CN113435446A - Inclined license plate correction method based on deep learning - Google Patents
Inclined license plate correction method based on deep learning Download PDFInfo
- Publication number
- CN113435446A CN113435446A CN202110767061.0A CN202110767061A CN113435446A CN 113435446 A CN113435446 A CN 113435446A CN 202110767061 A CN202110767061 A CN 202110767061A CN 113435446 A CN113435446 A CN 113435446A
- Authority
- CN
- China
- Prior art keywords
- license plate
- layer
- size
- corner
- local image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000013135 deep learning Methods 0.000 title claims abstract description 20
- 238000012937 correction Methods 0.000 title claims abstract description 18
- 238000001514 detection method Methods 0.000 claims abstract description 20
- 230000009466 transformation Effects 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 12
- 238000012800 visualization Methods 0.000 claims abstract description 7
- 239000010410 layer Substances 0.000 claims description 98
- 239000002356 single layer Substances 0.000 claims description 13
- 238000010586 diagram Methods 0.000 claims description 9
- 238000002372 labelling Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 238000012935 Averaging Methods 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a deep learning-based inclined license plate correction method, which comprises the following steps: step 1, obtaining a local image of a license plate, and constructing a visualization tool for marking a corner point of the license plate; step 2, constructing a target detection network model based on deep learning, and training the model; step 3, constructing a full convolution license plate type classification network and training the model; step 4, detecting the license plate, acquiring a local image of the license plate, and detecting license plate corner points in the local image by using the model obtained in the step 2 for positioning; step 5, judging the license plate type by using the license plate type classification network obtained in the step 3; and 6, performing corresponding perspective transformation according to the license plate type, adjusting the detected corner coordinates to a fixed value, and adjusting the length-width ratio of the corner coordinates to be close to the standard condition. The method has better robustness in a complex scene, can calculate the corrected coordinate value more accurately, improves the character recognition rate and enables the license plate recognition to be more accurate.
Description
Technical Field
The invention belongs to the technical field of intelligent transportation, and particularly relates to a method for correcting a tilted license plate based on deep learning.
Background
License plate recognition is an important component of intelligent traffic, the technology has been widely applied in various scenes, and inclined license plate correction still needs to be optimized for the traditional three-stage license plate recognition technology and the latest license plate recognition technology based on deep learning. The problem to be solved urgently in the field is to find an efficient and accurate method for correcting the inclined license plate. Due to the problem of shooting angle (light and environment), the license plate image in a natural scene is often collected to be a trapezoid, and the inclination of the license plate easily influences the accuracy of subsequent recognition. The license plates in China have different colors and single-layer and double-layer characters, so that the identification is more difficult. The traditional preprocessing detection and correction method cannot better cope with various complex scenes, and easily influences the license plate recognition rate. Therefore, the method has important practical significance for accurately correcting the license plate.
Disclosure of Invention
Based on the technical problems in the background art, the invention provides a method for correcting a tilted license plate based on deep learning. In order to conveniently and accurately mark the corner point position, a visual tool for marking the corner point of the license plate based on the MFC is designed, and the tool makes relevant constraints on the size and the shape of a marking frame; in order to realize license plate positioning with better effect, a license plate corner detection model based on a YOLO algorithm is designed; and finally, designing a full convolution network, and judging the license plate type by using the network. And performing angular point detection on the local license plate image obtained by license plate detection, and performing corresponding perspective transformation according to the type of the license plate, wherein the aspect ratio of the corrected license plate is closer to a standard value.
In order to achieve the purpose, the invention adopts the following technical scheme:
a deep learning-based inclined license plate correction method comprises the following steps:
step 1, obtaining a local image of a license plate, and constructing a visualization tool for marking a corner point of the license plate;
step 2, constructing a target detection network model based on deep learning, and training the model;
step 3, constructing a full convolution license plate type classification network and training the model;
step 4, detecting the license plate, acquiring a local image of the license plate, and detecting license plate corner points in the local image by using the model obtained in the step 2 for positioning;
step 5, judging the license plate type by using the license plate type classification network obtained in the step 3;
and 6, performing corresponding perspective transformation according to the license plate type, adjusting the detected corner coordinates to a fixed value, and adjusting the length-width ratio of the corner coordinates to be close to the standard condition.
In order to optimize the technical scheme, the specific measures adopted further comprise:
furthermore, a traditional license plate detector is arranged in the visualization tool for marking the corner points of the license plate in the step 1, and the marking frame of each corner point is a square; and automatically calculating the size of the labeling frame according to the coordinates of the local image of the license plate.
Furthermore, the size of the local image of the license plate is fixed to be 500 x 500, the size of the labeling frame is two thirds of the height of the license plate, and 4 angular points are labeled for each training image.
Further, the target detection network in step 2 is a fully-convoluted neural network, and angular points are detected on feature maps of two different sizes.
Further, the target detection network model based on deep learning constructed in step 2 is a corner detection model in YOLOv 3.
Further, the target detection network based on deep learning has 20 layers; the input of the network is a 3-channel image of 224 × 224, and the 1 st to 11 th layers are convolution layers with a convolution kernel size of 3 × 3; the 13 th layer is a yolo layer, and targets are detected on a 7 x 7 size feature map; the 14 th layer is a route layer and is used for acquiring a characteristic diagram of the 11 th layer; the 15 th layer is a convolution layer with convolution kernel size of 1 multiplied by 1; the 16 th layer is an upsampling layer, and the size of the feature map is changed into 14 multiplied by 14; the 17 th layer is a route layer and is used for splicing the characteristic diagrams of the 7 th layer and the 16 th layer; layer 18 is a convolutional layer with a convolutional kernel size of 3 × 3; the 20 th layer is the yolo layer, and targets are detected on a feature map of 14 × 14 size.
Further, the license plate type classification network trained in step 3 is used for classifying the license plate types into three types:
a) the single-layer new energy license plate comprises 8 characters;
b) the single-layer license plate containing 7 characters comprises a blue license plate, a yellow license plate and a police car license plate which are common on a car head;
c) the double-layer character license plate comprises a yellow license plate and a trailer license plate.
Furthermore, the license plate type classification network comprises a full convolution network of 8 convolution layers, a global average pooling layer, a full connection layer and a probability output layer; the network input size is 96 × 96 × 3, each convolutional layer uses a convolutional kernel of size 3 × 3, and the activation function is Relu.
Further, in step 6, the license plate categories are classified according to the detected angular point coordinates to perform perspective transformation, and the angular point coordinates are transformed into fixed coordinate points and the length-width ratio is adjusted to be close to the standard condition.
Further, the perspective transformation process in step 6 is as follows:
calculating 3 x 3 transformation matrix according to 4 pairs of license plate corner coordinatesM:
the calculation mode of the corrected license plate local image is as follows:
wherein ,srcthe local image of the license plate before correction is obtained;dstthe corrected local image of the license plate is obtained;M 11…M 33for transforming matricesMElements of the corresponding position.
The invention has the beneficial effects that: the traditional correction methods such as Hough transform, Randon transform and the like are difficult to cope with various complex scenes such as yin-yang license plates, worn license plates and the like, and the angular point positioning method based on deep learning target detection has better robustness in the complex scenes.
The traditional partial correction algorithm does not consider the problems of single-layer and double-layer and character number, namely how to calculate the corrected coordinate values more accurately, the double-layer license plate and the new energy license plate are common in an actual scene, the new coordinate values can influence the length-width ratio of the license plate and the character recognition rate in the later period, and the license plate type judgment network designed based on the method can calculate the corrected coordinate values more accurately, improve the character recognition rate and enable the license plate to be recognized more accurately.
Drawings
FIG. 1 is a schematic diagram of a local image of a license plate in the method for correcting an inclined license plate according to the present invention;
FIG. 2 is a schematic diagram of corner point labeling in the license plate number correction method according to the present invention;
FIG. 3 is a schematic diagram of a local image of a license plate after correction in the method for correcting an inclined license plate according to the present invention;
fig. 4 is a flow chart of an embodiment of the present invention.
Detailed Description
The present invention will now be described in further detail with reference to the accompanying drawings.
It should be noted that the terms "upper", "lower", "left", "right", "front", "back", etc. used in the present invention are for clarity of description only, and are not intended to limit the scope of the present invention, and the relative relationship between the terms and the terms is not limited by the technical contents of the essential changes.
The invention provides a deep learning-based inclined license plate correction method, which comprises the following steps:
step 1, obtaining a local image of a license plate, and constructing a visualization tool based on MFC (micro-fluidic controller) labeling license plate corner points, wherein a traditional license plate detector is arranged in the visualization tool, and a labeling frame of each corner point is a square; and automatically calculating the size of the labeling frame according to the coordinates of the local image of the license plate and displaying the geometric center point of the labeling frame. Wherein, the size of the local image of the license plate is fixed to be 500And 500, marking 4 corner points for each training image by using the tool, wherein the size of the marking frame is two thirds of the height of the license plate.
Step 2, constructing a target detection network model based on deep learning, and training the model;
a corner detection model of YOLOv3 was designed and trained using 10 ten thousand labeled samples. The initial learning rate was set to 0.0075, and the number of iterations was set to 56 ten thousand. The calculated amount of the whole network is 0.035BFlops, and the time consumption is about 1.4 milliseconds when one reasoning is carried out on the Invitta RTX 2080.
The target detection network based on deep learning has 20 layers (including two yolo layers), the input of the network is a 3-channel image of 224 × 224, the 1 st to 11 th layers are convolution layers with a convolution kernel size of 3 × 3, wherein the step size of the 1 st, 2 nd, 4 th, 6 th and 8 th layers is 2, and the step size of the rest layers is 1. Layer 13 is the yolo layer and targets are detected on a 7 × 7 feature map. And the 14 th layer is a route layer and is used for acquiring the 11 th layer characteristic diagram. The 15 th layer is a convolution layer with a convolution kernel size of 1 × 1, and the number of channels of the output feature map is changed to 16 in order to reduce the amount of calculation. The 16 th layer is an upsampling layer, which changes the feature map size to 14 × 14. And the 17 th layer is a route layer and is used for splicing the characteristic diagrams of the 7 th layer and the 16 th layer. Layer 18 is a convolutional layer with a convolutional kernel size of 3 × 3. The 20 th layer is the yolo layer, and targets are detected on a feature map of 14 × 14 size. The number of convolution kernels of the 1 st layer to the 11 th layer is 4, 8, 16, 24, 32, 48 in sequence, and the 12 th layer and the 19 th layer are convolution layers with convolution kernel size of 1 × 1, and since the network only detects one type of object (license plate corner), the number of convolution kernels of the two layers is 18.
And 3, constructing a full convolution license plate type classification network and training the model.
A full-convolution license plate type classifier is designed to classify license plates containing single-layer and double-layer characters and single-layer indefinite-length characters, and license plate images are divided into double-layer character license plates, single-layer license plates containing 7 characters and single-layer new energy license plates containing 8 characters. The number of training samples in each category is 20 ten thousand, the initial learning rate is set to be 0.003, the weight attenuation is set to be 0.0005, the momentum is 0.9, the maximum iteration number is 100 ten thousand, and the license plate type identification work can be completed by performing inference on the great RTX2080, which takes about 1.3 milliseconds.
The license plate type classification network is a full convolution network comprising 8 convolution layers, a global average pooling layer, a full connection layer and a probability output layer. The network input size is 96 × 96 × 3, each convolutional layer uses a convolutional kernel of size 3 × 3, and the activation function is Relu. Wherein, the step size of the 1 st, 2 nd, 4 th and 6 th layers is set as 2, and the step size of the rest layers is set as 1. The number of convolution kernels of the 1 st layer to the 8 th layer is 4, 8, 16, 24, 32 and 32 in sequence.
And 4, detecting the license plate, acquiring a local image of the license plate, and detecting license plate corner points in the local image by using the model obtained in the step 2 for positioning as shown in the figure 1. The detected license plate corner image is shown in fig. 2, wherein the frame of the detected corner is a normal rectangle, but the geometric center is substantially at the position of the corner.
And 5, judging the license plate type by using the license plate type classification network obtained in the step 3.
And 6, performing corresponding perspective transformation according to the license plate type, adjusting the detected corner coordinates to a fixed value, and adjusting the length-width ratio of the corner coordinates to be close to the standard condition.
The formula is as follows:
calculating 3 x 3 transformation matrix according to 4 pairs of license plate corner coordinatesM:
the calculation mode of the corrected license plate local image is as follows:
wherein ,srcthe local image of the license plate before correction is obtained;dstthe corrected local image of the license plate is obtained;M 11…M 33for transforming matricesMElements of the corresponding position.
Since the size of the license plate partial image which is not corrected is normalized to 500 × 500, and the license plate is approximately located at the geometric center of the image, the coordinate point after perspective transformation can be set to be a fixed value which is also located at the geometric center, and the corrected license plate partial image is as shown in fig. 3. Wherein, the time for making perspective transformation on the CPU is about 2.8 milliseconds. The CPU has a model number of Intel (R) Xeon (R) Gold 61482.40 GHz. The specific correction process is as follows:
(a) in step 2, coordinates of four corner points are detectedAnd performing reasoning once by using the license plate classifier to obtain the type of the license plate. Will be 500X 500 in sizeAnd carrying out coordinate transformation on the local image of the license plate according to types.
(b) The license plate type is a single-layer new energy license plate containing 8 characters: the coordinates are sequentially fixedly adjusted to (70, 198), (429, 198), (70, 302), (429, 302);
(c) the license plate type is a single-layer license plate containing 7 characters, such as a single blue license plate, a single yellow license plate and a police car license plate: the coordinates are sequentially fixedly adjusted to (80, 196), (419, 196), (80, 303), (419, 303);
(d) the license plate type is double-layer license plates such as yellow license plates and hanging license plates: the coordinates are fixed and adjusted to (80, 165), (419, 165), (80, 334), (419, 334) in this order.
The traditional correction methods such as Hough transform, Randon transform and the like are difficult to cope with various complex scenes such as yin-yang license plates, worn license plates and the like, and the angular point positioning method based on deep learning target detection has better robustness in the complex scenes.
The traditional partial correction algorithm does not consider the problems of single-layer and double-layer and character number, namely how to calculate the corrected coordinate values more accurately, double-layer license plates and new energy license plates are common in actual scenes, the new coordinate values can influence the length-width ratio of the license plates and the character recognition rate in the later period, and a license plate type judgment network is designed on the basis of the new coordinate values.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may be made by those skilled in the art without departing from the principle of the invention.
Claims (10)
1. A deep learning-based inclined license plate correction method is characterized by comprising the following steps:
step 1, obtaining a local image of a license plate, and constructing a visualization tool for marking a corner point of the license plate;
step 2, constructing a target detection network model based on deep learning, and training the model;
step 3, constructing a full convolution license plate type classification network and training the model;
step 4, detecting the license plate, acquiring a local image of the license plate, and detecting license plate corner points in the local image by using the model obtained in the step 2 for positioning;
step 5, judging the license plate type by using the license plate type classification network obtained in the step 3;
and 6, performing corresponding perspective transformation according to the license plate type, adjusting the detected corner coordinates to a fixed value, and adjusting the length-width ratio of the corner coordinates to be close to the standard condition.
2. The method for correcting the tilted license plate according to claim 1, wherein a conventional license plate detector is built in the visualization tool for marking the corner points of the license plate in the step 1, and the marking frame of each corner point is a square; and automatically calculating the size of the labeling frame according to the coordinates of the local image of the license plate.
3. The method of claim 2, wherein the size of the local image of the license plate is fixed to 500 x 500, the size of the labeling frame is two thirds of the height of the license plate, and each training image is labeled with 4 corner points.
4. The method of claim 1, wherein the target detection network in step 2 is a fully-convoluted neural network, and the corner points are detected on two different size feature maps.
5. The method for correcting the tilted license plate of claim 1, wherein the target detection network model based on deep learning constructed in the step 2 is a corner point detection model based on YOLOv 3.
6. The method according to claim 5, wherein the target detection network based on deep learning has 20 layers; the input of the network is a 3-channel image of 224 × 224, and the 1 st to 11 th layers are convolution layers with a convolution kernel size of 3 × 3; the 13 th layer is a yolo layer, and targets are detected on a 7 x 7 size feature map; the 14 th layer is a route layer and is used for acquiring a characteristic diagram of the 11 th layer; the 15 th layer is a convolution layer with convolution kernel size of 1 multiplied by 1; the 16 th layer is an upsampling layer, and the size of the feature map is changed into 14 multiplied by 14; the 17 th layer is a route layer and is used for splicing the characteristic diagrams of the 7 th layer and the 16 th layer; layer 18 is a convolutional layer with a convolutional kernel size of 3 × 3; the 20 th layer is the yolo layer, and targets are detected on a feature map of 14 × 14 size.
7. The method for correcting an inclined license plate according to claim 1, wherein the license plate type classification network trained in step 3 is used for classifying license plate types into three types:
a) the single-layer new energy license plate comprises 8 characters;
b) the single-layer license plate containing 7 characters comprises a blue license plate, a yellow license plate and a police car license plate which are common on a car head;
c) the double-layer character license plate comprises a yellow license plate and a trailer license plate.
8. The method of claim 7, wherein the license plate type classification network comprises a full convolution network of 8 convolution layers, a global averaging pooling layer, a full connection layer, and a probability output layer; the network input size is 96 × 96 × 3, each convolutional layer uses a convolutional kernel of size 3 × 3, and the activation function is Relu.
9. The method of claim 1, wherein the license plate class is classified according to the detected corner coordinates to perform perspective transformation in step 6, and the corner coordinates are transformed into fixed coordinate points and the aspect ratio is adjusted to be close to a standard condition.
10. The method for correcting the tilted license plate of claim 1, wherein the perspective transformation process in step 6 comprises the following steps:
calculating 3X 3 according to 4 pairs of license plate corner coordinatesTransformation matrixM:
the calculation mode of the corrected license plate local image is as follows:
wherein ,srcthe local image of the license plate before correction is obtained;dstthe corrected local image of the license plate is obtained;M 11…M 33for transforming matricesMElements of the corresponding position.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110767061.0A CN113435446B (en) | 2021-07-07 | 2021-07-07 | Deep learning-based inclined license plate correction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110767061.0A CN113435446B (en) | 2021-07-07 | 2021-07-07 | Deep learning-based inclined license plate correction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113435446A true CN113435446A (en) | 2021-09-24 |
CN113435446B CN113435446B (en) | 2023-10-31 |
Family
ID=77759343
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110767061.0A Active CN113435446B (en) | 2021-07-07 | 2021-07-07 | Deep learning-based inclined license plate correction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113435446B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114882492A (en) * | 2022-07-11 | 2022-08-09 | 浙江大华技术股份有限公司 | License plate recognition method, device, terminal and computer readable storage medium |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107563385A (en) * | 2017-09-02 | 2018-01-09 | 西安电子科技大学 | License plate character recognition method based on depth convolution production confrontation network |
CN110020651A (en) * | 2019-04-19 | 2019-07-16 | 福州大学 | Car plate detection localization method based on deep learning network |
CN110309828A (en) * | 2019-06-27 | 2019-10-08 | 浙江工业大学 | A kind of inclination license plate antidote |
CN111191611A (en) * | 2019-12-31 | 2020-05-22 | 同济大学 | Deep learning-based traffic sign label identification method |
CN111461113A (en) * | 2020-03-03 | 2020-07-28 | 华南理工大学 | Large-angle license plate detection method based on deformed plane object detection network |
CN111598089A (en) * | 2020-05-16 | 2020-08-28 | 湖南大学 | License plate correction and recognition method based on deep learning |
CN111666938A (en) * | 2020-05-21 | 2020-09-15 | 珠海大横琴科技发展有限公司 | Two-place double-license-plate detection and identification method and system based on deep learning |
CN111860509A (en) * | 2020-07-28 | 2020-10-30 | 湖北九感科技有限公司 | Coarse-to-fine two-stage non-constrained license plate region accurate extraction method |
CN112132130A (en) * | 2020-09-22 | 2020-12-25 | 福州大学 | Real-time license plate detection method and system for whole scene |
KR20200145174A (en) * | 2019-06-21 | 2020-12-30 | 인하대학교 산학협력단 | System and method for recognizing license plates |
CN112232371A (en) * | 2020-09-17 | 2021-01-15 | 福州大学 | American license plate recognition method based on YOLOv3 and text recognition |
CN112528994A (en) * | 2020-12-18 | 2021-03-19 | 南京师范大学 | Free-angle license plate detection method, license plate identification method and identification system |
CN112668663A (en) * | 2021-01-05 | 2021-04-16 | 南京航空航天大学 | Aerial photography car detection method based on YOLOv4 |
-
2021
- 2021-07-07 CN CN202110767061.0A patent/CN113435446B/en active Active
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107563385A (en) * | 2017-09-02 | 2018-01-09 | 西安电子科技大学 | License plate character recognition method based on depth convolution production confrontation network |
CN110020651A (en) * | 2019-04-19 | 2019-07-16 | 福州大学 | Car plate detection localization method based on deep learning network |
KR20200145174A (en) * | 2019-06-21 | 2020-12-30 | 인하대학교 산학협력단 | System and method for recognizing license plates |
CN110309828A (en) * | 2019-06-27 | 2019-10-08 | 浙江工业大学 | A kind of inclination license plate antidote |
CN111191611A (en) * | 2019-12-31 | 2020-05-22 | 同济大学 | Deep learning-based traffic sign label identification method |
CN111461113A (en) * | 2020-03-03 | 2020-07-28 | 华南理工大学 | Large-angle license plate detection method based on deformed plane object detection network |
CN111598089A (en) * | 2020-05-16 | 2020-08-28 | 湖南大学 | License plate correction and recognition method based on deep learning |
CN111666938A (en) * | 2020-05-21 | 2020-09-15 | 珠海大横琴科技发展有限公司 | Two-place double-license-plate detection and identification method and system based on deep learning |
CN111860509A (en) * | 2020-07-28 | 2020-10-30 | 湖北九感科技有限公司 | Coarse-to-fine two-stage non-constrained license plate region accurate extraction method |
CN112232371A (en) * | 2020-09-17 | 2021-01-15 | 福州大学 | American license plate recognition method based on YOLOv3 and text recognition |
CN112132130A (en) * | 2020-09-22 | 2020-12-25 | 福州大学 | Real-time license plate detection method and system for whole scene |
CN112528994A (en) * | 2020-12-18 | 2021-03-19 | 南京师范大学 | Free-angle license plate detection method, license plate identification method and identification system |
CN112668663A (en) * | 2021-01-05 | 2021-04-16 | 南京航空航天大学 | Aerial photography car detection method based on YOLOv4 |
Non-Patent Citations (2)
Title |
---|
IZIDIO, DMF等: "An embedded automatic license plate recognition system using deep learning", DESIGN AUTOMATION FOR EMBEDDED SYSTEMS, vol. 24, no. 1, pages 23 - 43, XP037085784, DOI: 10.1007/s10617-019-09230-5 * |
贺智龙;肖中俊;严志国;: "轻量级车牌识别卷积网络", 齐鲁工业大学学报, no. 04, pages 39 - 45 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114882492A (en) * | 2022-07-11 | 2022-08-09 | 浙江大华技术股份有限公司 | License plate recognition method, device, terminal and computer readable storage medium |
CN114882492B (en) * | 2022-07-11 | 2022-11-22 | 浙江大华技术股份有限公司 | License plate recognition method, device, terminal and computer readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN113435446B (en) | 2023-10-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109190444B (en) | Method for realizing video-based toll lane vehicle feature recognition system | |
CN111666938B (en) | Two-place double-license-plate detection and identification method and system based on deep learning | |
CN107729801A (en) | A kind of vehicle color identifying system based on multitask depth convolutional neural networks | |
CN111553201B (en) | Traffic light detection method based on YOLOv3 optimization algorithm | |
CN109344825A (en) | A kind of licence plate recognition method based on convolutional neural networks | |
CN108961235A (en) | A kind of disordered insulator recognition methods based on YOLOv3 network and particle filter algorithm | |
CN103793708B (en) | A kind of multiple dimensioned car plate precise positioning method based on motion correction | |
CN105354568A (en) | Convolutional neural network based vehicle logo identification method | |
CN106251353A (en) | Weak texture workpiece and the recognition detection method and system of three-dimensional pose thereof | |
CN106469304A (en) | Handwritten signature location positioning method in bill based on depth convolutional neural networks | |
CN108388871B (en) | Vehicle detection method based on vehicle body regression | |
CN110796700B (en) | Multi-object grabbing area positioning method based on convolutional neural network | |
CN112949633B (en) | Improved YOLOv 3-based infrared target detection method | |
CN108334881A (en) | A kind of licence plate recognition method based on deep learning | |
WO2020093624A1 (en) | Antenna downward inclination angle measurement method based on multi-scale detection algorithm | |
CN113095316B (en) | Image rotation target detection method based on multilevel fusion and angular point offset | |
CN111860509A (en) | Coarse-to-fine two-stage non-constrained license plate region accurate extraction method | |
CN111242002A (en) | Shared bicycle standardized parking judgment method based on computer vision | |
CN110659637A (en) | Electric energy meter number and label automatic identification method combining deep neural network and SIFT features | |
CN113435446B (en) | Deep learning-based inclined license plate correction method | |
CN110909772B (en) | High-precision real-time multi-scale dial pointer detection method and system | |
CN110733416B (en) | Lane departure early warning method based on inverse perspective transformation | |
CN112529001A (en) | License plate recognition method based on neural network | |
CN110309828B (en) | Inclined license plate correction method | |
CN114241438B (en) | Traffic signal lamp rapid and accurate identification method based on priori information |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |