- cluster
- underwater works
- point
- weakly supervised
- deep learning
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 41
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000013135 deep learning Methods 0.000 title claims abstract description 20
- 238000007689 inspection Methods 0.000 title claims abstract description 19
- 238000001514 detection method Methods 0.000 claims abstract description 32
- 238000012360 testing method Methods 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims abstract description 17
- 230000002950 deficient Effects 0.000 claims abstract description 15
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 10
- 230000002159 abnormal effect Effects 0.000 claims description 19
- 239000000284 extract Substances 0.000 claims description 7
- 238000012512 characterization method Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 4
- 230000001174 ascending effect Effects 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000003064 k means clustering Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- JEIPFZHSYJVQDO-UHFFFAOYSA-N iron(III) oxide Inorganic materials O=[Fe]O[Fe]=O JEIPFZHSYJVQDO-UHFFFAOYSA-N 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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
- G06T2207/20024—Filtering details
-
- 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
- G06T2207/30108—Industrial image inspection
- G06T2207/30132—Masonry; Concrete
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of underwater works defect inspection methods based on Weakly supervised deep learning, the following steps are included: S1, weak mark is carried out to input picture using semantic information as label, the convolutional neural networks model under training is Weakly supervised classifies to normal picture and defective image；S2 realizes depth conspicuousness detection algorithm using the third layer convolutional layer information of convolutional neural networks model；S3 is iterated the unified detection algorithm of cluster, one reliable underwater works image abnormity point classifier of training according to the testing result of depth conspicuousness algorithm；The unified detection classifier of iteration cluster is used for underwater works image data set and carries out assessment test by S4.A kind of underwater works defect inspection method based on Weakly supervised deep learning provided by the invention, solves the problems, such as that underwater works defects detection model is difficult to construct, and testing staff can preferably be assisted to complete the defects detection task to underwater works target.
Description
Technical field
Present invention relates particularly to a kind of underwater works defect inspection methods based on Weakly supervised deep learning, belong to image
Processing is analyzed and understands technical field.
Background technique
With the development of water resources and hydropower construction, more and more water conservancy projects are used in production and living.Dam, gate slot
Equal underwater works are easy the defects of being influenced by structural stress and underwater environment and generate crack, rotten rust.Currently based on light
The Underwater Imaging technology of image is learned because its higher resolution ratio and detection accuracy are widely used in detection under water.Based on depth
The automatic detection to defect may be implemented in the image detecting method of study, it is trained according to given sample data set, obtains
The model that expected detection effect can be completed to one, is applied in image detection.However, traditional big portion of deep learning method
Point the pixel labeled data collection of magnanimity is needed to be trained, underwater works defect inspection process due to underwater complex environment,
And the random multiplicity of structures defect is difficult to obtain magnanimity labeled data.
Summary of the invention
The technical problem to be solved by the present invention is to overcome the deficiencies of existing technologies, provide a kind of semanteme for only needing image
Information as mark is trained, be able to solve underwater works defects detection model be difficult to the problem of constructing based on Weakly supervised
The underwater works defect inspection method of deep learning.
The present invention proposes a kind of underwater works defect inspection method based on Weakly supervised deep learning, solves only scheming
The semantic information of picture is concentrated as the Weakly supervised training data of label, is constructed the model of an auxiliary detection, is realized to acquired
Underwater works image Semantic detection.
In order to solve the above technical problems, the technical solution adopted by the present invention are as follows:
A kind of underwater works defect inspection method based on Weakly supervised deep learning, it is characterised in that: including following step
Rapid: S1 carries out weak mark to input picture using semantic information as label, and the convolutional neural networks model under training is Weakly supervised is right
Normal picture and defective image are classified；
S2 realizes depth conspicuousness detection algorithm using the third layer convolutional layer information of convolutional neural networks model；
S3 is iterated the unified detection algorithm of cluster according to the testing result of depth conspicuousness algorithm, and training one reliable
Underwater works image abnormity point classifier；
The unified detection classifier of iteration cluster is used for underwater works image data set and carries out assessment test by S4.
Steps are as follows for depth conspicuousness detection algorithm in S2:
S21, using the characteristic pattern of the third layer t in convolutional neural networks model in convolutional layer as the centre for generating significant point
Image Im；
S22 searches I using maximum filtering devicemIn local maximum, and be saved into list L；
S23, by the descending sort L of intensity, intensity is expressed as strength (x), and wherein x is each of L element；
S24 is done as follows each of L element x: firstly, x is added to list γ to realize initialization, so
Afterwards, it traverses and belongs to I in 8 contiguous areas of each element xmElement y, if formula strength (y) ∈ (strength (x)-z,
Strength (x)) it sets up, then it enablesAnd it willLabeled as candidate point, ifThen willAs having
Effect maximum value is added in list γ；Secondly, traversal element8 contiguous areas in each element y, if the intensity of abutment points y
In section (strength (x)-z, strength (x)), it is also marked as candidate point；Continue to access the 8 of the candidate point
Contiguous area repeats marking operation as above, until finding that the intensity of all of its neighbor point y of some candidate point meets strength (y)
< strength (x)-z, ending said process, wherein z indicates the tolerance in this algorithm to gray value, by near the maximum
The range of seed filling determines, if accessed abutment points y is present in list L, using it as weaker maximum
Therefrom L is removed value；Finally, generate a significant point by the geometric center for calculating all elements in γ, geometric center by compared with
Maximum value characterization in wide image-region.
Steps are as follows for iteration cluster Unified Algorithm in S3:
S31 enables InAnd IaRespectively indicate normal picture and defective training set of images；
S32, if PnAnd PaTo utilize DSD algorithm from InAnd IaThe significant point set of middle extraction；
S33, iteration execute operations described below S times, wherein S > 1: to each width normal picture In, extract PnDotted intersection it is special
PCFM is levied, and uses K-means to cluster for normal cluster Q its dotted cross feature, cluster interior element Nq, q=1,2 ... Q
It indicates；To each defective image Ia, extract PaDotted cross feature indicate, and by its dotted cross feature use K-
Means cluster is abnormal cluster R, cluster interior element Ar, r=1,2 ... R is indicated；
S34 enables Q=QS, R=RS；
S35, to each of abnormal cluster R elements Ar, r=1,2 ... R is done as follows: in ArAnd NqBetween, meter
Calculate the distance between each point drq(Ar,Nq), q=1,2 ... then Q adjusts the distance d by ascending orderrqSequence；Then, normalizing is calculated
Change distance drq12=drq1/drq2, wherein drq1And drq2Respectively indicate ArThe adjacent cluster midpoint N nearest to itq1And Nq2Between
Distance estimates the distance between cluster using the Euclidean distance between cluster mass center；
S36, all distance d obtained from S35rq12In estimate average normalized distance dq12；
S37, to each of abnormal cluster elements ArIt is done as follows: if formula drq12< dq12It sets up, then by this yuan
Plain ArIt is rejected from abnormal cluster, is included into normal cluster.
In S4, assessment test is detected simultaneously by K-means clustering method handle significant point relevant to underwater works defect
It is identified.
Specific steps are as follows: defective underwater works image will be classified asIt is set as input picture, it is rightIn
Each significant point m is done as follows: extracting the dotted cross feature PCFM feature of point m；Then m to all cluster, that is, A is calculatedr
∪NqThe distance of each point in cluster；Class belonging in the K of m recently abutment points carries out majority voting, and what be will test is aobvious
It writes point and is classified as normal or abnormal point.
With the defect area of rectangle frame instruction underwater works.
A kind of beneficial effects of the present invention: underwater works defect inspection based on Weakly supervised deep learning provided by the invention
Survey method, it is only necessary to which the semantic information of image is trained as mark, can obtain an ideal image detection model.It should
Framework include depth conspicuousness detection and iteration cluster Unified Algorithm, only image semantic information as the Weakly supervised of label
Training data is concentrated, and solves the problems, such as that underwater works defects detection model is difficult to construct, and can preferably assist detection people
Member completes the defects detection task to underwater works target.
Detailed description of the invention
The training flow chart of underwater works defect inspection method of the Fig. 1 based on Weakly supervised deep learning；
The test flow chart of underwater works defect inspection method of the Fig. 2 based on Weakly supervised deep learning.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings, and following embodiment is only used for clearly illustrating the present invention
Technical solution, and not intended to limit the protection scope of the present invention.
As depicted in figs. 1 and 2, the present invention provides a kind of underwater works defects detection side based on Weakly supervised deep learning
Method, comprising the following steps:
Step 1 carries out weak mark to input picture using semantic information as label, the convolutional Neural net under training is Weakly supervised
Network model (WCNN) is realized to normal and abnormal image training set classification.
Step 2 realizes depth conspicuousness detection algorithm using the third layer convolutional layer information of convolutional neural networks model,
For detecting the significant point of defective underwater works image and zero defect image, depth conspicuousness detection algorithm (DSD) step
It is as follows:
A, using the characteristic pattern of the third layer t in convolutional neural networks model in convolutional layer as the middle graph for generating significant point
As Im；
B searches I using maximum filtering devicemIn local maximum, and be saved into list L；
C, by the descending sort L of intensity, intensity is expressed as strength (x), and wherein x is each of L element；
D is done as follows each of L element x: firstly, x is added to list γ with realize initialization, then,
It traverses and belongs to I in 8 contiguous areas of each element xmElement y, if formula strength (y) ∈ (strength (x)-z,
Strength (x)) it sets up, then it enablesAnd it willLabeled as candidate point, ifThen willAs having
Effect maximum value is added in list γ；Secondly, traversal element8 contiguous areas in each element y, if the intensity of abutment points y
In section (strength (x)-z, strength (x)), it is also marked as candidate point；Continue to access the 8 of the candidate point
Contiguous area repeats marking operation as above, until finding that the intensity of all of its neighbor point y of some candidate point meets strength (y)
< strength (x)-z, ending said process, wherein z indicates the tolerance in this algorithm to gray value, by near the maximum
The range of seed filling determines, if accessed abutment points y is present in list L, using it as weaker maximum
Therefrom L is removed value；Finally, generate a significant point by the geometric center for calculating all elements in γ, geometric center by compared with
Maximum value characterization in wide image-region.
Step 3 is iterated cluster Unified Algorithm (ICU) detection, repeatedly according to the testing result of depth conspicuousness algorithm
Generation cluster Unified Algorithm extracts the dotted cross feature (PCFM) being made of significant point, and labeled as abnormal or normal, training one
A reliable underwater works image abnormity point classifier.By utilizing extensive underwater works image under Weakly supervised environment
Data are trained WCNN network model and ICU classifier, obtain an ideal detection model.Iteration clusters Unified Algorithm
Steps are as follows:
A enables InAnd IaRespectively indicate normal picture and defective training set of images；
B, if PnAnd PaTo utilize DSD algorithm from InAnd IaThe significant point set of middle extraction；
C, iteration execute operations described below S times, wherein S > 1: to each width normal picture In, extract PnDotted cross feature
PCFM, and use K-means to cluster for normal cluster Q its dotted cross feature, cluster interior element Nq, q=1,2 ... Q table
Show；To each defective image Ia, extract PaDotted cross feature indicate, and by its dotted cross feature use K-means
Cluster is abnormal cluster R, cluster interior element Ar, r=1,2 ... R is indicated；
D enables Q=QS, R=RS；
E, to each of abnormal cluster R elements Ar, r=1,2 ... R is done as follows: in ArAnd NqBetween, it calculates
The distance between each point drq(Ar,Nq), q=1,2 ... then Q adjusts the distance d by ascending orderrqSequence；Then, normalization is calculated
Distance drq12=drq1/drq2, wherein drq1And drq2Respectively indicate ArThe adjacent cluster midpoint N nearest to itq1And Nq2Between away from
From, i.e., using cluster mass center between Euclidean distance come estimate cluster the distance between；
F, all distance d obtained from step erq12In estimate average normalized distance dq12；
G, to each of abnormal cluster elements ArIt is done as follows: if formula drq12< dq12It sets up, then by the element
ArIt is rejected from abnormal cluster, is included into normal cluster.
The unified detection classifier of iteration cluster is used for underwater works image data set and carries out assessment test by step 4.
Assessment test is detected and is identified by K-means clustering method handle significant point relevant to underwater works defect.Specifically
Step are as follows: defective underwater works image will be classified asIt is set as input picture, it is rightEach of significant point m
It is done as follows: extracting the dotted cross feature PCFM feature of point m；Then m to all cluster, that is, A is calculatedr∪NqEach point in cluster
Distance；Class belonging in the K of m recently abutment points carries out majority voting, and the significant point that will test is classified as normally
Or abnormal point.With the defect area of rectangle frame instruction underwater works.
A kind of underwater works defect inspection method based on Weakly supervised deep learning of the invention, in model training rank
Section, DSD are applied to zero defect and defective image, indicate underwater structure using the information extracted from the characteristic pattern of WCNN convolutional layer
Build the significant characteristics in object image.In test phase, DSD is applied to obtain significant point through the sorted defect image of WCNN
The defect of characterization.A kind of novel iteration of the invention clusters Unified Algorithm, is calculated based on clustering method the DSD under Weakly supervised
The significant point that method detects is classified, and trained and test phase is related to.In the training process, defective and normal structure is received
Object image is built, is clustered by K-means and their significant point is gathered in their vector representation.In test phase, mainly
It is responsible for handling the abnormal image of existing defects, obtains several abnormal points of underwater works defect, and referred to rectangle frame
Show the defect area of structures.
The above is only a preferred embodiment of the present invention, it should be pointed out that: for the ordinary skill people of the art
For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also answered
It is considered as protection scope of the present invention.
Claims (6)
1. a kind of underwater works defect inspection method based on Weakly supervised deep learning, it is characterised in that: the following steps are included:
S1 carries out weak mark to input picture using semantic information as label, and the convolutional neural networks model under training is Weakly supervised is right
Normal picture and defective image are classified；
S2 realizes depth conspicuousness detection algorithm using the third layer convolutional layer information of convolutional neural networks model；
S3 is iterated the unified detection algorithm of cluster, one reliable water of training according to the testing result of depth conspicuousness algorithm
Lower structures image abnormity point classifier；
The unified detection classifier of iteration cluster is used for underwater works image data set and carries out assessment test by S4.
2. a kind of underwater works defect inspection method based on Weakly supervised deep learning according to claim 1, special
Sign is: steps are as follows for depth conspicuousness detection algorithm in S2:
S21, using the characteristic pattern of the third layer t in convolutional neural networks model in convolutional layer as the intermediate image for generating significant point
Im；
S22 searches I using maximum filtering devicemIn local maximum, and be saved into list L；
S23, by the descending sort L of intensity, intensity is expressed as strength (x), and wherein x is each of L element；
S24 is done as follows each of L element x: firstly, x is added to list γ with realize initialization, then, time
It goes through in 8 contiguous areas of each element x and belongs to ImElement y, if formula strength (y) ∈ (strength (x)-z,
Strength (x)) it sets up, then it enablesAnd it willLabeled as candidate point, ifThen willAs having
Effect maximum value is added in list γ；Secondly, traversal element8 contiguous areas in each element y, if at the intensity of abutment points y
In section (strength (x)-z, strength (x)), it is also marked as candidate point；Continue to access the candidate point 8 are adjacent
Domain is connect, marking operation as above is repeated, until finding that the intensity of all of its neighbor point y of some candidate point meets strength (y) <
Strength (x)-z, ending said process, wherein z indicates the tolerance in this algorithm to gray value, by planting near the maximum
The range of son filling determines, if accessed abutment points y is present in list L, using it as weaker maximum value
Therefrom L is removed；Finally, generating a significant point by the geometric center for calculating all elements in γ, geometric center is by wider
Maximum value characterization in image-region.
3. a kind of underwater works defect inspection method based on Weakly supervised deep learning according to claim 1, special
Sign is: steps are as follows for iteration cluster Unified Algorithm in S3:
S31 enables InAnd IaRespectively indicate normal picture and defective training set of images；
S32, if PnAnd PaTo utilize DSD algorithm from InAnd IaThe significant point set of middle extraction；
S33, iteration execute operations described below S times, wherein S > 1: to each width normal picture In, extract PnDotted cross feature
PCFM, and use K-means to cluster for normal cluster Q its dotted cross feature, cluster interior element Nq, q=1,2 ... Q table
Show；To each defective image Ia, extract PaDotted cross feature indicate, and by its dotted cross feature use K-means
Cluster is abnormal cluster R, cluster interior element Ar, r=1,2 ... R is indicated；
S34 enables Q=QS, R=RS；
S35, to each of abnormal cluster R elements Ar, r=1,2 ... R is done as follows: in ArAnd NqBetween, it calculates each
The distance between a point drq(Ar,Nq), q=1,2 ... then Q adjusts the distance d by ascending orderrqSequence；Then, calculate normalization away from
From drq12=drq1/drq2, wherein drq1And drq2Respectively indicate ArThe adjacent cluster midpoint N nearest to itq1And Nq2The distance between,
The distance between cluster is estimated using the Euclidean distance between cluster mass center；
S36, all distance d obtained from S35rq12In estimate average normalized distance dq12；
S37, to each of abnormal cluster elements ArIt is done as follows: if formula drq12< dq12It sets up, then by the elements Ar
It is rejected from abnormal cluster, is included into normal cluster.
4. a kind of underwater works defect inspection method based on Weakly supervised deep learning according to claim 1, special
Sign is: in S4, assessment test is detected and is marked by K-means clustering method handle significant point relevant to underwater works defect
Knowledge comes out.
5. a kind of underwater works defect inspection method based on Weakly supervised deep learning according to claim 4, special
Sign is: specific steps are as follows: will be classified as defective underwater works imageIt is set as input picture, it is rightIn it is every
One significant point m is done as follows: extracting the dotted cross feature PCFM feature of point m；Then m to all cluster, that is, A is calculatedr∪
NqThe distance of each point in cluster；Class belonging in the K of m recently abutment points carries out majority voting, and what be will test is significant
Point is classified as normal or abnormal point.
6. a kind of underwater works defect inspection method based on Weakly supervised deep learning according to claim 5, special
Sign is: with the defect area of rectangle frame instruction underwater works.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910361732.6A CN110084804A (en) | 2019-04-30 | 2019-04-30 | A kind of underwater works defect inspection method based on Weakly supervised deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910361732.6A CN110084804A (en) | 2019-04-30 | 2019-04-30 | A kind of underwater works defect inspection method based on Weakly supervised deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110084804A true CN110084804A (en) | 2019-08-02 |
Family
ID=67418083
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910361732.6A Pending CN110084804A (en) | 2019-04-30 | 2019-04-30 | A kind of underwater works defect inspection method based on Weakly supervised deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110084804A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112508106A (en) * | 2020-12-08 | 2021-03-16 | 大连海事大学 | Underwater image classification method based on convolutional neural network |
| CN112700435A (en) * | 2021-01-12 | 2021-04-23 | 华南理工大学 | Wall defect detection method based on deep learning |
| CN114372983A (en) * | 2022-03-22 | 2022-04-19 | 武汉市富甸科技发展有限公司 | Shielding box coating quality detection method and system based on image processing |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106097315A (en) * | 2016-06-03 | 2016-11-09 | 河海大学常州校区 | A kind of underwater works crack extract method based on sonar image |
| CN108399406A (en) * | 2018-01-15 | 2018-08-14 | 中山大学 | The method and system of Weakly supervised conspicuousness object detection based on deep learning |
-
2019
- 2019-04-30 CN CN201910361732.6A patent/CN110084804A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106097315A (en) * | 2016-06-03 | 2016-11-09 | 河海大学常州校区 | A kind of underwater works crack extract method based on sonar image |
| CN108399406A (en) * | 2018-01-15 | 2018-08-14 | 中山大学 | The method and system of Weakly supervised conspicuousness object detection based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| DIMITRIS K. IAKOVIDIS等: "Detecting and Locating Gastrointestinal Anomalies Using Deep Learning and Iterative Cluster Unification", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112508106A (en) * | 2020-12-08 | 2021-03-16 | 大连海事大学 | Underwater image classification method based on convolutional neural network |
| CN112508106B (en) * | 2020-12-08 | 2024-05-24 | 大连海事大学 | Underwater image classification method based on convolutional neural network |
| CN112700435A (en) * | 2021-01-12 | 2021-04-23 | 华南理工大学 | Wall defect detection method based on deep learning |
| CN112700435B (en) * | 2021-01-12 | 2023-04-07 | 华南理工大学 | Wall defect detection method based on deep learning |
| CN114372983A (en) * | 2022-03-22 | 2022-04-19 | 武汉市富甸科技发展有限公司 | Shielding box coating quality detection method and system based on image processing |
| CN114372983B (en) * | 2022-03-22 | 2022-05-24 | 武汉市富甸科技发展有限公司 | Shielding box coating quality detection method and system based on image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108388927B (en) | Small sample polarization SAR terrain classification method based on deep convolution twin network | |
| CN105512640B (en) | A kind of people flow rate statistical method based on video sequence | |
| CN109285139A (en) | A kind of x-ray imaging weld inspection method based on deep learning | |
| CN108596055B (en) | Airport target detection method of high-resolution remote sensing image under complex background | |
| CN109101938B (en) | Multi-label age estimation method based on convolutional neural network | |
| Yang et al. | Automatic pixel-level crack detection for civil infrastructure using Unet++ and deep transfer learning | |
| CN110084804A (en) | A kind of underwater works defect inspection method based on Weakly supervised deep learning | |
| CN104298992B (en) | A kind of adaptive scale pedestrian recognition methods again based on data-driven | |
| CN111639587B (en) | Hyperspectral image classification method based on multi-scale spectrum space convolution neural network | |
| CN107808375B (en) | Merge the rice disease image detecting method of a variety of context deep learning models | |
| CN111368690A (en) | Deep learning-based video image ship detection method and system under influence of sea waves | |
| CN106408030A (en) | SAR image classification method based on middle lamella semantic attribute and convolution neural network | |
| CN109978872B (en) | White matter microstructure characteristic screening system and method based on white matter fiber tracts | |
| CN113780242A (en) | Cross-scene underwater sound target classification method based on model transfer learning | |
| CN114170527B (en) | Remote sensing target detection method using rotating frame representation | |
| CN108509980B (en) | Water level monitoring method based on dictionary learning | |
| Li et al. | SPM-IS: An auto-algorithm to acquire a mature soybean phenotype based on instance segmentation | |
| Zhou et al. | Convolutional neural networks–based model for automated sewer defects detection and classification | |
| Gou et al. | Pavement crack detection based on the improved faster-rcnn | |
| CN109543672A (en) | Object detecting method based on dense characteristic pyramid network | |
| CN109635726A (en) | A kind of landslide identification method based on the symmetrical multiple dimensioned pond of depth network integration | |
| Li et al. | A grid‐based classification and box‐based detection fusion model for asphalt pavement crack | |
| Xiong et al. | Pixel-Level patch detection from full-scale asphalt pavement images based on deep learning | |
| Liu et al. | A novel deep framework for change detection of multi-source heterogeneous images | |
| CN111291818A (en) | Non-uniform class sample equalization method for cloud mask |
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: 20190802 |