CN111986079A - Pavement crack image super-resolution reconstruction method and device based on generation countermeasure network - Google Patents
Pavement crack image super-resolution reconstruction method and device based on generation countermeasure network Download PDFInfo
- Publication number
- CN111986079A CN111986079A CN202010548216.7A CN202010548216A CN111986079A CN 111986079 A CN111986079 A CN 111986079A CN 202010548216 A CN202010548216 A CN 202010548216A CN 111986079 A CN111986079 A CN 111986079A
- Authority
- CN
- China
- Prior art keywords
- image
- resolution
- pavement crack
- super
- images
- 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
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000012549 training Methods 0.000 claims abstract description 16
- 238000005070 sampling Methods 0.000 claims abstract description 6
- 230000000007 visual effect Effects 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000002708 enhancing effect Effects 0.000 claims 1
- 238000013135 deep learning Methods 0.000 abstract description 6
- 238000001514 detection method Methods 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 15
- 230000000694 effects Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 4
- 238000002790 cross-validation Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 101100365548 Caenorhabditis elegans set-14 gene Proteins 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000003416 augmentation Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method and a device for reconstructing a pavement crack image super-resolution based on a generation countermeasure network, which are used for acquiring a high-resolution pavement crack image and performing down-sampling treatment to obtain the same number of low-resolution pavement crack images; pairing the low-resolution pavement crack images and the high-resolution pavement crack images in pairs, inputting the paired pavement crack images into a generation countermeasure network, and outputting the paired pavement crack images as a super-resolution image set; training to generate a confrontation network, optimizing to generate confrontation network parameters, and obtaining the optimized generated confrontation network as an image super-resolution reconstruction model. According to the method, the super-resolution reconstruction is performed on the pavement crack image by adopting a deep learning method, so that the accuracy of image reconstruction is improved; by improving the SRGAN method, the method is more suitable for the processed pavement crack image, is more natural in the representation of the detail texture of the reconstructed image, and can be directly used for pavement crack detection.
Description
Technical Field
The invention belongs to the field of pavement crack detection, and particularly relates to a method and a device for constructing a pavement crack image super-resolution reconstruction model based on a generated countermeasure network.
Background
In the collection process of the actual pavement crack image, the shooting equipment is usually interfered by a great amount of external noises such as movement speed, equipment quality and ambient light, and the collected pavement crack image often has the problem of poor image quality. Compared with a high-resolution image, the low-resolution image is very fuzzy in image quality, serious in information loss and not sharp enough in texture at the crack. If crack detection is performed directly by using a low-resolution picture, the final detection accuracy may not meet the actual requirements. In order to solve the hardware restriction of different imaging devices, a great deal of research is carried out by many scholars at home and abroad on how to improve the image resolution, and meanwhile, the method capable of increasing the inherent resolution of the image is called an image super-resolution reconstruction method. By the method, the accuracy of pavement crack identification can be improved while the restriction of hardware is removed, and an important reference value is provided for subsequent pavement maintenance.
At present, a great number of methods have been proposed in the field of image super-resolution reconstruction research, but the methods can be basically divided into two categories: one is the traditional non-deep learning method; another class is emerging deep learning approaches. The non-deep learning method also comprises methods based on interpolation, reconstruction, learning and the like. Based on the more classical bilinear interpolation, the adjacent interpolation and the like in the interpolation method, a high-resolution image is filled by simply calculating the low pixel value, and more image information is not added essentially, so that the visual effect is poor. Based on a more classical projection method, probability analysis and the like of a reconstruction method, a high-resolution image is reconstructed by iterating the prior information of the image. The method fully depends on image prior information, so that when the image resolution is low and the prior information is insufficient, the problem that the edge of the detail texture contains sawteeth still exists in a high-resolution image recovered by the method. Based on a classical sparse representation method, a neighborhood embedding method and the like of a learning method, prior information which is not contained in a low-resolution image is introduced through a machine learning strategy, compared with the former two methods, the reconstructed high-resolution image has a good effect, but a great difference still exists between the image details and the original image.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides a method and a device for reconstructing a pavement crack image super-resolution based on a generation countermeasure network, and solves the technical problem that the pavement crack image cannot be accurately and efficiently reconstructed in a super-resolution manner in the prior art.
In order to solve the technical problems, the invention adopts the following technical scheme:
a construction method of a pavement crack image super-resolution reconstruction model based on a generated countermeasure network comprises the following steps:
step 1, collecting N high-resolution pavement crack images as a target image set;
step 2, performing down-sampling treatment on the N high-resolution pavement crack images to obtain N low-resolution pavement crack images with the same quantity, wherein the N low-resolution pavement crack images are used as an image set to be reconstructed;
step 3, pairing the N low-resolution pavement crack images and the N high-resolution pavement crack images in pairs, and taking the N pairs of paired pavement crack images as an input image set;
step 4, training and generating a countermeasure network by taking the image set to be reconstructed in the input image set as input and the super-resolution image set as output;
the generation countermeasure network comprises a generator and a discriminator which are arranged in sequence;
and obtaining the optimized generation countermeasure network as an image super-resolution reconstruction model.
The invention also comprises the following technical characteristics:
specifically, the down-sampling processing of the high-resolution pavement crack image in the step 2 is to reduce the image size of the high-resolution pavement crack image by 4 times.
Specifically, the generator comprises 14 convolutional layers, wherein the 2 nd convolutional layer to the 11 th convolutional layer are combined in pairs to form a residual block, and a ReLU function is used as an activation function.
Specifically, the discriminator includes 8 convolutional layers, and the 1 st convolutional layer includes 64 convolutional kernels; then every 1 convolution layer, the number of convolution kernels is increased by 2 times until 512 convolution kernels of the last 8 th convolution layer; and the arbiter uses the LeakyReLU function as the activation function.
Specifically, when a confrontation network is generated in training, a loss function based on an L1 norm and a discriminator reconstruction error is designed, so that the competitiveness of a generator and a discriminator is balanced, and the visual quality of a generated image is enhanced;
assuming that the pixel-level error of the image follows a gaussian distribution, the L1 norm loss function has:
L(I)=|IHR-G(ILR)| (1)
in the formula IHRRepresenting a true high resolution pavement crack image, representing ILRCorresponding low-resolution pavement crack images, | | | represents calculation of L1 norm, and G represents a generator;
the loss function to be optimized and the updating rule of the generator and the discriminator are as follows:
wherein,
y=G(x;θD) (5)
in the formula (2-6), x represents a real high-resolution pavement crack image, z represents a corresponding low-resolution pavement crack image, y represents generation of a super-resolution pavement crack image,representing the penalty of the arbiter with respect to x and y, respectively, thetaG、θDRespectively representing relevant parameters of the generator and the discriminator; lambda [ alpha ]kDenotes the proportional gain of k, ktThe t-th updated value of k is represented, and the parameter k can balance the competitiveness of the generator and the discriminator; G. d represents a generator and a discriminator respectively; e [ L (G (z))]、E[L(x)]Representing the expected value of loss of the generated sample and the expected value of loss of the real sample, respectively, and gamma is the ratio of the expected value of loss of the generated sample and the expected value of loss of the real sample, which controls the trade-off between the diversity of the generated images and the visual quality.
A super-resolution reconstruction method based on a generated confrontation network pavement crack image comprises the following steps:
obtaining a low-resolution image to be reconstructed;
and inputting the low-resolution image to be reconstructed into the generated countermeasure network-based pavement crack image super-resolution reconstruction model obtained by the method for constructing the generated countermeasure network-based pavement crack image super-resolution reconstruction model, so as to obtain a target super-resolution image.
A pavement crack image super-resolution reconstruction device based on a generation countermeasure network comprises an image acquisition module and a reconstruction module;
the image acquisition module is used for acquiring a low-resolution image to be reconstructed;
the reconstruction module is used for inputting the low-resolution image to be reconstructed into the generated countermeasure network-based pavement crack image super-resolution reconstruction model obtained by the method for constructing the generated countermeasure network-based pavement crack image super-resolution reconstruction model, so as to obtain the target super-resolution image.
Compared with the prior art, the invention has the beneficial effects that:
1. the method adopts a deep learning method to carry out super-resolution reconstruction on the pavement crack image, and improves the accuracy of image reconstruction.
2. The invention can be more suitable for the road surface crack image processed by the invention by improving the SRGAN method, and the reconstructed image has more natural detailed texture expression, and can be directly used for detecting the road surface crack.
Drawings
FIG. 1 is a schematic process flow diagram;
FIG. 2 is a diagram of a network basic block;
FIG. 3 is a road surface crack image super-resolution reconstruction model training peak signal-to-noise ratio curve;
FIG. 4 is a comparison graph of road surface crack image super-resolution reconstruction results; wherein, (a) is an original high-resolution pavement crack image; (b) the method comprises the following steps of (1) adopting a Bicubic model to carry out super-resolution reconstruction on a pavement crack image; (c) the method is an effect diagram for performing super-resolution reconstruction on a pavement crack image by adopting an SRCNN model; (d) the method is an effect diagram for performing super-resolution reconstruction on a pavement crack image by adopting a DRCN model; (e) the method is an effect diagram for performing super-resolution reconstruction on a pavement crack image by adopting a traditional SRGAN model; (f) the invention is an effect diagram for performing super-resolution reconstruction on a pavement crack image by adopting the improved SRGAN model.
The details of the present invention are explained in further detail below with reference to the drawings and examples.
Detailed Description
In order to enable the reconstructed high-resolution image to be closer to the original high-resolution image, deep learning is introduced into the field of image super-resolution reconstruction, compared with a shallow learning method, the method can introduce more prior information, is prominent in noise reduction and deblurring, and the restored image can be recognized by more people in visual effect.
The following provides specific examples of the present invention, further illustrating the specific implementation process of the method of the present invention, to verify the beneficial effects of the present invention.
Example 1
The embodiment provides a method for constructing a super-resolution reconstruction model based on a generated confrontation network pavement crack image, which comprises the following steps of:
step 1, collecting N high-resolution pavement crack images as a target image set.
Step 2, performing down-sampling treatment on the N high-resolution pavement crack images to obtain N low-resolution pavement crack images with the same quantity, wherein the N low-resolution pavement crack images are used as an image set to be reconstructed;
the pavement crack image acquisition mode can adopt detection vehicle acquisition or intelligent mobile phone shooting;
in addition, the road surface image can be enlarged in modes of image overturning, different-angle rotation and the like.
In one embodiment, the sample data obtained after image augmentation collectively comprises 1000 high-resolution pavement crack images and 1000 low-resolution pavement crack images.
Step 3, pairing N low-resolution pavement crack images and N high-resolution pavement crack images in a one-to-one correspondence mode, and dividing the paired N pairs of pavement crack images into a training set and a cross validation set according to the proportion of 7: 3;
in this embodiment, the N pairs of pavement crack images may be further divided into a training set, a cross validation set, and a test set according to a ratio of 6:2:2, where the test set is used for an image reconstruction effect after the image super-resolution reconstruction model is trained.
Step 4, inputting an image set to be reconstructed in the input image set into a generation countermeasure network, and outputting the image set to be reconstructed as a super-resolution image set; the generation of the countermeasure network includes a generator and an arbiter arranged in sequence.
And 5, distinguishing the output super-resolution image set from the target image set to obtain an error value, performing reverse propagation on the error value, updating and iterating the weight in the generated countermeasure network, optimizing the parameters of the generated countermeasure network according to the error value, the target image set and the super-resolution image set, and obtaining the optimized generated countermeasure network as an image super-resolution reconstruction model.
FIG. 2 is a diagram of a network basic block;
the constructed super-resolution reconstruction model of the pavement crack image is an SRGAN model;
the SRGAN model comprises a generator and a discriminator; the generator includes 14 convolutional layers, wherein the 2 nd convolutional layer to the 11 th convolutional layer are combined two by two to form a residual block, and a ReLU function is used as an activation function.
The discriminator includes 8 convolutional layers, and the 1 st convolutional layer includes 64 convolutional kernels. Then every 1 convolutional layer, the number of convolution kernels is increased by 2 times until 512 convolution kernels of the last 8 th convolutional layer. And the arbiter uses a LeakyReLU function as an activation function;
in the early stage of the training process in the conventional GAN, the learning capability of the discriminator D far exceeds that of the generator G, thereby causing the problem of model collapse (model collapse). Therefore, in order to improve the problem of model collapse, a loss function based on the L1 norm and the reconstruction error of the discriminator is designed in network training, and the loss function can balance the competitiveness of a generator and the discriminator and can enhance the visual quality of a generated image.
Assuming that the pixel-level error of the image follows a gaussian distribution, the L1 norm loss function has:
L(I)=|IHR-G(ILR)| (1)
in the formula IHRRepresenting a true high resolution pavement crack image, representing ILRA corresponding low resolution pavement crack image;
the loss function to be optimized and the updating rule of the generator and the discriminator are as follows:
wherein,
y=G(x;θD) (5)
in the formula (2-6), x represents a real high-resolution pavement crack image, z represents a corresponding low-resolution pavement crack image, y represents generation of a super-resolution pavement crack image,representing the penalty of the arbiter with respect to x and y, respectively, thetaG、θDRepresenting the relevant parameters of the generator and the arbiter, respectively. Lambda [ alpha ]kDenotes the proportional gain of k, ktRepresenting the t-th updated value of k, the parameter k may balance the competitiveness of the generator and the arbiter. γ is the ratio of the expected loss of the generated sample to the expected loss of the real sample, which parameter can control the trade-off between the diversity of the generated image and the visual quality.
When an input image set (training set and cross-validation set) is input into a generation countermeasure network for training, and a SRGAN model is trained, an optimizer in the SRGAN model is set as an Adam optimizer, and when the number of times of training is less than 100, the learning rate is set to 0.01, when the number of times of training is 100 or more and less than 200, the learning rate is set to 0.001, and when the number of times of training is 200 or more, the learning rate is set to 0.0001.
As shown in fig. 3, a peak signal-to-noise ratio curve of the road surface crack image super-resolution reconstruction model under the embodiment is shown, and it can be seen that when the training reaches the 100 th time, the value of the peak signal-to-noise ratio basically tends to be stable, which indicates that the network training is completed.
Example 2
The embodiment provides a super-resolution reconstruction method based on a generated confrontation network pavement crack image, which comprises the following steps:
obtaining a low-resolution image to be reconstructed;
and inputting the low-resolution image to be reconstructed into a generated countermeasure network pavement crack image super-resolution reconstruction model obtained by a generated countermeasure network pavement crack image super-resolution reconstruction model construction method to obtain a target super-resolution image.
Calculating a peak signal-to-noise ratio of the image, comprising:
step 1, obtaining a mean square error MSE between the high-resolution image and the super-resolution image through a formula (7);
wherein m and n respectively represent the length and width of the image; i (I, j) and K (I, j) respectively represent pixel point values in the high-resolution image and the super-resolution image;
step 2, obtaining a peak signal-to-noise ratio PSNR through a formula (8);
wherein, MAXI 2Represents the maximum pixel value possible for the image; MSE represents the mean square error.
As shown in fig. 4, the comparison graph of the super-resolution reconstruction result of the pavement crack image is shown, wherein (a) is an acquired original high-resolution pavement crack image, (e) is an effect graph of performing super-resolution reconstruction of the pavement crack image by using a conventional SRGAN model, and (f) is an effect graph of performing super-resolution reconstruction of the image by using the method of the present invention (i.e., the improved SRGAN model of the present invention).
In the analysis of the image reconstruction result, the method, Bicubic, SRCNN, DRCN and SRGAN methods need to be compared with each other through standard evaluation indexes. The test was performed on sets 5, Set14, BSDS100 and 300 slit image sets, respectively, using a fixed 4-fold image magnification ratio, and the average PSNR values were calculated. The results of the different methods are shown in table 1. As can be seen from Table 1, the method performed well on each data set, with a PSNR value of 29.21dB on the crack data set.
TABLE 1 PSNR value comparison of different super-resolution methods on different data sets
Example 3
The embodiment provides a pavement crack image super-resolution reconstruction device based on a generation countermeasure network, which comprises an image acquisition module and a reconstruction module;
the image acquisition module is used for acquiring a low-resolution image to be reconstructed;
the reconstruction module is used for inputting the low-resolution image to be reconstructed into the generated countermeasure network-based pavement crack image super-resolution reconstruction model obtained by the generated countermeasure network-based pavement crack image super-resolution reconstruction model construction method, and obtaining the target super-resolution image.
It should be noted that the present invention is not limited to the above-mentioned embodiments, and based on the technical solutions disclosed in the present invention, those skilled in the art can make some substitutions and modifications to some technical features without creative efforts, and these substitutions and modifications are all within the protection scope of the present invention.
Claims (7)
1. A method for constructing a super-resolution reconstruction model based on a generated confrontation network pavement crack image is characterized by comprising the following steps:
step 1, collecting N high-resolution pavement crack images as a target image set;
step 2, performing down-sampling treatment on the N high-resolution pavement crack images to obtain N low-resolution pavement crack images with the same quantity, wherein the N low-resolution pavement crack images are used as an image set to be reconstructed;
step 3, pairing the N low-resolution pavement crack images and the N high-resolution pavement crack images in pairs, and taking the N pairs of paired pavement crack images as an input image set;
step 4, training and generating a countermeasure network by taking the image set to be reconstructed in the input image set as input and the super-resolution image set as output;
the generation countermeasure network comprises a generator and a discriminator which are arranged in sequence;
and obtaining the optimized generation countermeasure network as an image super-resolution reconstruction model.
2. The method for constructing the super-resolution reconstruction model based on the generated countermeasure network pavement crack image as claimed in claim 1, wherein the down-sampling of the high-resolution pavement crack image in the step 2 is to reduce the image size of the high-resolution pavement crack image by a factor of 4.
3. The method for constructing the super-resolution reconstruction model of the pavement crack image based on the generation countermeasure network as claimed in claim 1, wherein the generator comprises 14 convolutional layers, wherein the 2 nd convolutional layer to the 11 th convolutional layer are combined in pairs to form a residual block, and a ReLU function is used as an activation function.
4. The method for constructing the super-resolution reconstruction model of the pavement crack image based on the generation countermeasure network according to claim 1, wherein the discriminator comprises 8 convolution layers, and the 1 st convolution layer comprises 64 convolution kernels; then every 1 convolution layer, the number of convolution kernels is increased by 2 times until 512 convolution kernels of the last 8 th convolution layer; and the arbiter uses the LeakyReLU function as the activation function.
5. The method for constructing the super-resolution reconstruction model of the pavement crack image based on the generation countermeasure network as claimed in claim 1, wherein when the generation countermeasure network is trained, a loss function based on an L1 norm and a reconstruction error of a discriminator is designed for balancing the competitiveness of a generator and the discriminator and enhancing the visual quality of the generated image;
assuming that the pixel-level error of the image follows a gaussian distribution, the L1 norm loss function has:
L(I)=|IHR-G(ILR)| (1)
in the formula IHRRepresenting a true high resolution pavement crack image, representing ILRCorresponding low-resolution pavement crack images, | | | represents calculation of L1 norm, and G represents a generator;
the loss function to be optimized and the updating rule of the generator and the discriminator are as follows:
wherein,
y=G(x;θD) (5)
in the formula (2-6), x represents a real high-resolution pavement crack image, z represents a corresponding low-resolution pavement crack image, y represents generation of a super-resolution pavement crack image,representing the penalty of the arbiter with respect to x and y, respectively, thetaG、θDRespectively representing relevant parameters of the generator and the discriminator; lambda [ alpha ]kDenotes the proportional gain of k, ktThe t-th updated value of k is represented, and the parameter k can balance the competitiveness of the generator and the discriminator; G. d represents a generator and a discriminator respectively; e [ L (G (z))]、E[L(x)]Representing the expected value of loss of the generated sample and the expected value of loss of the real sample, respectively, and gamma is the ratio of the expected value of loss of the generated sample and the expected value of loss of the real sample, which controls the trade-off between the diversity of the generated images and the visual quality.
6. A super-resolution reconstruction method based on a generated confrontation network pavement crack image is characterized by comprising the following steps:
obtaining a low-resolution image to be reconstructed;
inputting a low-resolution image to be reconstructed into the super-resolution reconstruction model based on the generated confrontation network pavement crack image, which is obtained by the super-resolution reconstruction model based on the generated confrontation network pavement crack image construction method according to any one of claims 1 to 5, and obtaining a target super-resolution image.
7. A pavement crack image super-resolution reconstruction device based on a generation countermeasure network is characterized by comprising an image acquisition module and a reconstruction module;
the image acquisition module is used for acquiring a low-resolution image to be reconstructed;
the reconstruction module is used for inputting the low-resolution image to be reconstructed into the super-resolution reconstruction model based on the generated confrontation network pavement crack image, which is obtained by the super-resolution reconstruction model based on the generated confrontation network pavement crack image construction method according to any one of claims 1 to 5, so as to obtain the target super-resolution image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010548216.7A CN111986079A (en) | 2020-06-16 | 2020-06-16 | Pavement crack image super-resolution reconstruction method and device based on generation countermeasure network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010548216.7A CN111986079A (en) | 2020-06-16 | 2020-06-16 | Pavement crack image super-resolution reconstruction method and device based on generation countermeasure network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111986079A true CN111986079A (en) | 2020-11-24 |
Family
ID=73442246
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010548216.7A Pending CN111986079A (en) | 2020-06-16 | 2020-06-16 | Pavement crack image super-resolution reconstruction method and device based on generation countermeasure network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111986079A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113009447A (en) * | 2021-03-05 | 2021-06-22 | 长安大学 | Road underground cavity detection early warning method based on deep learning and ground penetrating radar |
CN116883362A (en) * | 2023-07-12 | 2023-10-13 | 四川大学工程设计研究院有限公司 | Crack detection method and system based on image recognition and image processing equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180075581A1 (en) * | 2016-09-15 | 2018-03-15 | Twitter, Inc. | Super resolution using a generative adversarial network |
CN110136063A (en) * | 2019-05-13 | 2019-08-16 | 南京信息工程大学 | A kind of single image super resolution ratio reconstruction method generating confrontation network based on condition |
CN110211045A (en) * | 2019-05-29 | 2019-09-06 | 电子科技大学 | Super-resolution face image method based on SRGAN network |
CN110889370A (en) * | 2019-11-26 | 2020-03-17 | 上海大学 | System and method for generating end-to-end side face synthesis front face of countermeasure network based on conditions |
-
2020
- 2020-06-16 CN CN202010548216.7A patent/CN111986079A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180075581A1 (en) * | 2016-09-15 | 2018-03-15 | Twitter, Inc. | Super resolution using a generative adversarial network |
CN110136063A (en) * | 2019-05-13 | 2019-08-16 | 南京信息工程大学 | A kind of single image super resolution ratio reconstruction method generating confrontation network based on condition |
CN110211045A (en) * | 2019-05-29 | 2019-09-06 | 电子科技大学 | Super-resolution face image method based on SRGAN network |
CN110889370A (en) * | 2019-11-26 | 2020-03-17 | 上海大学 | System and method for generating end-to-end side face synthesis front face of countermeasure network based on conditions |
Non-Patent Citations (1)
Title |
---|
丁明航, 邓然然, 邵 恒: "基于注意力生成对抗网络的图像超分辨率重建方法", 计算机***应用, pages 5 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113009447A (en) * | 2021-03-05 | 2021-06-22 | 长安大学 | Road underground cavity detection early warning method based on deep learning and ground penetrating radar |
CN116883362A (en) * | 2023-07-12 | 2023-10-13 | 四川大学工程设计研究院有限公司 | Crack detection method and system based on image recognition and image processing equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112734646B (en) | Image super-resolution reconstruction method based on feature channel division | |
CN111062872A (en) | Image super-resolution reconstruction method and system based on edge detection | |
CN107464217B (en) | Image processing method and device | |
CN110675321A (en) | Super-resolution image reconstruction method based on progressive depth residual error network | |
CN110796622B (en) | Image bit enhancement method based on multi-layer characteristics of series neural network | |
CN112419150B (en) | Image super-resolution reconstruction method of arbitrary multiple based on bilateral upsampling network | |
CN111932461A (en) | Convolutional neural network-based self-learning image super-resolution reconstruction method and system | |
CN111861886B (en) | Image super-resolution reconstruction method based on multi-scale feedback network | |
CN112435162B (en) | Terahertz image super-resolution reconstruction method based on complex domain neural network | |
CN116029902A (en) | Knowledge distillation-based unsupervised real world image super-resolution method | |
CN111583113A (en) | Infrared image super-resolution reconstruction method based on generation countermeasure network | |
CN111986079A (en) | Pavement crack image super-resolution reconstruction method and device based on generation countermeasure network | |
CN112669214A (en) | Fuzzy image super-resolution reconstruction method based on alternative direction multiplier algorithm | |
CN113538246A (en) | Remote sensing image super-resolution reconstruction method based on unsupervised multi-stage fusion network | |
CN114897694A (en) | Image super-resolution reconstruction method based on mixed attention and double-layer supervision | |
CN112950480A (en) | Super-resolution reconstruction method integrating multiple receptive fields and dense residual attention | |
CN103020905B (en) | For the sparse constraint adaptive N LM super resolution ratio reconstruction method of character image | |
CN117274059A (en) | Low-resolution image reconstruction method and system based on image coding-decoding | |
CN115526777A (en) | Blind over-separation network establishing method, blind over-separation method and storage medium | |
CN116051382A (en) | Data enhancement method based on deep reinforcement learning generation type antagonistic neural network and super-resolution reconstruction | |
CN112184552B (en) | Sub-pixel convolution image super-resolution method based on high-frequency feature learning | |
CN113096015A (en) | Image super-resolution reconstruction method based on progressive sensing and ultra-lightweight network | |
CN112070669A (en) | Super-resolution image reconstruction method for any fuzzy kernel | |
CN111524060A (en) | System, method, storage medium and device for blurring portrait background in real time | |
CN114519670A (en) | Image super-resolution reconstruction method based on convolutional neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |