CN111553423A - Handwriting recognition method based on deep convolutional neural network image processing technology - Google Patents
Handwriting recognition method based on deep convolutional neural network image processing technology Download PDFInfo
- Publication number
- CN111553423A CN111553423A CN202010354929.XA CN202010354929A CN111553423A CN 111553423 A CN111553423 A CN 111553423A CN 202010354929 A CN202010354929 A CN 202010354929A CN 111553423 A CN111553423 A CN 111553423A
- Authority
- CN
- China
- Prior art keywords
- neural network
- convolutional neural
- network
- image
- handwriting
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 45
- 238000012545 processing Methods 0.000 title claims abstract description 39
- 238000005516 engineering process Methods 0.000 title claims abstract description 22
- 238000005457 optimization Methods 0.000 claims description 37
- 238000012549 training Methods 0.000 claims description 33
- 238000012360 testing method Methods 0.000 claims description 23
- 239000000463 material Substances 0.000 claims description 21
- 238000012795 verification Methods 0.000 claims description 18
- 238000010606 normalization Methods 0.000 claims description 14
- 230000011218 segmentation Effects 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 8
- 238000005520 cutting process Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 5
- 238000004821 distillation Methods 0.000 claims description 4
- 239000013598 vector Substances 0.000 claims description 4
- 230000003631 expected effect Effects 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 abstract description 4
- 230000009466 transformation Effects 0.000 description 9
- 239000011159 matrix material Substances 0.000 description 6
- 230000015572 biosynthetic process Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000003786 synthesis reaction Methods 0.000 description 5
- 241001125929 Trisopterus luscus Species 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000003796 beauty Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Character Discrimination (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of image classification, and particularly discloses a handwriting recognition method based on a deep convolutional neural network image processing technology. The deep convolution-based neural network provided by the invention can accurately identify the directly extracted handwritten picture, and has high identification accuracy and high operation speed and convergence speed. The method is suitable for assisting in identifying the handwriting on the answer sheet or the answer sheet and outputting the handwriting as a standard font.
Description
Technical Field
The invention belongs to the technical field of image classification, relates to a handwriting recognition method, and particularly relates to a handwriting recognition method based on a deep convolutional neural network image processing technology.
Background
In recent years, various large-scale examinations such as college entrance examination and national english level examination adopt an online examination paper reading method, so that a large amount of material and labor are saved, and a large amount of time is saved. However, in the current online paper reading, only examinee test papers are photographed and stored, answer areas are sequentially divided according to the spacing distance of the handwriting in each piece of answer handwriting information to form a test quantity set and an answer area set so as to determine the number of the test questions and the number of the effective answer areas, and then each effective answer area of each piece of answer handwriting information is subjected to information identification and statistics to determine the answer of the test questions. The handwriting of each examinee is reserved by the examinees without exception, so that the teacher can sometimes make the examination and the examination trouble and get into a pit for word guessing, and if the examinees always judge the examination papers with beautiful writings, the teacher can also possibly check beauty and fatigue, which leads the subjective evaluation of the teacher to dominate.
With the birth of second generation artificial intelligence learning system transducer Flow developed by google based on distbel f (distributed deep learning platform of google corporation), the system is widely applied to the fields of image recognition, voice recognition and other multiple machine learning and deep learning. The sensor Flow is a system for transmitting a complex data structure to an artificial intelligent Neural Network for analysis and processing, the system identifies images based on a PC, and in addition, a Deep Convolutional Neural Network (DCNN) is also applied to the field of handwritten Chinese character identification. However, for the Chinese characters at present, only the print form can be directly extracted for efficient recognition, and the directly extracted handwriting form cannot be accurately recognized.
Disclosure of Invention
The invention aims to provide a handwriting recognition method based on a deep convolutional neural network image processing technology, so as to solve the problem that the directly extracted handwriting cannot be accurately recognized in the prior art.
In order to achieve the purpose, the technical method comprises the following steps:
a handwriting recognition method based on a deep convolutional neural network image processing technology comprises the following steps:
s1, extracting handwriting picture materials from the homework submitted by the student and storing the handwriting picture materials in a PKL format;
s2, reading data in a PKL format, carrying out binarization and normalization processing on the picture material, and then randomly distributing the processed picture material into a first test set, a first verification set and a first training set according to the proportion;
s3, obtaining a distilled data set by the first training set through a data distillation technology, inputting the distilled data set into a convolutional neural network with configured parameters and structures, and training the convolutional neural network into an image optimization network;
s4, inputting data of the first verification set into an image optimization network, classifying the first verification set by the image optimization network, adjusting parameters of the image optimization network according to classification results, training the image optimization network by the first training set again until the effect of classifying the first verification set reaches an expected effect, obtaining the image optimization network, and testing the classification accuracy of the image optimization network by the first test set;
s5, subdividing the picture material into a second test set, a second training set and a second verification set, building a deep convolutional neural network, training the convolutional neural network by using the second training set, verifying by using the second verification set and the second test set, and continuously adjusting network parameters until the expected accuracy is reached;
and S6, scanning the student test paper or answer sheet to read the student writing area, inputting the scanned picture into the image optimization network obtained in the step S4 after binarization and normalization processing, performing image optimization, performing character segmentation, inputting the image into the deep convolutional neural network, converting the handwriting of the student and outputting the handwriting as a standard font.
As a limitation: the handwritten picture materials in the step S1 include an HWDB1.1 data set, an MNIST data set, and a homemade data set with roman symbols, chemical symbols, and mathematical formulas as contents; the manufacturing process of the self-made data set comprises the following steps: the method comprises the steps of performing character segmentation on pictures containing roman symbols, chemical symbols and mathematical formula operation submitted by students to enable each character to be independently stored into an image, then performing binarization processing, adjusting the pixel size of the image, and manually marking out a standard font of each character.
As a further limitation: the method adopted by the binarization processing in the steps S2 and S4 is otsu method.
As a further limitation: the normalization processing in the steps S2, S4, and S5 employs a method standardized by min-max.
As another limitation: the image optimization network in step S3 is a reverse synthesis space transformation network, and the specific network structure sequentially includes, from top to bottom: two convolutional layers, one pooling layer, two fully-connected layers.
As a further limitation: the deep convolutional neural network in the step S4 has a structure of two convolutional layers, one pooling layer, and two fully-connected layers from top to bottom; and converting the output of the convolutional layer by adopting a Relu activation function, and classifying and outputting the one-dimensional vectors output by the full-link layer by adopting a softmax cross entropy loss method.
As a last definition: the character cutting method in step S6 specifically includes: horizontally projecting the picture, finding an upper limit and a lower limit of each line, and then cutting; and then carrying out vertical projection on each cut line, finding the left and right boundaries of each character, and then cutting a single character.
Due to the adoption of the scheme, compared with the prior art, the invention has the beneficial effects that:
(1) according to the handwriting recognition method based on the deep convolutional neural network image processing technology, a structural framework combining an image optimization network and a convolutional neural network is built and trained to form a complete and trained network framework, so that a directly extracted handwriting picture can be accurately recognized;
(2) according to the handwriting recognition method based on the deep convolutional neural network image processing technology, a large amount of handwriting picture materials are used for training an image optimization network, the optimization speed of the image optimization network on the handwriting picture materials is improved, the convolutional neural network is trained by using high-quality data, the recognition accuracy is improved, and the neural network can efficiently and accurately recognize and output pixel values as standard fonts;
(3) the handwriting recognition method based on the deep convolutional neural network image processing technology provided by the invention has the advantages that the handwriting picture material is stored into a readable RKL format and subjected to binarization processing, the operation speed is increased, normalization processing is performed, the output of a convolutional layer is converted by adopting a Relu activation function, and the convergence speed of the network is increased.
The invention is suitable for assisting in identifying the handwriting on the test paper or the answer sheet and outputting the handwriting as a standard font.
Drawings
The invention is described in further detail below with reference to the figures and the embodiments.
FIG. 1 is a flow chart of network training according to an embodiment of the present invention;
FIG. 2 is a flow chart of handwriting recognition according to an embodiment of the present invention;
FIG. 3 is a handwriting before recognition according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating standard fonts output after recognition according to an embodiment of the present invention.
Detailed Description
The present invention is further described with reference to the following examples, but it should be understood by those skilled in the art that the present invention is not limited to the following examples, and any modifications and variations based on the specific examples of the present invention are within the scope of the claims of the present invention.
Handwriting recognition method based on deep convolutional neural network image processing technology
A handwriting recognition method based on a deep convolutional neural network image processing technology comprises the following steps:
s1, extracting handwriting picture materials from the homework submitted by the student and storing the handwriting picture materials in a PKL format;
in the step, the handwritten picture materials comprise an HWDB1.1 data set, an MNIST data set and a homemade data set which takes Roman symbols, chemical symbols and mathematical formulas as contents; the manufacturing process of the self-made data set comprises the following steps: the method comprises the steps of performing character segmentation on pictures containing roman symbols, chemical symbols and mathematical formula operation submitted by students to enable each character to be independently stored into an image, then performing binarization processing, adjusting the pixel size of the image, and manually marking out a standard font of each character.
And S2, reading data in a PKL format, carrying out binarization and normalization processing on the picture material, and randomly distributing the data into a first test set, a first verification set and a first training set according to the ratio of 1:2: 7.
S3, obtaining a distilled data set by the first training set through a data distillation technology, inputting the distilled data set into a convolutional neural network with configured parameters and structures, and training the convolutional neural network into an image optimization network;
the image optimization network is a reverse synthesis space transformation network, and the network structure is as follows:
Conv(7×7,4)+Conv(7×7,8)+P+FC(48)+FC(8)
wherein Conv represents a convolution layer, P represents a pooling layer, FC represents a full-connection layer, the size and the number of convolution kernels are respectively shown in brackets, and the number of neurons is shown in brackets of the full-connection layer;
the data distillation technology comprises the following specific processes: generating a new data set by back-propagating iterative updates according to the first training set and the initialization weights of the convolutional network, i.e. using the first training set and the initialization weights of the convolutional network
θ0Is a fixed parameter, using a regenerated data setAnd new learning rateThrough one iteration, updatingAndthe formula of (1) is:
andobtained by iterative learning of gradient in advance, needs initialization and fixes parameter theta0UpdateAndobtaining data
The specific training method comprises the following steps: data to be recordedInputting the inverse synthesis space transformation network to obtain new deformation parameters, namely I (x), and the distortion parameters are p ═ p1, p2, … and p9]Affine transformation, and the related secondary coordinate transformation matrix is:
the training set is multiplied by the initialized warping parameter p, i.e. l (x) ═ g (p) i (x), because the coordinates after the warping transformation image calculation are non-integer, which results in image discontinuity, the invention performs bilinear interpolation on the transformed image l (x):
wherein, Vi C(ii) an ith pixel point representing the c sample, (x)i,yi) Denotes the i-th pixel coordinate of the image L (x), W denotesWidth of the image, H represents height of the image, (m, n) represents other points on the image;
inputting the image V (x) into a geometric predictor, wherein the geometric predictor is composed of a neural network or a small convolutional neural network, updating a parameter delta p of the geometric predictor by utilizing a back propagation BP algorithm, the delta p is a forward propagation prediction distortion increment of a backward synthesis space transformation network, and after updating the parameter delta p, further iteratively updating the distortion parameter p in a synthesis mode, namely the distortion parameter p is updatedThe corresponding transformation matrix is, where w (pin) is the initial matrix, w (pout) is the transformation matrix composed of the final updated warping parameters p:
W(pout)=W(ΔP)·W(pin)
multiplying the final matrix W (pout) consisting of p with the training set, i.e.
Im=I(x)·W(Pout)
And outputting the data Im.
S4, inputting the data of the first verification set into the image optimization network, classifying the first verification set by the image optimization network, adjusting parameters of the image optimization network according to classification results, training the image optimization network to achieve an expected effect of classifying the first verification set by means of the first training set again, and testing the classification accuracy of the image optimization network by means of the first test set.
S5, subdividing the picture material into a second test set, a second training set and a second verification set, building a deep convolutional neural network, training the convolutional neural network by using the second training set, verifying by using the second verification set and the second test set, and continuously adjusting network parameters until the expected accuracy is reached;
the structure of the deep convolutional neural network is as follows:
Conv(3×3,8)+Conv(3×3,16)+P+Conv(3×3,32)+Conv(3×3,64)+P+FC(100)+FC(200)
inputting the second training set into a fixed initialized convolutional neural network, extracting the characteristics of input data by the convolutional layer, and converting the output of the convolutional layer by adopting a Relu activation function, wherein the specific operation formula is as follows:
Y=f(U)
wherein Im is the output of the image optimization network, W is a convolution kernel, b is convolution layer bias, U is the convolution layer output, f is a Relu activation function, and Y is the output of the convolution layer output U after the Relu activation function;
the pooling layer performs sampling operation on input data samples in a two-dimensional space, and the specific calculation process is as follows:
the full connection layer reduces the dimension of the input two-dimensional feature matrix to a one-dimensional feature vector;
and classifying and outputting the one-dimensional vectors output by the full-connection layer by adopting a softmax cross entropy loss method.
And S6, scanning the student test paper or answer sheet to read the student writing area, inputting the scanned picture into the image optimization network obtained in the step S4 after binarization and normalization processing, performing image optimization, performing character segmentation, inputting the image into the deep convolutional neural network, converting the handwriting of the student and outputting the handwriting as a standard font.
The character segmentation method specifically comprises the following steps: horizontally projecting the picture, finding an upper limit and a lower limit of each line, and then cutting; and then carrying out vertical projection on each cut line, finding the left and right boundaries of each character, and then cutting a single character.
The method adopted by the binarization processing in the steps S2 and S4 is an otsu method, and the specific method comprises the following steps: for image I (x, y), the segmentation threshold for the foreground (i.e., object) and background is denoted as T, and the ratio of the foreground image to the entire image is denoted as ω0Average gray level mu of0(ii) a The proportion of the background image to the whole image is omega1Average gray of μ1The total average gray scale of the image is recorded as mu, the inter-class variance is recorded as g, and M × N is the total number of pixelsThe number of pixels in the image with the gray value of the pixel less than the threshold value T is recorded as N0The number of pixels having a pixel gray level greater than the threshold T is denoted by N1Then, then
The sum of the foreground pixels and the background pixels is N0+N1=M×N
The total ratio of the background image to the foreground image is omega0+ω1=1
The gray scale integration value in the 0-M gray scale interval is mu ═ mu0*ω0+μ1*ω1
The inter-class variance value is g ═ omega0*(μ-μ0)2+ω1*(μ-μ1)2=ω0*ω1*(μ0-μ1)2
The threshold T which maximizes the inter-class variance can be obtained by adopting a traversal method.
The normalization processing in steps S2, S4, and S5 uses a method of min-max normalization, and maps the data into the [0,1] interval by transforming the original data, and the calculation formula is:
x″=x′*(mx-mi)+mi
in the formula, max and min are the maximum value and the minimum value of one column, x "is the result, and mx and mi are the mapped interval values, that is, mx is 1 and mi is 0.
The network training flowchart of this embodiment is shown in fig. 1, the handwriting recognition process is shown in fig. 2, the handwriting of the "each" character in the student work is extracted, the "each" handwriting is shown in fig. 3, the input is input to the image optimization network obtained in step S4 after binarization and normalization processing for image optimization and character segmentation, and then the input is input to the deep convolutional neural network for converting the handwriting of the student into the standard fonts "each" shown in fig. 4.
Claims (10)
1. A handwriting recognition method based on a deep convolutional neural network image processing technology is characterized by comprising the following steps:
s1, extracting handwriting picture materials from the homework submitted by the student and storing the handwriting picture materials in a PKL format;
s2, reading data in a PKL format, carrying out binarization and normalization processing on the picture material, and then randomly distributing the processed picture material into a first test set, a first verification set and a first training set according to the proportion;
s3, obtaining a distilled data set by the first training set through a data distillation technology, inputting the distilled data set into a convolutional neural network with configured parameters and structures, and training the convolutional neural network into an image optimization network;
s4, inputting data of the first verification set into an image optimization network, classifying the first verification set by the image optimization network, adjusting parameters of the image optimization network according to classification results, training the image optimization network by the first training set again until the effect of classifying the first verification set reaches an expected effect, and testing the classification accuracy of the image optimization network by the first test set;
s5, subdividing the picture material into a second test set, a second training set and a second verification set, building a deep convolutional neural network, training the convolutional neural network by using the second training set, verifying by using the second verification set and the second test set, and continuously adjusting network parameters until the expected accuracy is reached;
and S6, scanning the student test paper or answer sheet to read the student writing area, inputting the scanned picture into the image optimization network obtained in the step S4 after binarization and normalization processing, performing image optimization, performing character segmentation, inputting the image into the deep convolutional neural network, converting the handwriting of the student and outputting the handwriting as a standard font.
2. The handwriting recognition method based on deep convolutional neural network image processing technology of claim 1, wherein the handwriting picture materials in step S1 include HWDB1.1 dataset, MNIST dataset, and homemade dataset with roman symbol, chemical symbol, mathematical formula as content; the manufacturing process of the self-made data set comprises the following steps: the method comprises the steps of performing character segmentation on pictures containing roman symbols, chemical symbols and mathematical formula operation submitted by students to enable each character to be independently stored into an image, then performing binarization processing, adjusting the pixel size of the image, and manually marking out a standard font of each character.
3. The handwriting recognition method based on deep convolutional neural network image processing technique of claim 1 or 2, wherein the method adopted by the binarization processing in steps S2 and S4 is otsu method.
4. The handwriting recognition method based on deep convolutional neural network image processing technique of claim 1 or 2, wherein the normalization process in steps S2, S4 and S5 is a min-max normalization method.
5. The handwriting recognition method based on deep convolutional neural network image processing technique of claim 3, wherein the normalization process in steps S2, S4 and S5 is a method standardized for min-max.
6. The handwriting recognition method based on deep convolutional neural network image processing technology according to any one of claims 1, 2 and 5, wherein the image optimization network in step S3 is an inverse synthetic space transform network, and the specific network structures sequentially from top to bottom are: two convolutional layers, one pooling layer, two fully-connected layers.
7. The handwriting recognition method based on deep convolutional neural network image processing technology of claim 3, wherein the image optimization network in step S3 is an inverse synthetic spatial transform network, and the specific network structures are sequentially from top to bottom: two convolutional layers, one pooling layer, two fully-connected layers.
8. The handwriting recognition method based on deep convolutional neural network image processing technology of claim 4, wherein the image optimization network in step S3 is an inverse synthetic spatial transform network, and the specific network structures are sequentially from top to bottom: two convolutional layers, one pooling layer, two fully-connected layers.
9. The handwriting recognition method based on deep convolutional neural network image processing technology of any one of claims 1, 2, 5, 7 and 8, wherein the structure of the deep convolutional neural network in step S4 is, from top to bottom, two convolutional layers, one pooling layer, two fully-connected layers; and converting the output of the convolutional layer by adopting a Relu activation function, and classifying and outputting the one-dimensional vectors output by the full-link layer by adopting a softmax cross entropy loss method.
10. The handwriting recognition method based on the deep convolutional neural network image processing technology of claim 9, wherein the character segmentation method in step S6 is specifically: horizontally projecting the picture, finding an upper limit and a lower limit of each line, and then cutting; and then carrying out vertical projection on each cut line, finding the left and right boundaries of each character, and then cutting a single character.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010354929.XA CN111553423A (en) | 2020-04-29 | 2020-04-29 | Handwriting recognition method based on deep convolutional neural network image processing technology |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010354929.XA CN111553423A (en) | 2020-04-29 | 2020-04-29 | Handwriting recognition method based on deep convolutional neural network image processing technology |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111553423A true CN111553423A (en) | 2020-08-18 |
Family
ID=72007836
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010354929.XA Pending CN111553423A (en) | 2020-04-29 | 2020-04-29 | Handwriting recognition method based on deep convolutional neural network image processing technology |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111553423A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112102126A (en) * | 2020-08-31 | 2020-12-18 | 湖北美和易思教育科技有限公司 | Student manual ability culture method and device based on neural network |
CN112308058A (en) * | 2020-10-25 | 2021-02-02 | 北京信息科技大学 | Method for recognizing handwritten characters |
CN112507864A (en) * | 2020-12-04 | 2021-03-16 | 河北地质大学 | Credit archive identification method based on convolutional neural network |
CN113076900A (en) * | 2021-04-12 | 2021-07-06 | 华南理工大学 | Test paper head student information automatic detection method based on deep learning |
CN113642550A (en) * | 2021-07-20 | 2021-11-12 | 南京红松信息技术有限公司 | Entropy maximization card-smearing identification method based on pixel probability distribution statistics |
CN114842486A (en) * | 2022-07-04 | 2022-08-02 | 南昌大学 | Handwritten chemical structural formula recognition method, system, storage medium and equipment |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107590497A (en) * | 2017-09-20 | 2018-01-16 | 重庆邮电大学 | Off-line Handwritten Chinese Recognition method based on depth convolutional neural networks |
CN109800746A (en) * | 2018-12-05 | 2019-05-24 | 天津大学 | A kind of hand-written English document recognition methods based on CNN |
-
2020
- 2020-04-29 CN CN202010354929.XA patent/CN111553423A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107590497A (en) * | 2017-09-20 | 2018-01-16 | 重庆邮电大学 | Off-line Handwritten Chinese Recognition method based on depth convolutional neural networks |
CN109800746A (en) * | 2018-12-05 | 2019-05-24 | 天津大学 | A kind of hand-written English document recognition methods based on CNN |
Non-Patent Citations (1)
Title |
---|
TONGZHOU WANG ET AL.: "DATASET DISTILLATION", 《ARXIV:1811.10959V1 [CS.LG] 27 NOV 2018》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112102126A (en) * | 2020-08-31 | 2020-12-18 | 湖北美和易思教育科技有限公司 | Student manual ability culture method and device based on neural network |
CN112308058A (en) * | 2020-10-25 | 2021-02-02 | 北京信息科技大学 | Method for recognizing handwritten characters |
CN112308058B (en) * | 2020-10-25 | 2023-10-24 | 北京信息科技大学 | Method for recognizing handwritten characters |
CN112507864A (en) * | 2020-12-04 | 2021-03-16 | 河北地质大学 | Credit archive identification method based on convolutional neural network |
CN113076900A (en) * | 2021-04-12 | 2021-07-06 | 华南理工大学 | Test paper head student information automatic detection method based on deep learning |
CN113642550A (en) * | 2021-07-20 | 2021-11-12 | 南京红松信息技术有限公司 | Entropy maximization card-smearing identification method based on pixel probability distribution statistics |
CN113642550B (en) * | 2021-07-20 | 2024-03-12 | 南京红松信息技术有限公司 | Entropy maximization card-coating identification method based on pixel probability distribution statistics |
CN114842486A (en) * | 2022-07-04 | 2022-08-02 | 南昌大学 | Handwritten chemical structural formula recognition method, system, storage medium and equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111553423A (en) | Handwriting recognition method based on deep convolutional neural network image processing technology | |
CN109886121B (en) | Human face key point positioning method for shielding robustness | |
CN111291629A (en) | Method and device for recognizing text in image, computer equipment and computer storage medium | |
CN111414906A (en) | Data synthesis and text recognition method for paper bill picture | |
CN113239954B (en) | Attention mechanism-based image semantic segmentation feature fusion method | |
CN111401156B (en) | Image identification method based on Gabor convolution neural network | |
CN110969171A (en) | Image classification model, method and application based on improved convolutional neural network | |
CN110674777A (en) | Optical character recognition method in patent text scene | |
CN113421318B (en) | Font style migration method and system based on multitask generation countermeasure network | |
CN113223025A (en) | Image processing method and device, and neural network training method and device | |
CN115147607A (en) | Anti-noise zero-sample image classification method based on convex optimization theory | |
CN115393861B (en) | Method for accurately segmenting handwritten text | |
CN111523622A (en) | Method for simulating handwriting by mechanical arm based on characteristic image self-learning | |
CN114444565A (en) | Image tampering detection method, terminal device and storage medium | |
CN114092938B (en) | Image recognition processing method and device, electronic equipment and storage medium | |
CN113657377B (en) | Structured recognition method for mechanical bill image | |
CN114972952A (en) | Industrial part defect identification method based on model lightweight | |
CN113963232A (en) | Network graph data extraction method based on attention learning | |
CN114220178A (en) | Signature identification system and method based on channel attention mechanism | |
CN113408418A (en) | Calligraphy font and character content synchronous identification method and system | |
Gayathri et al. | Optical Character Recognition in Banking Sectors Using Convolutional Neural Network | |
CN116612478A (en) | Off-line handwritten Chinese character scoring method, device and storage medium | |
CN109815889A (en) | A kind of across resolution ratio face identification method based on character representation collection | |
CN114550179A (en) | Method, system and equipment for guiding handwriting Chinese character blackboard writing | |
CN114241486A (en) | Method for improving accuracy rate of identifying student information of test paper |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200818 |
|
RJ01 | Rejection of invention patent application after publication |