CN111652012B - A Surface QR Code Localization Method Based on SSD Network Model - Google Patents
A Surface QR Code Localization Method Based on SSD Network Model Download PDFInfo
- Publication number
- CN111652012B CN111652012B CN202010393000.8A CN202010393000A CN111652012B CN 111652012 B CN111652012 B CN 111652012B CN 202010393000 A CN202010393000 A CN 202010393000A CN 111652012 B CN111652012 B CN 111652012B
- Authority
- CN
- China
- Prior art keywords
- code
- curved surface
- layer
- network
- ssd
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 230000004807 localization Effects 0.000 title description 3
- 238000012549 training Methods 0.000 claims abstract description 44
- 238000001514 detection method Methods 0.000 claims abstract description 19
- 238000000605 extraction Methods 0.000 claims abstract description 12
- 238000011176 pooling Methods 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 9
- 238000013434 data augmentation Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 206010037180 Psychiatric symptoms Diseases 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 2
- 230000007613 environmental effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 6
- 238000005286 illumination Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011897 real-time detection Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
- G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
- G06K7/14—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation using light without selection of wavelength, e.g. sensing reflected white light
- G06K7/1404—Methods for optical code recognition
- G06K7/1439—Methods for optical code recognition including a method step for retrieval of the optical code
- G06K7/1443—Methods for optical code recognition including a method step for retrieval of the optical code locating of the code in an image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
- G06K7/00—Methods or arrangements for sensing record carriers, e.g. for reading patterns
- G06K7/10—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation
- G06K7/14—Methods or arrangements for sensing record carriers, e.g. for reading patterns by electromagnetic radiation, e.g. optical sensing; by corpuscular radiation using light without selection of wavelength, e.g. sensing reflected white light
- G06K7/1404—Methods for optical code recognition
- G06K7/1408—Methods for optical code recognition the method being specifically adapted for the type of code
- G06K7/1417—2D bar codes
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Electromagnetism (AREA)
- Toxicology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
本发明公开了一种基于SSD网络模型的曲面QR码定位方法，首先建立曲面QR码数据集并对其进行预处理；搭建PPN‑SSD框架，PPN‑SSD框架以Inception网络为基础网络，添加辅助网络层作为网络的特征提取层和分类层；利用PPN‑SSD框架进行模型训练得到曲面QR码定位模型；利用训练好的曲面QR码定位模型对曲面QR码进行定位。本发明提供的复杂环境下曲面QR码定位方法，解决了目前算法难以定位曲面QR码的问题，并将复杂环境因素影响考虑其中，定位准确率高，检测速度快，实现了曲面QR码的实时检测。
The invention discloses a surface QR code positioning method based on an SSD network model. First, a surface QR code data set is established and preprocessed; a PPN-SSD framework is constructed. The PPN-SSD framework takes the Inception network as the basic network and adds auxiliary The network layer is used as the feature extraction layer and classification layer of the network; the PPN-SSD framework is used for model training to obtain the surface QR code positioning model; the surface QR code is positioned using the trained surface QR code positioning model. The method for locating a curved QR code in a complex environment provided by the invention solves the problem that the current algorithm is difficult to locate the curved QR code, and takes into account the influence of complex environmental factors. detection.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a curved surface QR code positioning method based on an SSD network model.
Background
With the rapid development of information technology and the popularization of smart phones, the market application range of the QR code is continuously expanded and is seen everywhere in our lives. The application improves the convenience degree of our life, and is closely related to our life. However, due to various application scenarios and different acquisition conditions of the QR code, the identification process of the QR code is more complicated, and the conventional identification method cannot meet the needs of people in production and life, which requires improvement of the identification technology of the QR code to improve the processing speed and the identification rate.
The QR code positioning aims at extracting the area where the QR code is located as accurately as possible, is the first step of the whole QR code identification process, and plays a vital role as the basis of subsequent distortion correction and decoding. With the complexity of the application scene of the QR code and the diversification of the acquisition environment, various distortion conditions of the QR code occur in practical application, wherein the condition of the curved surface distorted QR code image is complex, and in addition, the QR code is influenced by the complex environment, and the requirements on the performance of the positioning algorithm are higher and higher.
The development of QR code positioning algorithm is from positioning according to QR code image finding graphics to positioning by taking QR code positioning as a target detection task, and then to processing by taking QR code positioning as an image classification problem. The researches greatly improve the positioning performance of the QR code, however, the methods basically process the undistorted or plane-distorted QR code, and the study on the curved surface distortion is less.
Disclosure of Invention
The invention provides a curved surface QR code positioning method based on an SSD network model, aiming at solving the problem that the existing QR code positioning method is difficult to accurately position the area where a QR code is located in an image under the conditions of complex environment and curved surface distortion of the QR code.
In order to achieve the above purpose, the technical means adopted is as follows:
a curved surface QR code positioning method based on an SSD network model comprises the following steps:
s1, establishing a curved surface QR code data set and preprocessing the curved surface QR code data set;
s2, building a PPN-SSD frame, wherein the PPN-SSD frame takes an inclusion network as a basic network, and an auxiliary network layer is added to serve as a feature extraction layer and a classification layer of the network;
s3, performing model training by using the PPN-SSD frame to obtain a curved surface QR code positioning model;
and S4, positioning the curved surface QR code by using the trained curved surface QR code positioning model.
Preferably, the preprocessing in step S1 specifically includes: labeling the curved surface QR code dataset, and respectively filling the number, the length, the width, the target category and the target position information of each image into a csv format file, so as to convert the curved surface QR code dataset into the csv format file; and dividing the converted file data into a training set and a testing set.
Preferably, in the step S2, the PPN-SSD framework performs feature extraction by using a pre-trained model of the inclusion v3 network on the ImageNet dataset as a base network.
Preferably, in the step S2, the feature extraction layer of the PPN-SSD framework includes a convolutional layer "mixed 1_ c" layer and five maximum pooling layers "max _ pool _ 1" layer, "max _ pool _ 2" layer, "max _ pool _ 3" layer, "max _ pool _ 4" layer and "max _ pool _ 5" layer.
Preferably, in the step S2, the classification layer of the PPN-SSD framework employs a convolution shared across all scales to predict the category score and the candidate box position information of the target image.
Preferably, the step S3 specifically includes:
inputting the training set as training input data, and performing data augmentation operation on the training set;
initializing the network: using a pre-training model obtained by training the inclusion v3 network on the ImageNet data set as an initial value of the network parameter to obtain a basic network;
after passing through the middle layer of the basic network, performing feature extraction by using the five maximum pooling layers;
based on the extracted feature map, predicting the category score and the candidate frame position information of the target image through the classification layer to obtain a predicted target frame;
calculating the error between the predicted target frame and the real target frame of the target image, and then updating the weight of an error function by a back propagation algorithm;
and performing iterative training until the loss value of the error function is smaller than a preset threshold value or the maximum iteration times is reached, and stopping training to obtain the curved surface QR code positioning model.
Preferably, in the step S3, the data augmentation operation includes: randomly selecting the images of the training set to adjust the contrast, brightness and saturation, rotating, twisting, cutting or shielding the images in the training set, adding Gaussian noise to the images in the training set, and turning the images in the training set up and down or turning the images in a mirror image mode.
Preferably, in step S3, the error function is specifically:
where N is the number of matching predicted target boxes, LconfIs a loss of confidence, LlocIs the loss of position, x represents the target object, c represents the confidenceDegree, l represents the predicted target frame, g represents the real target frame, and α is the weight occupied by the positioning loss.
Preferably, the step S4 further includes:
carrying out curved surface QR code positioning on the test set by using a trained curved surface QR code positioning model, and setting IoU coefficient threshold values, namely predicting the area intersection ratio between the target frame and the real target frame by using the detection result; if the IoU coefficient of the detection result is larger than the set IoU coefficient threshold value, outputting the detection result; otherwise, the detection result is discarded.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the method is designed aiming at the curved surface QR code positioning in the complex environment, adopts a forward-propagation deep convolutional neural network PPN-SSD frame, and can extract the semantic features of deep images in a self-adaptive manner according to object categories due to the strong feature learning capability, so that the target detection algorithm has good adaptability and robustness. The curved surface QR code positioning model can accurately position the area of the QR code in the image under the influence of factors such as disordered background, over-long distance between image acquisition equipment and the QR code, partial shielding, poor illumination condition, motion blur and the like, and has strong adaptability when the QR code has curved surface distortion. The method solves the problem that the existing algorithm is difficult to position the curved surface QR code, considers the influence of complex environmental factors, has high positioning accuracy and high detection speed, and realizes the real-time detection of the curved surface QR code.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is a diagram of a PPN-SSD-attachment network structure according to the present invention.
FIG. 3 is a flowchart of a method for obtaining a curved QR code positioning model by performing model training using a PPN-SSD framework in the present invention.
FIG. 4 is a diagram illustrating IoU coefficients according to the present invention.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;
it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
A curved surface QR code positioning method based on SSD network model, as shown in fig. 1 and 2, includes the following steps:
s1, establishing a curved surface QR code data set and preprocessing the curved surface QR code data set;
because the curved surface QR code positioning method provided by this embodiment considers the influence of the complex environment factors, in this step, an image data set that includes various complex environments as much as possible is established, that is, a curved surface QR code image is captured in complex environments such as a cluttered background, an excessively long distance between an image acquisition device and the QR code, a partial occlusion, a poor illumination condition, and a motion blur, and the file name of the image data is named as the format from 10001. jpg.
Marking the established curved surface QR code dataset, and respectively filling the number, the length, the width, the target category and the target position information of each image into a csv format file, thereby converting the curved surface QR code dataset into the csv format file; and dividing the converted file data into a training set and a testing set.
S2, building a PPN-SSD frame, wherein the PPN-SSD frame takes an inclusion network as a basic network, and an auxiliary network layer is added to serve as a feature extraction layer and a classification layer of the network;
the PPN (Powing Pyramid network) -SSD (Single Shot Multi Box Detector) framework is a forward-propagation deep convolutional neural network, a series of default prediction boxes with different sizes and proportions are generated on each point of each extracted feature map to obtain a target candidate region, and then a convolution shared on all scales is used for predicting category scores and candidate box position information. The built PPN-SSD framework is composed of a network layer of an inclusion network model, a convolution layer and five pooling layers. The concrete construction is as follows:
a basic network stage: the dimension of input data is 480 × 3, after the data is input, the data is processed by an inclusion network, namely, a series of stage operations such as a convolution layer, a Batch Normalization layer, a ReLU layer and a pooling layer are performed, and the size of an output feature map is 35 × 35 × 512;
and (3) a maximum pooling stage: the stage comprises five maximum pooling layers, namely a "max _ pool _ 1" layer, a "max _ pool _ 2" layer, a "max _ pool _ 3" layer, a "max _ pool _ 4" layer and a "max _ pool _ 5" layer, wherein the maximum pooling operation of each layer is that the filter size is 2 × 2, the step size is 2, and the output feature map sizes are 17 × 17 × 512, 8 × 8 × 512, 4 × 4 × 512, 2 × 2 × 512 and 1 × 1 × 512 respectively;
in the PPN-SSD framework of this embodiment, a convolutional layer "mixed 1_ c" layer and five maximum pooling layers "max _ pool _ 1" layer, "max _ pool _ 2" layer, "max _ pool _ 3" layer, "max _ pool _ 4" layer and "max _ pool _ 5" layer are selected as feature extraction layers.
And generating a series of default prediction frames with different sizes and proportions on each point of each extracted feature map to obtain a target candidate region.
And predicting the obtained target candidate region by using convolution shared on all scales to obtain category scores and candidate frame position information, eliminating redundant prediction frames, and obtaining a detection result with the score being larger than a set threshold value.
The classification layer of the PPN-SSD framework of the present embodiment employs one convolution shared across all scales to predict the category score and candidate box position information of the target image.
S3, performing model training by using the PPN-SSD frame to obtain a curved surface QR code positioning model; in the training stage, firstly, matching a default prediction frame with a marked real target frame by the algorithm, and if the matching value is greater than a set threshold value, determining that the matching is successful, wherein the prediction target frame is a positive sample; otherwise, it is a negative sample. As shown in fig. 3, the specific training process is as follows:
inputting the training set as training input data, and performing data augmentation operation on the training set; wherein the data augmentation operation comprises: randomly selecting images of the training set to adjust the contrast, brightness and saturation, rotating, distorting, cutting or shielding the images in the training set, adding Gaussian noise to the images in the training set, and turning the images in the training set up and down or turning the images in a mirror image manner, so that the data set is improved.
Initializing the network: using a pre-training model obtained by training the inclusion v3 network on the ImageNet data set as an initial value of the network parameter to obtain a basic network;
after passing through the middle layer of the basic network, performing feature extraction by using the five maximum pooling layers;
based on the extracted feature map, predicting the category score and the candidate frame position information of the target image through the classification layer to obtain a predicted target frame;
calculating the error between the predicted target frame and the real target frame of the target image, and then updating the weight of an error function by a back propagation algorithm;
performing iterative training until the loss value of the error function is smaller than a preset threshold or the maximum iteration times is reached, and stopping training to obtain an optimal network model, namely a curved surface QR code positioning model;
where the loss of the PPN-SSD framework consists of two parts, localization and classification, the error function is thus defined as a weighted sum of localization error and confidence error. The calculation method is as follows:
where N is the number of matching predicted target boxes, LconfIs a loss of confidence, LlocThe method comprises the following steps of determining the positioning loss, wherein x represents a target object, c represents confidence coefficient, l represents a predicted target frame, g represents a real target frame, and alpha is the weight occupied by the positioning loss.
S4, positioning the curved surface QR code by using the trained curved surface QR code positioning model;
in this embodiment, a curved surface QR code positioning test is performed on the test set by using a trained curved surface QR code positioning model. And directly outputting the detection result of the curved surface QR code positioning model on the test set so as to evaluate the performance of the network. In this embodiment, the coefficient threshold IoU is set to 0.5, and if the coefficient IoU of the detection result (as shown in fig. 4, i.e. the intersection ratio of the regions between the predicted target frame and the actual target frame of the detection result) is greater than the set coefficient threshold IoU, the detection result is output; otherwise, the detection result is discarded.
Through tests, the curved surface QR code positioning method can cope with various complex environment conditions, such as situations of background clutter, too far distance between image acquisition equipment and a QR code, partial shielding, poor illumination condition, motion blur and the like, realizes real-time detection, achieves higher positioning accuracy, and has better adaptability and robustness.
The terms describing positional relationships in the drawings are for illustrative purposes only and are not to be construed as limiting the patent;
it should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (6)
1. A curved surface QR code positioning method based on an SSD network model is characterized by comprising the following steps:
s1, establishing a curved surface QR code data set and preprocessing the curved surface QR code data set;
s2, building a PPN-SSD frame, wherein the PPN-SSD frame takes an inclusion network as a basic network, and an auxiliary network layer is added to serve as a feature extraction layer and a classification layer of the network;
the PPN-SSD framework adopts a pre-training model of an inclusion v3 network on an ImageNet data set as a basic network to extract features;
the feature extraction layer of the PPN-SSD frame comprises a convolution layer 'mixed 1_ c' layer and five maximum pooling layers 'max _ pool _ 1' layer, 'max _ pool _ 2' layer, 'max _ pool _ 3' layer, 'max _ pool _ 4' layer and 'max _ pool _ 5' layer as feature extraction layers;
generating a series of default prediction frames with different sizes and different proportions on each point of each extracted feature map to obtain a target candidate area;
predicting the obtained target candidate area by using a convolution shared on all scales to obtain category scores and candidate frame position information, eliminating redundant prediction frames, and obtaining a detection result with the score larger than a set threshold value;
the classification layer of the PPN-SSD framework adopts a convolution shared on all scales to predict the category score and the candidate frame position information of the target image;
s3, performing model training by using the PPN-SSD frame to obtain a curved surface QR code positioning model;
and S4, positioning the curved surface QR code by using the trained curved surface QR code positioning model.
2. The curved surface QR code positioning method based on the SSD network model according to claim 1, wherein the preprocessing of step S1 specifically comprises: labeling the curved surface QR code dataset, and respectively filling the number, the length, the width, the target category and the target position information of each image into a csv format file, so as to convert the curved surface QR code dataset into the csv format file; and dividing the converted file data into a training set and a testing set.
3. The curved surface QR code positioning method based on the SSD network model according to claim 2, wherein the step S3 specifically comprises:
inputting the training set as training input data, and performing data augmentation operation on the training set;
initializing the network: using a pre-training model obtained by training the inclusion v3 network on an ImageNet data set as an initial value of a parameter of the curved surface QR code positioning model to obtain a basic network;
after passing through the middle layer of the basic network, performing feature extraction by using the five maximum pooling layers;
based on the extracted feature map, predicting the category score and the candidate frame position information of the target image through the classification layer to obtain a predicted target frame;
calculating the error between the predicted target frame and the real target frame of the target image, and then updating the weight of an error function by a back propagation algorithm;
and performing iterative training until the loss value of the error function is smaller than a preset threshold value or the maximum iteration times is reached, and stopping training to obtain the curved surface QR code positioning model.
4. The method for positioning the QR code based on the SSD network model according to claim 3, wherein in the step S3, the data augmentation operation comprises: randomly selecting the images of the training set to adjust the contrast, brightness and saturation, rotating, twisting, cutting or shielding the images in the training set, adding Gaussian noise to the images in the training set, and turning the images in the training set up and down or turning the images in a mirror image mode.
5. The method for positioning the QR code according to claim 4, wherein in the step S3, the error function is specifically:
where N is the number of matching predicted target boxes, LconfIs a loss of confidence, LlocThe method comprises the following steps of determining the positioning loss, wherein x represents a target object, c represents confidence coefficient, l represents a predicted target frame, g represents a real target frame, and alpha is the weight occupied by the positioning loss.
6. The curved surface QR code positioning method based on SSD network model according to claim 2, wherein the step S4 further comprises:
carrying out curved surface QR code positioning on the test set by using a trained curved surface QR code positioning model, and setting IoU coefficient threshold values, namely predicting the area intersection ratio between the target frame and the real target frame by using the detection result; if the IoU coefficient of the detection result is larger than the set IoU coefficient threshold value, outputting the detection result; otherwise, the detection result is discarded.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010393000.8A CN111652012B (en) | 2020-05-11 | 2020-05-11 | A Surface QR Code Localization Method Based on SSD Network Model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010393000.8A CN111652012B (en) | 2020-05-11 | 2020-05-11 | A Surface QR Code Localization Method Based on SSD Network Model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111652012A CN111652012A (en) | 2020-09-11 |
| CN111652012B true CN111652012B (en) | 2021-10-29 |
Family
ID=72346847
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010393000.8A Active CN111652012B (en) | 2020-05-11 | 2020-05-11 | A Surface QR Code Localization Method Based on SSD Network Model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111652012B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110659664A (en) * | 2019-08-02 | 2020-01-07 | 杭州电子科技大学 | SSD-based method for high-precision identification of small objects |
| CN110765803A (en) * | 2019-10-11 | 2020-02-07 | 首都医科大学宣武医院 | An invisible two-dimensional code and its identification system |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11093745B2 (en) * | 2017-05-09 | 2021-08-17 | Blue River Technology Inc. | Automated plant detection using image data |
| CN107563372B (en) * | 2017-07-20 | 2021-01-29 | 济南中维世纪科技有限公司 | License plate positioning method based on deep learning SSD frame |
| CN107705324A (en) * | 2017-10-20 | 2018-02-16 | 中山大学 | A kind of video object detection method based on machine learning |
| CN108154105B (en) * | 2017-12-21 | 2020-08-07 | 深圳先进技术研究院 | Underwater biological detection and identification method and device, server and terminal equipment |
| US20190333302A1 (en) * | 2018-06-07 | 2019-10-31 | Ysehak Kagnew | Smart and Secure Delivery Door for Packages |
| CN109684967A (en) * | 2018-12-17 | 2019-04-26 | 东北农业大学 | A kind of soybean plant strain stem pod recognition methods based on SSD convolutional network |
| CN109886997B (en) * | 2019-01-23 | 2023-07-11 | 平安科技（深圳）有限公司 | Identification frame determining method and device based on target detection and terminal equipment |
| CN110598687A (en) * | 2019-09-18 | 2019-12-20 | 上海眼控科技股份有限公司 | Vehicle identification code detection method and device and computer equipment |
-
2020
- 2020-05-11 CN CN202010393000.8A patent/CN111652012B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110659664A (en) * | 2019-08-02 | 2020-01-07 | 杭州电子科技大学 | SSD-based method for high-precision identification of small objects |
| CN110765803A (en) * | 2019-10-11 | 2020-02-07 | 首都医科大学宣武医院 | An invisible two-dimensional code and its identification system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111652012A (en) | 2020-09-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109977918B (en) | An Optimization Method for Object Detection and Localization Based on Unsupervised Domain Adaptation | |
| WO2021036059A1 (en) | Image conversion model training method, heterogeneous face recognition method, device and apparatus | |
| CN111950453A (en) | Optional-shape text recognition method based on selective attention mechanism | |
| JP2020508531A (en) | Image quality evaluation method and image quality evaluation system | |
| CN111967470A (en) | Text recognition method and system based on decoupling attention mechanism | |
| CN114663346A (en) | Strip steel surface defect detection method based on improved YOLOv5 network | |
| CN111079683A (en) | Remote sensing image cloud and snow detection method based on convolutional neural network | |
| CN113762269B (en) | Chinese character OCR recognition method, system and medium based on neural network | |
| CN105469047A (en) | Chinese detection method based on unsupervised learning and deep learning network and system thereof | |
| CN113487610B (en) | Herpes image recognition method and device, computer equipment and storage medium | |
| CN116844217B (en) | Image processing system and method for generating face data | |
| CN115861462B (en) | Training method and device for image generation model, electronic equipment and storage medium | |
| CN118350464A (en) | Conversational target positioning method and device based on arbitrary granularity text input | |
| CN116486419A (en) | A Calligraphy Character Recognition Method Based on Siamese Convolutional Neural Network | |
| CN110991412A (en) | Face recognition method and device, storage medium and electronic equipment | |
| CN110163156A (en) | It is a kind of based on convolution from the lip feature extracting method of encoding model | |
| CN116012709A (en) | High-resolution remote sensing image building extraction method and system | |
| CN116310128A (en) | Dynamic environment monocular multi-object SLAM method based on instance segmentation and three-dimensional reconstruction | |
| CN116416503A (en) | Small sample target detection method, system and medium based on multi-mode fusion | |
| CN114241495B (en) | Data enhancement method for off-line handwritten text recognition | |
| CN116071307A (en) | Pavement defect detection model building method, detection method, storage medium and equipment | |
| Tian et al. | BAN, a barcode accurate detection network | |
| CN115292538A (en) | Map line element extraction method based on deep learning | |
| CN111652012B (en) | A Surface QR Code Localization Method Based on SSD Network Model | |
| CN112991281A (en) | Visual detection method, system, electronic device and medium |
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 |