CN110322442A - A kind of building surface crack detecting method based on SegNet - Google Patents
A kind of building surface crack detecting method based on SegNet Download PDFInfo
- Publication number
- CN110322442A CN110322442A CN201910623302.7A CN201910623302A CN110322442A CN 110322442 A CN110322442 A CN 110322442A CN 201910623302 A CN201910623302 A CN 201910623302A CN 110322442 A CN110322442 A CN 110322442A
- Authority
- CN
- China
- Prior art keywords
- segnet
- model
- crack
- image
- building surface
- 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
- 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
- G06T2207/20—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
Abstract
The present invention relates to a kind of building surface crack detecting method based on SegNet, including step S1: acquisition primitive bridge crack image data set and original wall crack image data set；Step S2: the image of step S1 acquisition is annotated, the profile of crackle is marked out；Step S3: image preprocessing is carried out；Step S4: carrying out model training, realizes SegNet model using Keras frame, and under Anaconda platform, and pretreated data set is sent into improved SegNet model and is trained, saves best model after the completion of training；Step S5: crack detection is carried out using the best model that step S4 is saved.The present invention can effectively adapt to environmental change, reach preferable detection effect, play an important role for the accuracy rate and efficiency that promote building surface crack detection.
Description
Technical field
The present invention relates to building field of crack detection, especially a kind of building surface crack detection based on SegNet
Method.
Background technique
Over time, the buildings such as bridge and wall can generate two different crackles, and one is due to load
And the stress crackle generated, another kind are the non-stress crackle that generates due to internal factors such as restrained deformations.It is acted on by load
Caused crack is unstable on component, and the beauty for not only influencing building has also resulted in the durability of building, held
Carry the reduced performances such as power and waterproofness, with crackle forming continuous expansion will badly damaged building, and bring serious safety
Potential problem.Therefore fracture inspect periodically particularly important.
In traditional crack detection, artificial on-site land survey is needed mostly, and resources costs phase not high there are detection efficiency
To height, and it is easy the presence of the problems such as detection blind spot.Manned surveys can be helped to be effectively reduced using computer image processing technology
Cost.However, reducing the accuracy of detection since image procossing is vulnerable to Environmental Noise Influence, such as illumination.
Summary of the invention
In view of this, the purpose of the present invention is to propose to a kind of building surface crack detecting method based on SegNet, energy
Environmental change is effectively adapted to, preferable detection effect is reached, for promoting the accuracy rate and effect of building surface crack detection
Rate plays an important role.
The present invention is realized using following scheme: a kind of building surface crack detecting method based on SegNet, specific to wrap
Include following steps:
Step S1: acquisition primitive bridge crack image data set and original wall crack image data set；
Step S2: the image of step S1 acquisition is annotated, the profile of crackle is marked out；
Step S3: image preprocessing is carried out；
Step S4: carrying out model training, realizes SegNet model using Keras frame, and under Anaconda platform, will locate in advance
Data set after reason is sent into improved SegNet model and is trained, and saves best model after the completion of training；
Step S5: crack detection is carried out using the best model that step S4 is saved.
Further, in step S1, the resolution ratio of primitive bridge crack image data set is 1024 × 1024 pixels；It is original
The resolution ratio of wall crack image data set is 256 × 256 pixels.
Further, step S2 is marked specifically, in order to protrude the cracks in image using a binaryzation
Artwork master, pixel be 0 and 255；It is wherein concentrated in bridge crack image data and marks crackle at white, background is labeled as black
Color；Since metope background is largely white, crackle is marked into black in wall crack image data set, background mark at
White；After the completion of mark, the crackle that the brightness of image is adjusted to most secretly obtain binaryzation is marked into image.
Further, step S3 specifically: primitive bridge crack image is cut, to original wall crack image into
The enhancing of row data.
Further, step S4 includes: to be changed to initial SegNet model the segmentation of 21 seed types to 2 seed types
More classification situations are changed to two classification by segmentation；And inputted using gray level image, input size is changed by original 360 × 480
At 256 × 256.
Further, step S4 includes: over-fitting in order to prevent, by all convolutional layer and ReLU in SegNet model
BN layers are added between activation primitive.
Further, step S4 includes: and changes stochastic gradient descent in SegNet model into order to improve convergence rate and adopt
With Adam optimizer, if while continuous 5 wheel of training of setting, modelling effect, which does not improve, just decays to original five for learning rate
/ mono-, wherein setting the minimum of learning rate as 0.00001.
Further, step S5 specifically: use sliding window scan method, crackle picture to be detected is swept
It retouches, and is sequentially sent to the trained SegNet model of step S4 and carries out classification judgement；Wherein bridge crack image data is concentrated, when
When judging result is crackle, the pixel value of corresponding region retains, conversely, then pixel value assigns 0；In wall crack image data set,
When judging result is crackle, the pixel value of corresponding region retains；Conversely, then pixel value assigns 1.
Further, before being detected, picture to be detected is done into boundary zero padding, obtained picture size is long,
Wide is 256 integral multiple；Then filled image is scanned with 256 × 256 sliding window, scanning sequency be from
It is left-to-right, from top to bottom.
Compared with prior art, the invention has the following beneficial effects:
1, without saving the characteristic pattern of entire coded portion when the SegNet model that uses of the present invention indexes down-sampled using pondization,
It can be realized the lightweight network model for possessing less parameters, it is low to request memory.
2, the SegNet model that uses of the present invention did not had to carry out deconvolution in the up-sampling stage, simplified model, can be with
Reach preferable cutting effect.
3, the present invention can be suitable for the crack detection under different condition；Traditional image processing method is easy by illumination
The influence of the external environments such as condition variation, background interference variation, and the building surface of the invention based on SegNet is split
For marks detection method, the method proposed is attained by acceptable essence without being post-processed under various complex backgrounds
Degree, better than traditional edge detection method.
4, the present invention can be realized using sliding window scanning technique and be detected to the crack image of arbitrary size.
Detailed description of the invention
Fig. 1 is the overall flow schematic diagram of the embodiment of the present invention.
Fig. 2 is the SegNet model method flow diagram of the embodiment of the present invention.
Fig. 3 is the bridge crack data set cutting effect example of the embodiment of the present invention.
Fig. 4 is the wall crack data collection reinforcing effect example of the embodiment of the present invention.
Fig. 5 is original image, mark figure and the testing result figure using present invention method under bridge crack data set.
Fig. 6 is original image, mark figure and the testing result figure using present invention method under wall crack data collection.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and embodiments.
It is noted that described further below be all exemplary, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms used herein has usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular
Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
As shown in Figure 1, a kind of building surface crack detecting method based on SegNet is present embodiments provided, it is specific to wrap
Include following steps:
Step S1: acquisition primitive bridge crack image data set and original wall crack image data set；Including network
Disclosed primitive bridge crack data collection 2068 is opened and the original wall crack data collection 110 of laboratory team mobile phone shooting is opened；
Step S2: the image of step S1 acquisition is annotated, the profile of crackle is marked out；
Step S3: image preprocessing is carried out；
Step S4: carrying out model training, realizes SegNet model using Keras frame, and under Anaconda platform, will locate in advance
Data set after reason is sent into improved SegNet model and is trained, and saves best model after the completion of training；
Step S5: crack detection is carried out using the best model that step S4 is saved.
In the present embodiment, in step S1, the resolution ratio of primitive bridge crack image data set is 1024 × 1024 pixels；
The resolution ratio of original wall crack image data set is 256 × 256 pixels.
In the present embodiment, step S2 is marked specifically, in order to protrude the cracks in image using one two
The artwork master of value, pixel are 0 and 255；Wherein concentrate crackle mark in bridge crack image data into white, background mark
For black；Since metope background is largely white, crackle is marked into black, background mark in wall crack image data set
Form white；After the completion of mark, the crackle that the brightness of image is adjusted to most secretly obtain binaryzation is marked into image.
Preferably, annotation tool marks editing machine (Semantic using the semantic object based on Web in step S2
Segmentation Editor).
In the present embodiment, step S3 specifically: primitive bridge crack image is cut, to original wall crackle figure
As carrying out data enhancing.
Wherein, since the pixel of photo resolution 1024 × 1024 is not suitable for neural network in bridge crack data set
Training, therefore its pixel value is cut into 256 × 256 using python code, the picture for being free of crackle is rejected, is ultimately formed
5180 bridge crack samples pictures are divided into training set, test set and verifying according to the ratio of 8:1:1 and collect.
Wherein, wall crack data is concentrated since data set only has 110, and data volume is too small to be easy to cause over-fitting, institute
Increase sample size in a manner of using data enhancing, main use rotates and turn over and combine the sides such as fuzzy and addition noise
Formula improves sample size, ultimately forms 326 wall crackle pictures, is divided into training set, test set according to the ratio of 8:1:1 and tests
Card collection.
In the present embodiment, step S4 includes: to be changed to initial SegNet model the segmentation of 21 seed types to 2 types
More classification situations are changed to two classification by the segmentation of type；And inputted using gray level image, will input size by original 360 ×
480 are changed to 256 × 256.
In the present embodiment, step S4 includes: over-fitting in order to prevent, in SegNet model by all convolutional layers with
BN layers are added between ReLU activation primitive.
In the present embodiment, step S4 includes: in order to improve convergence rate, by stochastic gradient descent in SegNet model
(SGD) change into using Adam optimizer (learning rate 0.0001), if while continuous 5 wheel of training of setting, modelling effect do not improve
Learning rate is just decayed into original 1/5th, wherein setting the minimum of learning rate as 0.00001.
Preferably, in the present embodiment, in step S4, two experimental data sets train 100 epoch, training process
In be not that each round all saves parameter, when penalty values verifying collection on obtain optimum value when then save parameter, if training later
It does not obtain more preferable effect and just no longer saves parameter.
In the present embodiment, step S5 specifically: use sliding window scan method, crackle picture to be detected is carried out
Scanning, and be sequentially sent to the trained SegNet model of step S4 and carry out classification judgement；Wherein bridge crack image data is concentrated,
When judging result is crackle, the pixel value of corresponding region retains, conversely, then pixel value assigns 0；Wall crack image data set
In, when judging result is crackle, the pixel value of corresponding region retains；Conversely, then pixel value assigns 1.
In the present embodiment, during actual experiment, it is contemplated that the picture of shooting is all larger-size picture, will be schemed
It is unpractical that the size that piece is cut into 256 × 256 carries out detection again, therefore before being detected, by picture to be detected
Do boundary zero padding, obtained picture size length and width is 256 integral multiple, and in addition obtain with expand after it is onesize
Figure A, scheme A value be full 0；Then filled image is scanned with 256 × 256 sliding window, scanning sequency is
From left to right, the small figure from top to bottom, obtained in order to cutting carries out prediction and the result of prediction is placed in the correspondence position of figure A
It sets, finally obtains complete prognostic chart A, cutting recovery is carried out to figure A and obtains the script size of image.
In the concrete case of application the present embodiment method:
As shown in Figure 1, being annotated acquired image using the Note tool, pressed after bridge crack data images are cut
Training set, test set and verifying collection are randomly divided into according to certain proportion；By wall crack data collection according to one after data enhance
Certainty ratio is randomly divided into training set, test set and verifying collection.Then crack detection depth is carried out using improved SegNet model
Study, obtains preservation model after optimal models, finally tests final result using sliding window scanning technique.
Use Anaconda science platform and Keras frame (using TensorFlow as rear end) in Intel Core
The network of explanation is realized in the computer of i5-4430 CPU.Using the ReduceLROnPlateau method of Keras come according to mould
The situation of change regularized learning algorithm rate of the performance indicator of type training stage.In addition, the ModelCheckpoint using Keras will be instructed
The smallest Model Weight of penalty values saves as best model on training set during white silk.
As shown in Fig. 2, SegNet network used in the present embodiment, SegNet realizes end using coding and decoded structure
To the semantic segmentation at end.The coded portion of SegNet is modified on VGG16, and is corresponded with decoding layer, is in symmetric figure
Shape finally seeks the probability value in each classification of each pixel using Softmax layers.SegNet coding mode is the same as CNN's
Convolution operation is the same, and the extraction of feature is realized using convolutional layer, then the size of input feature vector figure is reduced using pondization operation, reduces
Number of parameters accelerates model training speed.SegNet decoding process still uses zero padding convolution, and the process is mainly for wadding warp
The profile information of up-sampling is crossed, so that the information for allowing pondization operation to abandon obtains in decoding process.
Fig. 3, Fig. 4 are effect exemplary diagram after data set processing；In Fig. 3, (a) is original image, is (b) mark figure, (c) is to cut
Schematic diagram.In Fig. 4, (a) is original image, is (b) rotation, (c) is flip horizontal, (d) is flip vertical, is (d) scaling, (f) is
Brightness change, (g) be it is fuzzy, (h) cut for mistake, be (i) elastic deformation, (j) be salt-pepper noise.
Fig. 5, Fig. 6 are final result exemplary diagram.In Fig. 5, (a) is three exemplary original images, (b) corresponding for (a)
Mark figure (c) is (a) corresponding testing result figure.In Fig. 6, (a) is three exemplary original images, (b) is (a) corresponding mark
Figure (c) is (a) corresponding testing result figure.
It should be understood by those skilled in the art that, embodiments herein can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the application
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the application, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The application is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present application
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
The above described is only a preferred embodiment of the present invention, being not that the invention has other forms of limitations, appoint
What those skilled in the art changed or be modified as possibly also with the technology contents of the disclosure above equivalent variations etc.
Imitate embodiment.But without departing from the technical solutions of the present invention, according to the technical essence of the invention to above embodiments institute
Any simple modification, equivalent variations and the remodeling made, still fall within the protection scope of technical solution of the present invention.
Claims (9)
1. a kind of building surface crack detecting method based on SegNet, which comprises the following steps:
Step S1: acquisition primitive bridge crack image data set and original wall crack image data set；
Step S2: the image of step S1 acquisition is annotated, the profile of crackle is marked out；
Step S3: image preprocessing is carried out；
Step S4: carrying out model training, realizes SegNet model using Keras frame, and under Anaconda platform, will locate in advance
Data set after reason is sent into improved SegNet model and is trained, and saves best model after the completion of training；
Step S5: crack detection is carried out using the best model that step S4 is saved.
2. a kind of building surface crack detecting method based on SegNet according to claim 1, which is characterized in that step
In rapid S1, the resolution ratio of primitive bridge crack image data set is 1024 × 1024 pixels；Original wall crack image data set
Resolution ratio be 256 × 256 pixels.
3. a kind of building surface crack detecting method based on SegNet according to claim 1, which is characterized in that step
Rapid S2 is specifically, in order to protrude the cracks in image, and for mark using the artwork master of a binaryzation, pixel is 0 He
255；It is wherein concentrated in bridge crack image data and marks crackle at white, background is labeled as black；Since metope background is big
Part is white, is marked crackle at black in wall crack image data set, background is marked into white；After the completion of mark,
The crackle that the brightness of image is adjusted to most secretly obtain binaryzation is marked into image.
4. a kind of building surface crack detecting method based on SegNet according to claim 1, which is characterized in that step
Rapid S3 specifically: primitive bridge crack image is cut, data enhancing is carried out to original wall crack image.
5. a kind of building surface crack detecting method based on SegNet according to claim 1, which is characterized in that step
Rapid S4 includes: that initial SegNet model is changed to segmentation to 2 seed types to the segmentation of 21 seed types, i.e., will classify situation more
It is changed to two classification；And inputted using gray level image, input size is changed to 256 × 256 by original 360 × 480.
6. a kind of building surface crack detecting method based on SegNet according to claim 1, which is characterized in that step
Rapid S4 includes: over-fitting in order to prevent, will be added between all convolutional layers and ReLU activation primitive in SegNet model
BN layers.
7. a kind of building surface crack detecting method based on SegNet according to claim 1, which is characterized in that step
Rapid S4 includes: to change stochastic gradient descent in SegNet model using Adam optimizer, simultaneously into order to improve convergence rate
If continuous 5 wheel of training of setting, modelling effect, which does not improve, just decays to original 1/5th for learning rate, wherein setting learning rate
Minimum be 0.00001.
8. a kind of building surface crack detecting method based on SegNet according to claim 1, which is characterized in that step
Rapid S5 specifically: use sliding window scan method, crackle picture to be detected is scanned, and be sequentially sent to step S4 instruction
The SegNet model perfected carries out classification judgement；Wherein bridge crack image data is concentrated, corresponding when judging result is crackle
The pixel value in region retains, conversely, then pixel value assigns 0；It is right when judging result is crackle in wall crack image data set
The pixel value in region is answered to retain；Conversely, then pixel value assigns 1.
9. a kind of building surface crack detecting method based on SegNet according to claim 8, which is characterized in that
Before being detected, picture to be detected is done into boundary zero padding, obtained picture size length and width is 256 integral multiple；It connects
Filled image is scanned with 256 × 256 sliding window, scanning sequency be from left to right, from top to bottom.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910623302.7A CN110322442A (en) | 2019-07-11 | 2019-07-11 | A kind of building surface crack detecting method based on SegNet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910623302.7A CN110322442A (en) | 2019-07-11 | 2019-07-11 | A kind of building surface crack detecting method based on SegNet |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110322442A true CN110322442A (en) | 2019-10-11 |
Family
ID=68121911
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910623302.7A Pending CN110322442A (en) | 2019-07-11 | 2019-07-11 | A kind of building surface crack detecting method based on SegNet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110322442A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112967249A (en) * | 2021-03-03 | 2021-06-15 | 南京工业大学 | Intelligent identification method for manufacturing errors of prefabricated pier reinforcing steel bar holes based on deep learning |
| CN113436138A (en) * | 2021-03-31 | 2021-09-24 | 成都飞机工业（集团）有限责任公司 | Image preprocessing method for aviation structural part identification based on digital image |
| CN113792783A (en) * | 2021-09-13 | 2021-12-14 | 陕西师范大学 | Automatic identification method and system for dough mixing stage based on deep learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108710919A (en) * | 2018-05-25 | 2018-10-26 | 东南大学 | A kind of crack automation delineation method based on multi-scale feature fusion deep learning |
| CN108805863A (en) * | 2018-05-02 | 2018-11-13 | 南京工程学院 | The method of depth convolutional neural networks combining form detection image variation |
| CN108876780A (en) * | 2018-06-26 | 2018-11-23 | 陕西师范大学 | Bridge Crack image crack detection method under a kind of complex background |
| CN109376773A (en) * | 2018-09-30 | 2019-02-22 | 福州大学 | Crack detecting method based on deep learning |
-
2019
- 2019-07-11 CN CN201910623302.7A patent/CN110322442A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108805863A (en) * | 2018-05-02 | 2018-11-13 | 南京工程学院 | The method of depth convolutional neural networks combining form detection image variation |
| CN108710919A (en) * | 2018-05-25 | 2018-10-26 | 东南大学 | A kind of crack automation delineation method based on multi-scale feature fusion deep learning |
| CN108876780A (en) * | 2018-06-26 | 2018-11-23 | 陕西师范大学 | Bridge Crack image crack detection method under a kind of complex background |
| CN109376773A (en) * | 2018-09-30 | 2019-02-22 | 福州大学 | Crack detecting method based on deep learning |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112967249A (en) * | 2021-03-03 | 2021-06-15 | 南京工业大学 | Intelligent identification method for manufacturing errors of prefabricated pier reinforcing steel bar holes based on deep learning |
| CN113436138A (en) * | 2021-03-31 | 2021-09-24 | 成都飞机工业（集团）有限责任公司 | Image preprocessing method for aviation structural part identification based on digital image |
| CN113436138B (en) * | 2021-03-31 | 2023-07-21 | 成都飞机工业（集团）有限责任公司 | Image preprocessing method for aviation structure identification based on digital image |
| CN113792783A (en) * | 2021-09-13 | 2021-12-14 | 陕西师范大学 | Automatic identification method and system for dough mixing stage based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110378222B (en) | Method and device for detecting vibration damper target and identifying defect of power transmission line | |
| CN109509187B (en) | Efficient inspection algorithm for small defects in large-resolution cloth images | |
| Ping et al. | A deep learning approach for street pothole detection | |
| CN110322442A (en) | A kind of building surface crack detecting method based on SegNet | |
| CN110991444B (en) | License plate recognition method and device for complex scene | |
| CN105740910A (en) | Vehicle object detection method and device | |
| CN109671071A (en) | A kind of underground piping defect location and grade determination method based on deep learning | |
| CN109840483B (en) | Landslide crack detection and identification method and device | |
| CN113240691A (en) | Medical image segmentation method based on U-shaped network | |
| CN110060251A (en) | A kind of building surface crack detecting method based on U-Net | |
| CN108681689B (en) | Frame rate enhanced gait recognition method and device based on generation of confrontation network | |
| CN110390673A (en) | Cigarette automatic testing method based on deep learning under a kind of monitoring scene | |
| CN111353396A (en) | Concrete crack segmentation method based on SCSEOCUnet | |
| CN110689000A (en) | Vehicle license plate identification method based on vehicle license plate sample in complex environment | |
| CN116206185A (en) | Lightweight small target detection method based on improved YOLOv7 | |
| CN113610024B (en) | Multi-strategy deep learning remote sensing image small target detection method | |
| CN116486246A (en) | Intelligent recognition method for bridge underwater image diseases based on convolutional neural network | |
| CN113902792A (en) | Building height detection method and system based on improved RetinaNet network and electronic equipment | |
| CN110135423A (en) | The training method and optical character recognition method of text identification network | |
| CN108711150A (en) | A kind of end-to-end pavement crack detection recognition method based on PCA | |
| CN117437201A (en) | Road crack detection method based on improved YOLOv7 | |
| CN105354833B (en) | A kind of method and apparatus of shadow Detection | |
| CN115619719A (en) | Pine wood nematode infected wood detection method based on improved Yolo v3 network model | |
| CN115330743A (en) | Method for detecting defects based on double lights and corresponding system | |
| CN116152674A (en) | Dam unmanned aerial vehicle image crack intelligent recognition method based on improved U-Net model |
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: 20191011 |
|
| RJ01 | Rejection of invention patent application after publication |