CN114155523A - End-to-end efficient and accurate license plate detection and identification method - Google Patents
End-to-end efficient and accurate license plate detection and identification method Download PDFInfo
- Publication number
- CN114155523A CN114155523A CN202111521371.0A CN202111521371A CN114155523A CN 114155523 A CN114155523 A CN 114155523A CN 202111521371 A CN202111521371 A CN 202111521371A CN 114155523 A CN114155523 A CN 114155523A
- Authority
- CN
- China
- Prior art keywords
- license plate
- character
- plate frame
- characteristic diagram
- convolution
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000010586 diagram Methods 0.000 claims abstract description 54
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 14
- 238000011176 pooling Methods 0.000 claims abstract description 9
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 7
- 230000005540 biological transmission Effects 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 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
- G06F18/24—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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Character Discrimination (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an end-to-end efficient and accurate license plate detection and identification method, which comprises the steps of collecting a license plate image, extracting shallow layer characteristics and license plate frame characteristics in the license plate image by a license plate detection module through a convolutional neural network to form a license plate frame characteristic diagram; returning the license plate frame feature map to the license plate boundary coordinates through the full connecting layer; the license plate recognition module shares the shallow layer characteristics extracted by the license plate detection module to obtain the character characteristics of the license plate image and form a character combined characteristic graph group; the character combination feature map group and the license plate frame feature map are spliced after region-of-interest pooling operation to obtain key region features of the characters; and after the character key region features are identified by the classifier, the license plate number after identification is obtained. The information input in the end-to-end license plate detection and identification method comprises the image of the license plate, the position area of the license plate can be positioned in one forward transmission, the number of the license plate can be identified, the whole model is light, flexible and efficient, and the license plate detection and identification can be completed quickly and accurately.
Description
Technical Field
The invention relates to the field of license plate detection and identification, in particular to an end-to-end efficient and accurate license plate detection and identification method.
Background
The existing license plate detection and identification method is divided into a multi-stage license plate detection and identification method and a single-stage license plate detection and identification method. At present, most license plate detection and identification methods adopt a multi-stage method, and the main defect is that two tasks of detection and identification are separated, so that the accuracy and the time efficiency of the whole method are greatly influenced. The single-stage method is based on a neural network, license plate detection and recognition are completed through a unified network model, the method uses fewer parameters to link detection and recognition tasks, and has better precision and efficiency, but the general difficulty is higher.
The multi-stage method comprises a license plate detection algorithm used for positioning a license plate region in an image, a license plate segmentation algorithm used for segmenting license plate characters, and a license plate recognition algorithm used for carrying out character recognition on the segmented characters. The method is mainly divided into a traditional method and a deep learning method.
The traditional license plate detection and recognition algorithm cannot extract enough abundant feature information due to the limitation of a feature extraction mode, so that the accuracy is limited and the generalization performance is poor. The existing license plate detection and recognition algorithm based on deep learning has certain advantages in accuracy, but the complex structure can reduce the efficiency of the network.
There has been a small amount of research work in attempting to use single-stage license plate detection and recognition. Compared with a multi-stage algorithm, the single-stage algorithm can better utilize the characteristics of high correlation between license plate detection and license plate recognition, and simultaneously reduces the parameters of the model, so that the algorithm is faster and more effective. However, the existing end-to-end license plate detection and recognition algorithm based on deep learning is still limited by the problems of long time, low efficiency and the like caused by network depth.
Disclosure of Invention
The invention provides an end-to-end efficient and accurate license plate detection and identification method, which aims to overcome the technical problems of low license plate detection and identification efficiency and the like in the prior art.
In order to achieve the purpose, the technical scheme of the invention is as follows:
an end-to-end efficient and accurate license plate detection and identification method is characterized by comprising the following steps:
step 1, collecting a license plate image and inputting the license plate image into a license plate detection module;
step 2, extracting shallow layer characteristics and license plate frame characteristics in the license plate image through a convolutional neural network by a license plate detection module to form a license plate frame characteristic diagram;
step 3, returning license plate boundary coordinates from the license plate frame characteristic graph through the full connecting layer;
step 4, the license plate recognition module shares the shallow feature extracted by the license plate detection module, and obtains the character feature of the license plate image through convolution operation to form two groups of character combined feature maps;
step 5, splicing the two groups of character combined feature maps and the license plate frame feature map after region-of-interest pooling operation to obtain key region features of the characters;
and 6, identifying the character key region characteristics through a classifier to obtain the number of the license plate after identification.
Furthermore, the license plate detection module has 7 convolution layers, the first convolution layer is input as a license plate image and convolutes to output a first license plate frame characteristic diagram, and transmitted to a second convolution layer, the second convolution layer convolves the first license plate frame characteristic diagram and outputs a second license plate frame characteristic, and transmitted to a third convolution layer, which convolves the second license plate frame characteristic diagram and outputs a third license plate frame characteristic, and transmitted to a fourth convolution layer, which convolves the third license plate frame characteristic diagram and outputs a fourth license plate frame characteristic, and transmitted to a fifth convolution layer, which convolves the fourth license plate frame characteristic diagram and outputs a fifth license plate frame characteristic, and transmitted to a sixth convolution layer, which convolves the fifth license plate frame feature map and outputs a sixth license plate frame feature, and transmitting the third license plate frame characteristic graph to a third convolution layer, and convolving the third license plate frame characteristic graph by the third convolution layer and outputting a third license plate frame characteristic graph.
Further, the step 2 further comprises: and a seventh license plate frame characteristic output by a seventh convolutional layer in the convolutional neural network is an attention parameter of the character characteristic and is used for assisting the license plate recognition module to position the character characteristic.
Further, the license plate recognition module comprises a first recognition convolutional layer and a second recognition convolutional layer, the license plate recognition module shares the shallow layer features extracted by the license plate detection module, and the first recognition convolutional layer and the second recognition convolutional layer are used for extracting character features of a license plate image to form a first character feature map and a second character feature map.
Further, a first license plate frame feature map in a first convolution layer of a convolution neural network in the license plate detection module is selected as input information of a first recognition convolution layer in the license plate recognition module, and a first character feature map is formed through convolution; selecting a third license plate frame characteristic diagram in a third convolution layer of a convolution neural network in the license plate detection module as input information of a second recognition convolution layer in the license plate recognition module, and performing convolution to form a second character characteristic diagram; splicing the second license plate frame characteristic diagram and the first character characteristic diagram to form a first character combination characteristic diagram; splicing the fourth license plate frame characteristic diagram and the second character characteristic diagram to form a second character combination characteristic diagram; and the first combined feature map is restored through convolution dimension reduction operation to obtain a first character combined feature map, and the second combined feature map is restored through convolution dimension reduction operation to obtain a second character combined feature map.
Further, the step 5 specifically includes:
step 5.1, obtaining three key character region characteristics with the same size through region-of-interest pooling operation of the first character combined characteristic diagram, the second character combined characteristic diagram and the license plate frame characteristic diagram at the sixth layer respectively;
and 5.2, splicing the three character key area characteristics with the same size to obtain the character key area characteristics.
Further, the step 6 specifically includes:
Lossdec(pb,gb)=∑N∑m∈{cx,cy,w,h}smoothL1(pbm-gbm) (2)
wherein LOSS (pb, pn, gb, gn) is an integral LOSS model, namely, a recognized license plate number, and comprises a license plate detection module LOSS model Lossdec(pb, gb) and license plate recognition module Loss model Lossreg(pn, gn), wherein N is the Size of the convolutional neural network Batch Size, m is the coordinate of the license plate frame, cx is the coordinate of the center point x of the license plate frame, cy is the coordinate of the center point y of the license plate frame, w is the width of the license plate frame, h is the height of the license plate frame, and smoothL1And in the loss function, pb is the license plate boundary coordinate in the character key region characteristic, gb is the real license plate boundary coordinate, pn is the recognized license plate number, gn is the real license plate number, M is the length of the license plate number, i represents the ith license plate character, j is the jth class of a license plate character in the convolutional neural network, and each license plate character has c classes.
Has the advantages that:
1. the information input in the end-to-end efficient and accurate license plate detection and identification method comprises an image of a license plate, and the position area of the license plate can be positioned in one forward transmission and the license plate number in the position area can be identified;
2. the license plate detection and recognition system can efficiently detect and recognize the license plate, integrates two subtasks of detection and recognition, and avoids dividing the detection and recognition into two independent stages, thereby saving the execution time.
3. The characteristics of different levels extracted by the convolutional neural network are applied, and the characteristics of the network sharing shallow layer are simultaneously used for detecting and identifying tasks, so that repeated extraction of the characteristics is avoided, and the time efficiency is improved.
4. The relevance between the license plate detection and the license plate recognition is emphasized, and the license plate character recognition is more accurate by combining the license plate boundary coordinates and the character key region characteristic model. Meanwhile, after the license plate is identified, two tasks of license plate detection and identification are optimized in a combined mode, and the detection performance and the identification performance are further improved.
5. The whole model is light, flexible and efficient, and can quickly and accurately complete the license plate detection and identification.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flow chart of an end-to-end efficient and accurate license plate detection and identification method of the present invention;
FIG. 2 is a diagram of an end-to-end efficient and accurate license plate detection network architecture according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment provides an end-to-end efficient and accurate license plate detection and identification method, as shown in fig. 1-2, comprising the following steps:
step 1, collecting a license plate image and inputting the license plate image into a license plate detection module;
step 2, extracting shallow features and license plate frame features in the license plate image through a convolutional neural network by a license plate detection module to form a license plate frame feature map, wherein the shallow features are basic features of the image, such as color, texture, edges and the like;
step 3, returning license plate boundary coordinates from the license plate frame characteristic graph through the full connecting layer;
step 4, the license plate recognition module shares the shallow feature extracted by the license plate detection module, and obtains the character feature of the license plate image through convolution operation to form two groups of character combined feature maps;
step 5, splicing the two groups of character combined feature maps and the license plate frame feature map after region-of-interest pooling operation to obtain key region features of the characters;
and 6, identifying the character key region characteristics through M classifiers to obtain the license plate number with the length of M after identification.
Specifically, seven convolution layers in the license plate detection module are used for extracting the frame characteristics of the license plate, and a regression module consisting of three full-connection layers is used for regressing a license plate boundary coordinate model; of the seven convolutional layers, each convolutional layer is composed of five parts: convolution operation, Batch Normalization (BN), ReLU activation, Max Pool (Max Pool), and Dropout; the feature map output by the last convolutional layer is input into the next three fully-connected layers for regression of the coordinates (i.e., the frame) of the license plate region.
The license plate detection module has 7 convolution layers, the first convolution layer inputs a license plate image and outputs a first license plate frame characteristic diagram (namely an output result of the first convolution layer) in a convolution mode, the first convolution layer is transmitted to the second convolution layer, the second convolution layer convolves the first license plate frame characteristic diagram and outputs a second license plate frame characteristic (namely an output result of the second convolution layer), the second convolution layer is transmitted to the third convolution layer, the third convolution layer convolves the second license plate frame characteristic diagram and outputs a third license plate frame characteristic (namely an output result of the third convolution layer), the fourth convolution layer convolves the third license plate frame characteristic diagram and outputs a fourth license plate frame characteristic (namely an output result of the fourth convolution layer), the fourth convolution layer convolves the fourth license plate frame characteristic diagram and outputs a fifth license plate frame characteristic (namely an output result of the fifth convolution layer), and transmitting to a sixth convolutional layer, which convolves the fifth license plate frame feature map and outputs a sixth license plate frame feature (i.e., an output result of the sixth convolutional layer), and transmitting to a seventh convolutional layer, which convolves the sixth license plate frame feature map and outputs a seventh license plate frame feature (i.e., an output result of the seventh convolutional layer).
In a specific embodiment, the step 2 further includes:
and the seventh license plate frame characteristic output by the seventh convolutional layer in the convolutional neural network is an attention parameter (attention) of the character characteristic and is used for assisting the license plate recognition module to position the character characteristic.
In a specific embodiment, the license plate recognition module comprises a first recognition convolutional layer and a second recognition convolutional layer, the license plate recognition module shares the shallow layer features extracted by the license plate detection module, and the first recognition convolutional layer and the second recognition convolutional layer are used for extracting character features of a license plate image to form a first character feature map and a second character feature map.
In a specific embodiment, a first license plate frame feature map in a first convolution layer of a convolution neural network in a license plate detection module is selected as input information of the first recognition convolution layer in a license plate recognition module, and a first character feature map is formed through convolution; selecting a third license plate frame characteristic diagram in a third convolution layer of a convolution neural network in the license plate detection module as input information of a second recognition convolution layer in the license plate recognition module, and performing convolution to form a second character characteristic diagram; splicing the second license plate frame characteristic diagram and the first character characteristic diagram to form a first character combination characteristic diagram; splicing the fourth license plate frame characteristic diagram and the second character characteristic diagram to form a second character combination characteristic diagram; the first combined feature map is reduced through convolution dimension reduction operation to obtain a first character combined feature map, and the second combined feature map is reduced through convolution dimension reduction operation to obtain a second character combined feature map
In a specific embodiment, the step 5 specifically includes:
step 5.1, obtaining three key character region characteristics with the same size through region-of-interest pooling operation of the first character combined characteristic diagram, the second character combined characteristic diagram and the license plate frame characteristic diagram at the sixth layer respectively;
and 5.2, splicing the three character key area characteristics with the same size to obtain the character key area characteristics.
In a specific embodiment, the step 6 specifically includes:
Lossdec(pb,gb)=∑N∑m∈{cx,cy,w,h}smoothL1(pbm-gbm) (2)
wherein LOSS (pb, pn, gb, gn) is an integral LOSS model, namely, a recognized license plate number, and the LOSS comprises a license plate detection module LOSS model Lossdec(pb, gb) and license plate recognition module Loss model Lossreg(pn, gn), wherein N is the Size of the convolutional neural network Batch Size, m is the coordinate of the license plate frame, cx is the coordinate of the center point x of the license plate frame, cy is the coordinate of the center point y of the license plate frame, w is the width of the license plate frame, h is the height of the license plate frame, and smoothL1And in the loss function, pb is the license plate boundary coordinate in the character key region characteristic, gb is the real license plate boundary coordinate, pn is the recognized license plate number, gn is the real license plate number, M is the length of the license plate number, i represents the ith license plate character, j is the jth class of a license plate character in the convolutional neural network, and each license plate character has c classes.
In a specific embodiment, the license plate recognition module selects a first character combined feature map, a second character combined feature map and a license plate frame feature map at a sixth layer to perform ROI pooling respectively, then combines the pooled feature maps, sends the combined feature maps into seven parallel classifiers, namely classifiers 1-7 in FIG. 2, and finally outputs a recognized license plate model, namely a predicted license plate number; the extraction and selection of the three feature maps input into the ROI pooling becomes the attention of the license plate recognition module.
In FIG. 2, Conv-d1, Conv-d2, Conv-d3, Conv-d4, Conv-d5, Conv-d6 and Conv-d7 are 7 convolutional layers of a license plate detection module respectively, Conv-r1 and Conv-r2 are a first recognition convolutional layer and a second recognition convolutional layer of a license plate recognition module respectively, and Conv-r3 and Conv-r4 are convolution dimension reduction operations and output a first character combination feature map and a second character combination feature map respectively.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (7)
1. An end-to-end efficient and accurate license plate detection and identification method is characterized by comprising the following steps:
step 1, collecting a license plate image and inputting the license plate image into a license plate detection module;
step 2, extracting shallow layer characteristics and license plate frame characteristics in the license plate image through a convolutional neural network by a license plate detection module to form a license plate frame characteristic diagram;
step 3, returning license plate boundary coordinates from the license plate frame characteristic graph through the full connecting layer;
step 4, the license plate recognition module shares the shallow feature extracted by the license plate detection module, and obtains the character feature of the license plate image through convolution operation to form two groups of character combined feature maps;
step 5, splicing the two groups of character combined feature maps and the license plate frame feature map after region-of-interest pooling operation to obtain key region features of the characters;
and 6, identifying the character key region characteristics through a classifier to obtain the number of the license plate after identification.
2. The method of claim 1, wherein the method comprises the steps of: the license plate detection module has 7 convolution layers, the first convolution layer is input as a license plate image and convolutes to output a first license plate frame characteristic diagram, and transmitted to a second convolution layer, the second convolution layer convolves the first license plate frame characteristic diagram and outputs a second license plate frame characteristic, and transmitted to a third convolution layer, which convolves the second license plate frame characteristic diagram and outputs a third license plate frame characteristic, and transmitted to a fourth convolution layer, which convolves the third license plate frame characteristic diagram and outputs a fourth license plate frame characteristic, and transmitted to a fifth convolution layer, which convolves the fourth license plate frame characteristic diagram and outputs a fifth license plate frame characteristic, and transmitted to a sixth convolution layer, which convolves the fifth license plate frame feature map and outputs a sixth license plate frame feature, and transmitting the third license plate frame characteristic graph to a third convolution layer, and convolving the third license plate frame characteristic graph by the third convolution layer and outputting a third license plate frame characteristic graph.
3. The method for detecting and identifying a license plate of a vehicle with high efficiency and accuracy from end to end according to claim 2, wherein the step 2 further comprises: and a seventh license plate frame characteristic output by a seventh convolutional layer in the convolutional neural network is an attention parameter of the character characteristic and is used for assisting the license plate recognition module to position the character characteristic.
4. The method of claim 3, wherein the method comprises the steps of: the license plate recognition module comprises a first recognition convolutional layer and a second recognition convolutional layer, the license plate recognition module shares the shallow layer characteristics extracted by the license plate detection module, and the first recognition convolutional layer and the second recognition convolutional layer are used for extracting the character characteristics of the license plate image to form a first character characteristic diagram and a second character characteristic diagram.
5. The method of claim 4, wherein the method comprises the steps of: a first license plate frame feature map in a first convolution layer of a convolution neural network in a license plate detection module is selected as input information of a first recognition convolution layer in a license plate recognition module, and a first character feature map is formed through convolution; selecting a third license plate frame characteristic diagram in a third convolution layer of a convolution neural network in the license plate detection module as input information of a second recognition convolution layer in the license plate recognition module, and performing convolution to form a second character characteristic diagram; splicing the second license plate frame characteristic diagram and the first character characteristic diagram to form a first character combination characteristic diagram; splicing the fourth license plate frame characteristic diagram and the second character characteristic diagram to form a second character combination characteristic diagram; and the first combined feature map is restored through convolution dimension reduction operation to obtain a first character combined feature map, and the second combined feature map is restored through convolution dimension reduction operation to obtain a second character combined feature map.
6. The method for detecting and identifying the license plate of claim 5, wherein the step 5 is specifically as follows:
step 5.1, obtaining three key character region characteristics with the same size through region-of-interest pooling operation of the first character combined characteristic diagram, the second character combined characteristic diagram and the license plate frame characteristic diagram at the sixth layer respectively;
and 5.2, splicing the three character key area characteristics with the same size to obtain the character key area characteristics.
7. The method for detecting and identifying the license plate of claim 6, wherein the step 6 is specifically as follows:
Lossdec(pb,gb)=∑N∑m∈{cx,cy,w,h}smoothL1(pbm-gbm) (2)
wherein LOSS (pb, pn, gb, gn) is an integral LOSS model, namely, a recognized license plate number, and the LOSS comprises a license plate detection module LOSS model Lossdec(pb, gb) and license plate recognition module Loss model Lossreg(pn, gn), wherein N is the Size of the convolutional neural network Batch Size, m is the coordinate of the license plate frame, cx is the coordinate of the center point x of the license plate frame, cy is the coordinate of the center point y of the license plate frame, w is the width of the license plate frame, h is the height of the license plate frame, and smoothL1And in the loss function, pb is the license plate boundary coordinate in the character key region characteristic, gb is the real license plate boundary coordinate, pn is the recognized license plate number, gn is the real license plate number, M is the length of the license plate number, i represents the ith license plate character, j is the jth class of a license plate character in the convolutional neural network, and each license plate character has c classes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111521371.0A CN114155523A (en) | 2021-12-13 | 2021-12-13 | End-to-end efficient and accurate license plate detection and identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111521371.0A CN114155523A (en) | 2021-12-13 | 2021-12-13 | End-to-end efficient and accurate license plate detection and identification method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114155523A true CN114155523A (en) | 2022-03-08 |
Family
ID=80450686
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111521371.0A Pending CN114155523A (en) | 2021-12-13 | 2021-12-13 | End-to-end efficient and accurate license plate detection and identification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114155523A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115376119A (en) * | 2022-10-25 | 2022-11-22 | 珠海亿智电子科技有限公司 | License plate recognition method and device, license plate recognition equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018233038A1 (en) * | 2017-06-23 | 2018-12-27 | 平安科技(深圳)有限公司 | Deep learning-based method, apparatus and device for recognizing license plate, and storage medium |
CN109344825A (en) * | 2018-09-14 | 2019-02-15 | 广州麦仑信息科技有限公司 | A kind of licence plate recognition method based on convolutional neural networks |
CN110796643A (en) * | 2019-10-18 | 2020-02-14 | 四川大学 | Rail fastener defect detection method and system |
CN111444840A (en) * | 2020-03-26 | 2020-07-24 | 中科海微(北京)科技有限公司 | Automatic detection method and system for fake-licensed vehicle |
CN112149661A (en) * | 2020-08-07 | 2020-12-29 | 珠海欧比特宇航科技股份有限公司 | License plate recognition method, device and medium |
-
2021
- 2021-12-13 CN CN202111521371.0A patent/CN114155523A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018233038A1 (en) * | 2017-06-23 | 2018-12-27 | 平安科技(深圳)有限公司 | Deep learning-based method, apparatus and device for recognizing license plate, and storage medium |
CN109344825A (en) * | 2018-09-14 | 2019-02-15 | 广州麦仑信息科技有限公司 | A kind of licence plate recognition method based on convolutional neural networks |
CN110796643A (en) * | 2019-10-18 | 2020-02-14 | 四川大学 | Rail fastener defect detection method and system |
CN111444840A (en) * | 2020-03-26 | 2020-07-24 | 中科海微(北京)科技有限公司 | Automatic detection method and system for fake-licensed vehicle |
CN112149661A (en) * | 2020-08-07 | 2020-12-29 | 珠海欧比特宇航科技股份有限公司 | License plate recognition method, device and medium |
Non-Patent Citations (1)
Title |
---|
史建伟;章韵;: "基于改进YOLOv3和BGRU的车牌识别系统", 计算机工程与设计, no. 08, 16 August 2020 (2020-08-16) * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115376119A (en) * | 2022-10-25 | 2022-11-22 | 珠海亿智电子科技有限公司 | License plate recognition method and device, license plate recognition equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111210443B (en) | Deformable convolution mixing task cascading semantic segmentation method based on embedding balance | |
CN109902806B (en) | Method for determining target bounding box of noise image based on convolutional neural network | |
CN111639564B (en) | Video pedestrian re-identification method based on multi-attention heterogeneous network | |
CN109766805B (en) | Deep learning-based double-layer license plate character recognition method | |
CN111104903A (en) | Depth perception traffic scene multi-target detection method and system | |
CN111462140B (en) | Real-time image instance segmentation method based on block stitching | |
CN111753682B (en) | Hoisting area dynamic monitoring method based on target detection algorithm | |
CN111626120B (en) | Target detection method based on improved YOLO-6D algorithm in industrial environment | |
CN107564007B (en) | Scene segmentation correction method and system fusing global information | |
CN111476248A (en) | CNN method and apparatus using 1x1 convolution for image recognition | |
CN111401145A (en) | Visible light iris recognition method based on deep learning and DS evidence theory | |
CN113850136A (en) | Yolov5 and BCNN-based vehicle orientation identification method and system | |
CN115471670A (en) | Space target detection method based on improved YOLOX network model | |
CN112669343A (en) | Zhuang minority nationality clothing segmentation method based on deep learning | |
CN114693966A (en) | Target detection method based on deep learning | |
CN113487610B (en) | Herpes image recognition method and device, computer equipment and storage medium | |
CN114155523A (en) | End-to-end efficient and accurate license plate detection and identification method | |
CN111199255A (en) | Small target detection network model and detection method based on dark net53 network | |
CN113537119B (en) | Transmission line connecting part detection method based on improved Yolov4-tiny | |
CN113763412A (en) | Image processing method and device, electronic equipment and computer readable storage medium | |
CN112132145A (en) | Image classification method and system based on model extended convolutional neural network | |
CN115953743A (en) | Parking space state identification method based on improved YOLO model | |
CN113223006B (en) | Lightweight target semantic segmentation method based on deep learning | |
CN115359487A (en) | Rapid railcar number identification method, equipment and storage medium | |
CN115424012A (en) | Lightweight image semantic segmentation method based on context 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 |