CN111832401A - Electronic marking recognition method - Google Patents
Electronic marking recognition method Download PDFInfo
- Publication number
- CN111832401A CN111832401A CN202010498443.3A CN202010498443A CN111832401A CN 111832401 A CN111832401 A CN 111832401A CN 202010498443 A CN202010498443 A CN 202010498443A CN 111832401 A CN111832401 A CN 111832401A
- Authority
- CN
- China
- Prior art keywords
- image
- neural network
- convolutional neural
- training
- deep convolutional
- 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.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 21
- 238000012360 testing method Methods 0.000 claims abstract description 65
- 238000012549 training Methods 0.000 claims abstract description 35
- 238000001514 detection method Methods 0.000 claims abstract description 18
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 238000004140 cleaning Methods 0.000 claims abstract description 6
- 238000013527 convolutional neural network Methods 0.000 claims description 43
- 230000004913 activation Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000012937 correction Methods 0.000 claims description 8
- 238000002790 cross-validation Methods 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 8
- 238000003709 image segmentation Methods 0.000 claims description 7
- 230000000694 effects Effects 0.000 abstract description 4
- 238000002372 labelling Methods 0.000 abstract description 4
- 238000013528 artificial neural network Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 19
- 238000004590 computer program Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 4
- 230000006872 improvement Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 2
- 230000003213 activating effect Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/40—Document-oriented image-based pattern recognition
- G06V30/41—Analysis of document content
- G06V30/413—Classification of content, e.g. text, photographs or tables
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Educational Administration (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- Educational Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Game Theory and Decision Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Primary Health Care (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to an electronic marking recognition method, which comprises the steps of constructing an image data set for target detection, mainly comprising the steps of collecting, cleaning and labeling real images, and training and testing a deep convolution neural network, and mainly comprising the steps of preprocessing images, designing a network structure, training, testing, actually measuring and determining a network model with an optimal effect, wherein the third part is used for using the optimal model, mainly comprising the steps of giving out coordinates and classification information of output images, eliminating interference factors and improving marking recognition accuracy, so that no special printing requirements are required on a test paper, a teacher can use any dark color pen to correct in the test paper, and all the places can be recognized, the requirements on the background of the test paper are reduced, the requirements on the definition of the collected images are reduced, and the requirements on shooting equipment are reduced.
Description
Technical Field
The invention relates to the technical field of information processing, in particular to an electronic marking identification method.
Background
The on-line examination paper marking is a novel examination paper marking method which is based on a computer network technology and an electronic scanning technology, aims to control subjective questions and examination paper marking errors and realize a test fairness principle, controls the information points filled by examinees to be automatically judged and classified by objective questions through a computer program, scans and sends the subjective questions to the computer program after paper correction is carried out on the subjective questions by an examination paper teacher, identifies and scores the examinee's answer paper with correction symbols after correction by the teacher through the computer program, and finally automatically counts and synthesizes examinee scores through the computer program. The identification and detection method in the existing automatic paper marking system has high requirements on the background and definition of answer sheets or test papers, requires that the surface of the test paper is clean and tidy, cannot be smeared, and needs a specified writing pen, otherwise, the identification and detection results are influenced, and the automatic paper marking fails.
Disclosure of Invention
The invention aims to solve the technical problem of providing a novel electronic paper marking identification method for detecting the correction symbols of teachers in test paper processing and effectively and accurately reading the target positions and target classification information of the correction symbols.
In order to solve the technical problems, the technical scheme provided by the invention is as follows: an electronic marking recognition method comprises the following steps:
step one, constructing an image data set for target detection, wherein the image data set for target detection comprises the steps of carrying out image acquisition on test paper image data obtained by an image segmentation function in an examination paper marking system, cleaning the image data after image acquisition, discarding unqualified samples, marking position information of correction symbols and classification information on the cleaned image data,
secondly, training and testing the deep convolutional neural network on the image data set of the target detection, wherein the training and testing of the deep convolutional neural network comprises preprocessing the image, designing, training and testing the deep convolutional neural network on the preprocessed image to obtain an optimal network model,
the design, training and test of the deep convolutional neural network comprise that the deep convolutional neural network is designed for extracting characteristics according to the quantity, the characteristic complexity and the computing resources of images in a data set, the characteristic graphs obtained by deconvolution and combination of characteristic graphs of a shallow layer, a middle layer and a deep layer of the convolutional neural network are used for detecting a target, samples in the data set are divided into a training set, a cross validation set and a test set for deep convolutional neural network training, a model after the deep convolutional neural network training is deployed on the precision of pictures collected on a test line in a server, and a network model with optimal performance is determined according to a test result,
thirdly, predicting the optimal network model input image, wherein the method for predicting the optimal network model input image comprises the steps of intercepting an image area to be identified in a test paper by an image segmentation function in an examination paper marking system, acquiring and preprocessing the intercepted image, discarding unqualified samples, predicting by using the optimal network model to obtain coordinate information and classification of symbols in the test paper,
and fourthly, counting the scores of the test paper according to the result returned by the optimal network model, and analyzing and displaying the answering conditions.
After adopting the structure, the invention has the following advantages: by using an image data set for constructing target detection, training and testing a deep convolutional neural network, determining a network model with an optimal effect and using the optimal model, interference factors are eliminated, and the identification accuracy of paper marking is improved, so that no special printing requirements are required on a test paper, a teacher can use any dark color pen to correct in the test paper, positions can be identified, the requirements on the background of the test paper are reduced, the requirements on the definition of collected images are reduced, and the requirements on shooting equipment are reduced.
As an improvement, an activation function formula of the convolution calculation of the deep convolution neural network is set as a Leaky Relu activation function, a loss function formula of the convolution calculation of the deep convolution neural network comprises confidence coefficient loss, classification loss and boundary box coordinate loss, wherein the boundary box coordinate loss is set as GIoUlossAnd (4) a calculation method.
As an improvement, the image preprocessing comprises means reduction, variance reduction, random rotation and bilinear interpolation.
As an improvement, the number of training sets trained by the deep convolutional neural network accounts for 80%, the number of cross validation sets accounts for 10%, and the number of test sets accounts for 10%, and the training purpose is achieved by setting hyper-parameters in the convolutional neural network.
Detailed Description
The present invention is described in further detail below.
An electronic marking recognition method comprises the following steps:
step one, constructing an image data set for target detection, wherein the image data set for target detection comprises the steps of carrying out image acquisition on test paper image data obtained by an image segmentation function in an examination paper marking system, cleaning the image data after image acquisition, discarding unqualified samples, marking position information of correction symbols and classification information on the cleaned image data,
secondly, training and testing the deep convolutional neural network on the image data set of the target detection, wherein the training and testing of the deep convolutional neural network comprises preprocessing the image, designing, training and testing the deep convolutional neural network on the preprocessed image to obtain an optimal network model,
the design, training and test of the deep convolutional neural network comprise that the deep convolutional neural network is designed for extracting characteristics according to the quantity, the characteristic complexity and the computing resources of images in a data set, the characteristic graphs obtained by deconvolution and combination of characteristic graphs of a shallow layer, a middle layer and a deep layer of the convolutional neural network are used for detecting a target, samples in the data set are divided into a training set, a cross validation set and a test set for deep convolutional neural network training, a model after the deep convolutional neural network training is deployed on the precision of pictures collected on a test line in a server, and a network model with optimal performance is determined according to a test result,
thirdly, predicting the optimal network model input image, wherein the method for predicting the optimal network model input image comprises the steps of intercepting an image area to be identified in a test paper by an image segmentation function in an examination paper marking system, acquiring and preprocessing the intercepted image, discarding unqualified samples, predicting by using the optimal network model to obtain coordinate information and classification of symbols in the test paper,
and fourthly, counting the scores of the test paper according to the result returned by the optimal network model, and analyzing and displaying the answering conditions.
Setting an activation function formula of the deep convolutional neural network convolution calculation as a Leaky Relu activation function, setting a loss function formula of the deep convolutional neural network convolution calculation as a confidence coefficient loss, a classification loss and a bounding box coordinate loss, wherein the bounding box coordinate loss is set as GIoUlossAnd (4) a calculation method.
The image preprocessing comprises mean value reduction, variance reduction, random rotation and bilinear interpolation.
The number of training sets trained by the deep convolutional neural network accounts for 80%, the number of cross validation sets accounts for 10%, and the number of test sets accounts for 10%.
When the method is implemented specifically, the target detection content in the intelligent marking system has three parts. The first part is to construct an image data set for target detection, and mainly comprises the steps of collecting, cleaning and labeling of real images; the second part is training and testing of the deep convolutional neural network, and mainly comprises image preprocessing, network structure design, training, testing, actual measurement and determination of a network model with an optimal effect; the third part uses the optimal model, mainly including the coordinates and classification information of the given output image. The method comprises the following specific steps:
(1) constructing an image dataset for object detection
S1-1): image acquisition
About 3 thousands of image data sets are collected according to the region to be detected of the test paper obtained by the image segmentation function in the paper marking system.
S1-2): image cleaning
And detecting image samples in the data set, and discarding unqualified samples.
S1-3): labeling
And labeling the image data in the data set by using software, wherein the labeled content comprises the position information and the classification information of the batch modification symbols.
(2) Training and testing of deep convolutional neural networks
S2-1): image pre-processing
All images are subjected to preprocessing operation, such as mean value reduction, variance reduction, random rotation, bilinear interpolation and the like.
S2-2): design of deep convolutional neural network
And designing a 28-layer deep convolutional neural network according to the number of images in the data set, the complexity of the features and the computing resources for extracting the features, wherein feature maps obtained by deconvoluting and combining feature maps of shallow, medium and deep layers of the convolutional neural network are used for detecting a target.
It mainly includes the following functions:
(1-3-1) activation function
Activation functions have a very important role in deep convolutional neural network models to learn and understand very complex and nonlinear functions. The activating function adds nonlinear factors and solves the problem which cannot be solved by a linear model. The target detection network uses Leaky Relu as an activation function, and the mathematical formula is shown as follows.
Wherein λiIs a very small constant.
(1-3-2) loss function
Confidence coefficient loss, classification loss and boundary box coordinate loss are respectively designed in the network, wherein the boundary box coordinate loss uses GIoUlossSee equation (2-2).
Wherein lambda is the confidence of judging whether there is an object in the grid, GroudTruthwAnd GroundTruthhThe width and height of the real box are indicated, and the width and height of the predicted box are indicated. The GIoU is an improvement over IoU (intersection ratio of bounding boxes), and its formula is as follows:
where C is the minimum convex set of bounding box A and bounding box B.
S2-3): deep convolutional neural network training
Dividing samples in a data set into a training set, a cross validation set and a test set: wherein the number of training sets is 80%, the number of cross validation sets is 10%, and the number of test sets is 10%. The training purpose is achieved by setting hyper-parameters in the convolutional neural network.
S2-4): testing of deep convolutional neural networks
And deploying the model in the S (2-3) in a server for testing the accuracy of the collected pictures on the line.
S2-5): obtaining an optimal network model
And determining a network model which performs the best according to the test result in the S2-4), and applying the network model to the actual scene.
(3) Predicting an input image using an optimal model
S3-1) obtaining the image area to be predicted in the test paper
The method of acquiring the image is the same as the method used to construct the data set.
S3-2): image pre-processing
The same processing method as that of S2-1).
S3-3): predicting location coordinates and classification of objects using models
And the method is applied in an actual scene, and coordinate information and classification of the symbols in the test paper are obtained.
S3-4): statistical scoring and analysis
According to the result returned by the network model, the score of the test paper is counted and the answering condition is analyzed, so that an important basis is provided for the functions of class learning, student portrait and the like.
The method and the device have the advantages that the image data set for constructing the target detection, the training and testing of the deep convolutional neural network, the network model with the optimal effect and the use of the optimal model are used, interference factors are eliminated, and the identification accuracy rate of the paper marking is improved, so that no special printing requirements are required on the test paper, a teacher can use any dark color pen to correct the position in the test paper, the position can be identified, the requirements on the background of the test paper are reduced, the requirements on the definition of the acquired image are reduced, and the requirements on shooting equipment are reduced.
The present invention and its embodiments have been described above, but the description is not limitative, and the actual structure is not limited thereto. It should be understood that those skilled in the art should understand that they can easily make various changes, substitutions and alterations herein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (4)
1. An electronic marking recognition method is characterized in that: it comprises the following steps:
step one, constructing an image data set for target detection, wherein the image data set for target detection comprises the steps of carrying out image acquisition on test paper image data obtained by an image segmentation function in an examination paper marking system, cleaning the image data after image acquisition, discarding unqualified samples, marking position information of correction symbols and classification information on the cleaned image data,
secondly, training and testing a deep convolutional neural network on the image data set of the target detection, wherein the training and testing of the deep convolutional neural network comprises preprocessing the image, designing, training and testing the deep convolutional neural network on the preprocessed image to obtain an optimal network model, the designing, training and testing of the deep convolutional neural network comprises designing the deep convolutional neural network for extracting features according to the number, the feature complexity and the computing resources of the image in the data set, using the feature maps obtained by deconvolution and combination of the feature maps of the shallow, the medium and the deep layers of the convolutional neural network for detecting the target, dividing the samples in the data set into a training set, a cross validation set and a testing set for carrying out deep convolutional neural network training, deploying the model trained by the deep convolutional neural network on the image precision collected on a test line in a server, determining a network model with optimal performance according to the test result,
thirdly, predicting the optimal network model input image, wherein the method for predicting the optimal network model input image comprises the steps of intercepting an image area to be identified in a test paper by an image segmentation function in an examination paper marking system, acquiring and preprocessing the intercepted image, discarding unqualified samples, predicting by using the optimal network model to obtain coordinate information and classification of symbols in the test paper,
and fourthly, counting the scores of the test paper according to the result returned by the optimal network model, and analyzing and displaying the answering conditions.
2. The electronic scoring recognition method of claim 1, wherein: setting an activation function formula of the deep convolutional neural network convolution calculation as a Leaky Relu activation function, setting a loss function formula of the deep convolutional neural network convolution calculation as a confidence coefficient loss, a classification loss and a bounding box coordinate loss, wherein the bounding box coordinate loss is set as GIoUlossAnd (4) a calculation method.
3. The electronic scoring recognition method of claim 1, wherein: the image preprocessing comprises mean value reduction, variance reduction, random rotation and bilinear interpolation.
4. The electronic scoring recognition method of claim 1, wherein: the number of training sets trained by the deep convolutional neural network accounts for 80%, the number of cross validation sets accounts for 10%, and the number of test sets accounts for 10%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010498443.3A CN111832401A (en) | 2020-06-04 | 2020-06-04 | Electronic marking recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010498443.3A CN111832401A (en) | 2020-06-04 | 2020-06-04 | Electronic marking recognition method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111832401A true CN111832401A (en) | 2020-10-27 |
Family
ID=72897799
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010498443.3A Withdrawn CN111832401A (en) | 2020-06-04 | 2020-06-04 | Electronic marking recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111832401A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112990180A (en) * | 2021-04-29 | 2021-06-18 | 北京世纪好未来教育科技有限公司 | Question judging method, device, equipment and storage medium |
-
2020
- 2020-06-04 CN CN202010498443.3A patent/CN111832401A/en not_active Withdrawn
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112990180A (en) * | 2021-04-29 | 2021-06-18 | 北京世纪好未来教育科技有限公司 | Question judging method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210374466A1 (en) | Water level monitoring method based on cluster partition and scale recognition | |
| CN112699876B (en) | Automatic reading method for various meters of gas collecting station | |
| CN107590498A (en) | A kind of self-adapted car instrument detecting method based on Character segmentation level di- grader | |
| CN108921163A (en) | A kind of packaging coding detection method based on deep learning | |
| CN105787522B (en) | Handwriting-based writing attitude evaluation method and system | |
| CN202939608U (en) | Test paper inspecting system | |
| CN105046200B (en) | Electronic paper marking method based on straight line detection | |
| CN104077577A (en) | Trademark detection method based on convolutional neural network | |
| CN106033535B (en) | Electronic paper marking method | |
| CN112819093A (en) | Man-machine asynchronous recognition method based on small data set and convolutional neural network | |
| CN110414563A (en) | Total marks of the examination statistical method, system and computer readable storage medium | |
| CN115345881B (en) | Pavement disease detection method based on computer vision | |
| CN110288612B (en) | Nameplate positioning and correcting method and device | |
| CN110659637A (en) | Electric energy meter number and label automatic identification method combining deep neural network and SIFT features | |
| CN109523514A (en) | To the batch imaging quality assessment method of Inverse Synthetic Aperture Radar ISAR | |
| CN112489026A (en) | Asphalt pavement disease detection method based on multi-branch parallel convolution neural network | |
| CN115359295A (en) | Decoupling knowledge distillation hardware target detection method and system | |
| CN111652117A (en) | Method and medium for segmenting multi-document image | |
| CN111832401A (en) | Electronic marking recognition method | |
| CN108089753B (en) | Positioning method for predicting fingertip position by using fast-RCNN | |
| CN109472257B (en) | Character layout determining method and device | |
| CN115393861A (en) | Method for accurately segmenting handwritten text | |
| CN110738641A (en) | Image processing and KELM based qualitative detection method for concentration of medical reagent | |
| CN110298236A (en) | A kind of braille automatic distinguishing method for image and system based on deep learning | |
| CN114694128A (en) | Pointer instrument detection method and system based on abstract metric learning |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20201027 |
|
| WW01 | Invention patent application withdrawn after publication |