CN111583130A - Method for recovering license plate image for LPR - Google Patents
Method for recovering license plate image for LPR Download PDFInfo
- Publication number
- CN111583130A CN111583130A CN202010291139.1A CN202010291139A CN111583130A CN 111583130 A CN111583130 A CN 111583130A CN 202010291139 A CN202010291139 A CN 202010291139A CN 111583130 A CN111583130 A CN 111583130A
- Authority
- CN
- China
- Prior art keywords
- network
- license plate
- training
- image
- lpr
- 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 35
- 238000012549 training Methods 0.000 claims abstract description 61
- 238000011084 recovery Methods 0.000 claims abstract description 35
- 238000012360 testing method Methods 0.000 claims abstract description 34
- 238000012795 verification Methods 0.000 claims abstract description 16
- 230000000694 effects Effects 0.000 claims abstract description 8
- 230000011218 segmentation Effects 0.000 claims description 19
- 238000012937 correction Methods 0.000 claims description 13
- 230000009467 reduction Effects 0.000 claims description 10
- 238000005516 engineering process Methods 0.000 claims description 7
- 208000011580 syndromic disease Diseases 0.000 claims description 6
- 238000012935 Averaging Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 2
- 230000006870 function Effects 0.000 description 12
- 238000011176 pooling Methods 0.000 description 9
- 238000013527 convolutional neural network Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000010200 validation analysis Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000009966 trimming Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000004438 eyesight Effects 0.000 description 1
- 230000004313 glare Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/73—Deblurring; Sharpening
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for recovering a license plate image for LPR, which comprises the following steps: after a series of operations are carried out on the images in the known data set, the images are proportionally divided into a training set, a verification set and a test set; establishing a model of an image recovery network for license plate recognition, and training the model by using a training set to obtain a corresponding training model; the verification set is used for checking the accuracy of the training model, so that the hyper-parameters of the model are adjusted, and the model is optimized to obtain better performance; and inputting the test set image into the determined optimal model, testing the generalization performance of the test set image, and observing the recovery effect of the license plate image. The scheme redesigns the structure of the license plate image recovery network, increases the auxiliary network to optimize the recovery quality of the image, and obviously increases the robustness of the LPR; in addition, a good effect is obtained by a method of combining denoising and correcting networks, so that the license plate recognition accuracy is quite high, and the license plate recognition network is a fast and accurate recognition network.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method for recovering a license plate image for LPR.
Background
Along with the increasing of the national economic strength, the living standard of people is also greatly improved, and more families have own private cars. The accompanying traffic problems are increasing, and the problem of traffic vehicle management is one of the important problems in city management nowadays. For this reason, an Intelligent Transportation System (ITS) has been developed.
ITS is a technology that combines existing scientific technologies (such as computer counting, sensor technology, image processing technology, etc.) for transportation, service control, etc., and License Plate Recognition (LPR) is an important basic link. The current common method is to locate the license plate area by using the color and texture characteristics of the license plate, then to perform appropriate enhancement processing on the image, and then to recognize the image to obtain the license plate information.
Due to the development and application of convolutional neural networks (CNN for short), tasks in many computer vision fields are greatly developed, and meanwhile, many LPR methods based on CNN are also applied to solve and recognize license plate images in the real world. However, many existing methods are based on capturing high quality images. It is still difficult to identify images of lower quality in the face of blurred, oblique, etc. images captured in harsh environments (e.g., glare, night, haze, etc.), or captured in motion.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a license plate image recovery method with strong robustness, high identification accuracy and good performance.
The purpose of the invention is realized by the following technical scheme:
a method for recovering a license plate image for LPR mainly comprises the following specific steps:
step S1: after a series of operations are performed on the images in the known data set, the images are proportionally divided into a training set, a validation set and a test set.
Specifically, the step S1 further includes the following steps:
step S11: adopting a plurality of famous license plate recognition data sets VTLP, and carrying out verification and test on a training set, a verification set and a test set according to the following steps of 6:2:2 to a ratio of 2.
Step S12: in order to increase the amount of training data, four sub-pictures are generated by adopting rotation with different angles on the training set, and the four sub-pictures are doubled by a size conversion and segmentation method; the original training picture is marked as IHThe divided four rotated subgraphs are
i ∈ { -30 °, -15 °, +15 °, +30 ° }, and the size-transformed subgraph is marked as
The picture after binary segmentation by pixel is recorded as
The character count value is C.
In a preferred embodiment of the present invention, in step S1, the series of operations includes averaging, defogging and trimming operations.
Step S2: and establishing a model of an image recovery network for license plate recognition, and training the model by using a training set to obtain a corresponding training model.
Specifically, the step S2 further includes the following steps:
s21: setting a main network and an auxiliary network of a recovery network, wherein the main network comprises two sub-networks, and the auxiliary network comprises two decoder modules and then respectively trains the sub-networks and the modules;
s22: training noise reduction subnetwork MD；
S23: training the syndrome network MR；
S24: training pixel segmentation module AS；
S25: training character technology module AC；
S26: the four loss function weights are added.
Step S3: and the verification set is used for checking the accuracy of the training model, so that the hyper-parameters of the model are adjusted, and the model is optimized to obtain better performance.
Further, the step S3 further includes: after obtaining the training model through step S2, the LPR network can obtain the output image result provided by the correction module through connection with the syndrome network
And then adjusting the hyper-parameters of the model.
Step S4: and inputting the test set image into the determined optimal model, testing the generalization performance of the test set image, and observing the recovery effect of the license plate image.
Further, the step S4 further includes: providing the test set picture to a license plate recognition network, and respectively obtaining recognition results LPR through a license plate image recovery network and an LPR networkresult。
The working process and principle of the invention are as follows: the method for recovering the license plate image for the LPR redesigns the structure of the license plate image recovery network, increases the auxiliary network to optimize the recovery quality of the image, and obviously increases the robustness of the LPR; in addition, the recovery network provided by the scheme adopts a method of combining denoising and correction networks, a good effect is obtained, and the LPR network adopts a detector which is high in accuracy and rapid in identification at present, so that the accuracy of license plate identification is quite high, and the LPR network is a rapid and accurate identification network.
Compared with the prior art, the invention also has the following advantages:
(1) the method for recovering the license plate image for the LPR can well process and recover the license plate image with low quality, and the recovered image can be accurately identified by the current popular LPR network, so that the method is suitable for the license plate image obtained by capturing under different environments.
(2) The method for recovering the license plate image for the LPR, provided by the invention, aims at low-quality images, recovers the images through a series of end-to-end convolutional neural networks based on deep learning, and then takes the recovered images to the LPR network for recognition, so that interference factors can be obviously reduced, and the license plate can be quickly and accurately recognized.
(3) The method for recovering the license plate image for the LPR is designed aiming at the structure of a license plate image recovery network, an auxiliary network is added to optimize the recovery quality of the image, and the robustness during license plate recognition can be enhanced. The recovery network of the invention combines denoising and correction networks, and maximizes the quality of the recovered image by the structure of the auxiliary network, which is unique and innovative.
Drawings
Fig. 1 is a flowchart of a method for recovering a license plate image for LPR according to the present invention.
Fig. 2 is a schematic structural diagram of the entire license plate recognition network provided by the present invention.
FIG. 3 is a schematic structural diagram of a U-Net-based network provided by the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer and clearer, the present invention is further described below with reference to the accompanying drawings and examples.
Example 1:
as shown in fig. 1 to 3, the present embodiment discloses a method for recovering a license plate image for an LPR, which mainly includes the following specific steps:
step S1: after a series of operations are performed on the images in the known data set, the images are proportionally divided into a training set, a validation set and a test set.
Specifically, the step S1 further includes the following steps:
step S11: adopting a plurality of famous license plate recognition data sets VTLP, and carrying out verification and test on a training set, a verification set and a test set according to the following steps of 6:2:2 to a ratio of 2.
Step S12: in order to increase the amount of training data, four sub-pictures are generated by adopting rotation with different angles on the training set, and the four sub-pictures are doubled by a size conversion and segmentation method; the original training picture is marked as IHThe divided four rotated subgraphs are
i ∈ { -30 °, -15 °, +15 °, +30 ° }, and the size-transformed subgraph is marked as
The picture after binary segmentation by pixel is recorded as
The character count value is C.
In a preferred embodiment of the present invention, in step S1, the series of operations includes averaging, defogging and trimming operations.
Step S2: and establishing a model of an image recovery network for license plate recognition, and training the model by using a training set to obtain a corresponding training model.
Specifically, the step S2 further includes the following steps:
s21: setting a main network and an auxiliary network of a recovery network, wherein the main network comprises two sub-networks, and the auxiliary network comprises two decoder modules and then respectively trains the sub-networks and the modules;
s22: training noise reduction subnetwork MD；
S23: training the syndrome network MR；
S24: training pixel segmentation module AS；
S25: training character technology module AC；
S26: the four loss function weights are added.
Step S3: and the verification set is used for checking the accuracy of the training model, so that the hyper-parameters of the model are adjusted, and the model is optimized to obtain better performance.
Further, the step S3 further includes: after obtaining the training model through step S2, the LPR network can obtain the output image result provided by the correction module through connection with the syndrome network
And then adjusting the hyper-parameters of the model.
Step S4: and inputting the test set image into the determined optimal model, testing the generalization performance of the test set image, and observing the recovery effect of the license plate image.
Further, the step S4 further includes: providing the test set picture to a license plate recognition network, and respectively obtaining recognition results LPR through a license plate image recovery network and an LPR networkresult。
The working process and principle of the invention are as follows: the method for recovering the license plate image for the LPR redesigns the structure of the license plate image recovery network, increases the auxiliary network to optimize the recovery quality of the image, and obviously increases the robustness of the LPR; in addition, the recovery network provided by the scheme adopts a method of combining denoising and correction networks, a good effect is obtained, and the LPR network adopts a detector which is high in accuracy and rapid in identification at present, so that the accuracy of license plate identification is quite high, and the LPR network is a rapid and accurate identification network.
Example 2:
referring to fig. 1 to 3, the present embodiment discloses a method for restoring a license plate image for an LPR, including the following steps:
s1: after a series of operations such as averaging, defogging, cropping, etc., are performed on the images in the known dataset, the resulting picture size is 572 x 572, which scales the images into a training set, a validation set, and a test set.
The method comprises the following specific steps:
s11: a data set VTLP is adopted, 10650 license plate pictures are contained in the data set VTLP, and then the data set is divided into a training set, a verification set and a test set according to the ratio of 6:2:2, wherein the training set, the verification set and the test set respectively contain 6390 license plate pictures, 2130 license plate pictures and 2130 license plate pictures.
S12: then expanding the training set with correct labels, and obtaining a label graph IHExpansion into four subgraphs by rotation transformation
i ∈ { -30 °, -15 °, +15 °, +30 ° }, and after size transformation, four photons are generated
The subgraph after binary segmentation is recorded as
Marking the character label as C;
s2: establishing a model of an image recovery network for license plate recognition, and training the model by using a training set to obtain a corresponding training model, wherein the specific steps comprise the following steps:
s21: the recovery network comprises a backbone network and an auxiliary network, wherein the backbone network comprises two sub-networks, the first sub-network takes the low-quality image as input and outputs the low-quality image as a recovery image. The second is a correction network that corrects the output from the noise reduction network. The auxiliary network comprises a text counting module and a pixel segmentation module, the input of the text counting module and the pixel segmentation module is the sum of the encoder picture characteristics of the noise reduction network and the correction network, and the output of the text counting module and the pixel segmentation module is a picture after character number and binary segmentation respectively. The recovery network of the invention adopts a U-Net structure, the noise reduction network and the correction network both comprise an encoder module and a decoder module, and the counting network and the segmentation network only have the decoder module. Both sub-networks are built based on a U-Net structure, namely the U-Net structure as shown in the third figure, and the U-Net structure is selected because the U-Net structure can improve the detail information of the target in the image without negatively influencing the image generation. The structure based on U-Net is adopted, and jump connection is added, so that low-level semantic information of the image can be shared.
S211: both sub-networks of the backbone network comprise an encoder and a decoder, wherein the network structure of the encoder comprises a first convolution layer conv1(1,2) (3 × 3 × 32) × 2, step 2; the second convolution layer conv2(1,2) (3 × 3 × 64) × 2, step 2; the third convolution layer conv3(1,2) (3 × 3 × 128) × 2, step 2; the fourth convolution layer conv4(1,2) (3 × 3 × 256) × 2, step 2; the fifth convolution layer conv5(1,2) (3 × 3 × 512) × 2, step 2; first pooling layer pool1(2 × 2) max pooling, step size 2; second pooling layer pool2(2 × 2) max pooling, step size 2; third pooling layer pool3(2 × 2) max pooling, step size 2; fourth pooling layer pool4(2 × 2) max pooling, step size 2; it should be noted that a BN module is added before each convolutional layer, the activation functions connected after the convolutional layers are all LeakyReLU functions, the feature maps generated after the second convolutional kernel in each convolutional layer are led out to the feature maps after the upsampled convolutional layers in the decoder network through skip connection to be cascaded, and a corresponding maximum pooling layer is connected between every two convolutional layers.
S212: the network structure of the two decoders employed by the sub-networks of the backbone network and the decoder employed by the pixel segmentation module of the auxiliary network are the same. Wherein the decoder comprises: the sixth convolution layer conv6(1,2) (3 × 3 × 256) × 2, step 2; the seventh convolution layer conv7(1,2) (3 × 3 × 128) × 2, step 2; the eighth convolution layer conv8(1,2) (3 × 3 × 64) × 2, step 2; the ninth convolution layer conv9(1,2) (3 × 3 × 32) × 2, step 2; the tenth convolution layer conv10(1 × 1 × 3), step 1; a first upsampling layer upsamplable, concatenating con5_2 and conv4_ 2; a second up-sampling layer upsamplable, concatenating con6_2 and conv3_ 2; a third up-sampling layer upsamplable, namely con7_2 and conv2_2 are cascaded; the fourth upsampling layer upsamplable concatenates con8_2 and conv1_ 2. The upsampling layers are respectively connected between the two convolutional layers, and each convolutional layer is subjected to a LeakyReLU activation function.
S213: the text counting module of the auxiliary network uses CNN to make predictions of the number of texts to accomplish a simple classification task, wherein the decoder network structure comprises five convolutional layers of sizes (N × 512), (1 × 1 × 256), (1 × 1 × 128), (1 × 1 × 64), (1 × 1 × 1 × 1).
S214: the connection mode of the backbone network is as follows: the low-quality pictures input into the recovery network pass through the noise reduction sub-network, and then the output result is input into the correction sub-network. In this case, the encoders of the noise reduction network and the correction network each have an output, and the sum of the output characteristic maps of these two encoders is denoted as a characteristic set F and then fed to the counting decoder and the segmentation decoder in the auxiliary network, respectively. The text counting can accurately separate each text character of the license plate, and the number of texts is predicted, so that the picture is more suitable for a subsequent LPR network to detect. The image of the license plate can be clearer by pixel segmentation, and text recognition is easier to perform.
S22: sub-drawing
i ∈ { -30 °, -15 °, +15 °, +30 ° } to the noise reduction sub-network MDFor training, where w is a parameter of the noise reduction network, the loss function of this sub-network is represented as follows:
s23: sub-drawing
Feeding correction sub-network MRFor training, where w is a parameter of the correction sub-network, the loss function is expressed as follows:
s24: sub-drawing
Feeding a Pixel splitting decoder Module ASFor training, where F represents the feature set after the encoder sums of the two subnets of the main network,
indicating the probability that the certain pixel is a license plate region.
The cross entropy loss function formula showing the true classification of pixels is as follows:
s25: the count tag value C is fed to the pixel segmentation decoder module ACFor training, CpredIndicates the predicted value, CG,TRepresenting the tag value, the loss function is as follows:
s26: finally the loss function of the whole recovery network can be expressed as the sum of the weights of all sub-objective functions:
s27: training an optimal model according to the total loss function
S3: after each epoch is completed, the accuracy of the training model is checked by using a verification set, so that the hyper-parameters of the model are adjusted, and the model is optimized to obtain better performance, and the method comprises the following specific steps:
s31: inputting the license plate pictures of the verification set into a recovery network according to the training model obtained in the step S2 to obtain an output result
Then inputting the result into an LPR network for target detection, obtaining the accuracy result of the picture, and then properly adjusting the hyper-parameter optimization model according to the result.
S4: inputting the test set image into the determined optimal model, testing the generalization performance of the test set image, and observing how the recovery effect of the license plate image is, the specific steps comprise:
s41: picture I of test settestAgain input to the recovery network and the output result thereof input to the LPR network, and the resulting output can be expressed as：
LPRresult＝LPR(MR(MD(Itest)))
S42: and then comparing the accuracy of the judgment result with the existing SOTA license plate recognition network model. Summarizing, the excellent degree of the network performance of the invention is the optimal network in the network structure adopting the same data set at present, and the identification accuracy is up to 93%.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.
Claims (6)
1. A method for recovering a license plate image for LPR is characterized by comprising the following steps:
step S1: after a series of operations are carried out on the images in the known data set, the images are proportionally divided into a training set, a verification set and a test set;
step S2: establishing a model of an image recovery network for license plate recognition, and training the model by using a training set to obtain a corresponding training model;
step S3: the verification set is used for checking the accuracy of the training model, so that the hyper-parameters of the model are adjusted, and the model is optimized to obtain better performance;
step S4: and inputting the test set image into the determined optimal model, testing the generalization performance of the test set image, and observing the recovery effect of the license plate image.
2. The method for restoring a license plate image for LPR of claim 1, wherein said step of S1 further comprises the steps of:
step S11: adopting a plurality of famous license plate recognition data sets VTLP, and carrying out verification and test on a training set, a verification set and a test set according to the following steps of 6:2:2 to divide;
step S12: to increase the amount of training data, the pairsThe training set generates four sub-pictures by adopting rotation with different angles, and doubles the four sub-pictures by a size conversion and segmentation method; the original training picture is marked as IHThe divided four rotated subgraphs are
The subgraph after the size transformation is marked as
The picture after binary segmentation by pixel is recorded as
The character count value is C.
3. The method for restoring a license plate image for LPR of claim 1, wherein said step of S2 further comprises the steps of:
s21: setting a main network and an auxiliary network of a recovery network, wherein the main network comprises two sub-networks, and the auxiliary network comprises two decoder modules and then respectively trains the sub-networks and the modules;
s22: training noise reduction subnetwork MD；
S23: training the syndrome network MR；
S24: training pixel segmentation module AS；
S25: training character technology module AC；
S26: the four loss function weights are added.
4. The method for restoring a license plate image for LPR of claim 1, wherein said step of S3 further comprises: after obtaining the training model through step S2, the LPR network can obtain the output image result provided by the correction module through connection with the syndrome network
And then adjust the modelAnd (4) super-parameter.
5. The method for restoring a license plate image for LPR of claim 1, wherein said step of S4 further comprises: providing the test set picture to a license plate recognition network, and respectively obtaining recognition results LPR through a license plate image recovery network and an LPR networkresult。
6. The method for restoring a license plate image for LPR of claim 1, wherein said sequence of operations of step S1 comprises averaging, defogging and cropping.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010291139.1A CN111583130A (en) | 2020-04-14 | 2020-04-14 | Method for recovering license plate image for LPR |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010291139.1A CN111583130A (en) | 2020-04-14 | 2020-04-14 | Method for recovering license plate image for LPR |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111583130A true CN111583130A (en) | 2020-08-25 |
Family
ID=72111549
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010291139.1A Pending CN111583130A (en) | 2020-04-14 | 2020-04-14 | Method for recovering license plate image for LPR |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111583130A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111882004A (en) * | 2020-09-28 | 2020-11-03 | 北京易真学思教育科技有限公司 | Model training method, question judging method, device, equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100177982A1 (en) * | 2009-01-09 | 2010-07-15 | Sony Corporation | Image processing device, image processing method, program, and imaging device |
| CN110866561A (en) * | 2019-11-18 | 2020-03-06 | 佛山市南海区广工大数控装备协同创新研究院 | Plastic bottle color sorting method based on image recognition |
-
2020
- 2020-04-14 CN CN202010291139.1A patent/CN111583130A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100177982A1 (en) * | 2009-01-09 | 2010-07-15 | Sony Corporation | Image processing device, image processing method, program, and imaging device |
| CN110866561A (en) * | 2019-11-18 | 2020-03-06 | 佛山市南海区广工大数控装备协同创新研究院 | Plastic bottle color sorting method based on image recognition |
Non-Patent Citations (1)
| Title |
|---|
| YOUNKWAN LEE等: "SNIDER: Single Noisy Image Denoising and Rectification for Improving License Plate Recognition" * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111882004A (en) * | 2020-09-28 | 2020-11-03 | 北京易真学思教育科技有限公司 | Model training method, question judging method, device, equipment and storage medium |
| CN111882004B (en) * | 2020-09-28 | 2021-01-05 | 北京易真学思教育科技有限公司 | Model training method, question judging method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113052210B (en) | Rapid low-light target detection method based on convolutional neural network | |
| Wang et al. | Spatial attention for multi-scale feature refinement for object detection | |
| Chen et al. | Adaptive convolution for object detection | |
| CN112581409B (en) | Image defogging method based on end-to-end multiple information distillation network | |
| CN114155527A (en) | Scene text recognition method and device | |
| CN112990065B (en) | Vehicle classification detection method based on optimized YOLOv5 model | |
| CN112784834A (en) | Automatic license plate identification method in natural scene | |
| CN115205114A (en) | High-resolution image splicing improved algorithm based on ORB (object-oriented bounding box) features | |
| CN116503709A (en) | Vehicle detection method based on improved YOLOv5 in haze weather | |
| CN113591861B (en) | Image processing method, device, computing equipment and storage medium | |
| Tao et al. | F-PVNet: Frustum-level 3-D object detection on point–voxel feature representation for autonomous driving | |
| Song et al. | Robofusion: Towards robust multi-modal 3d obiect detection via sam | |
| CN111583130A (en) | Method for recovering license plate image for LPR | |
| CN114239667A (en) | Garbage detection classification and inference method based on deep learning | |
| CN116363535A (en) | Ship detection method in unmanned aerial vehicle aerial image based on convolutional neural network | |
| Guo et al. | Real-time pantograph anomaly detection using unsupervised deep learning and K-nearest neighbor classification | |
| CN112200840B (en) | Moving object detection system in visible light and infrared image combination | |
| Ghoneim et al. | Using super resolution to enhance license plates recognition accuracy | |
| US20240070808A1 (en) | Enhancement of coarse resolution imagery using super-resolution and convolution neural networks to be used for fit-for-purpose systematic land titling and land valuation | |
| Zhang et al. | Chinese license plate recognition using machine and deep learning models | |
| CN114387277A (en) | Method and device for training neural network to realize target segmentation based on event sequence | |
| CN110853040A (en) | Image collaborative segmentation method based on super-resolution reconstruction | |
| CN113160217A (en) | Method, device and equipment for detecting foreign matters in circuit and storage medium | |
| CN110826564A (en) | Small target semantic segmentation method and system in complex scene image | |
| CN110765864A (en) | Image pedestrian re-identification system and method based on resolution irrelevant features |
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 |