AU2020103315A4 - A method for digitizing writings in antiquity - Google Patents
A method for digitizing writings in antiquity Download PDFInfo
- Publication number
- AU2020103315A4 AU2020103315A4 AU2020103315A AU2020103315A AU2020103315A4 AU 2020103315 A4 AU2020103315 A4 AU 2020103315A4 AU 2020103315 A AU2020103315 A AU 2020103315A AU 2020103315 A AU2020103315 A AU 2020103315A AU 2020103315 A4 AU2020103315 A4 AU 2020103315A4
- Authority
- AU
- Australia
- Prior art keywords
- single character
- document
- writings
- antiquity
- digitizing
- 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 36
- 238000001514 detection method Methods 0.000 claims abstract description 39
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000013145 classification model Methods 0.000 claims abstract description 15
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 238000002372 labelling Methods 0.000 claims abstract description 4
- 238000005070 sampling Methods 0.000 claims abstract description 4
- 238000010586 diagram Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000005520 cutting process Methods 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims description 3
- 238000005260 corrosion Methods 0.000 claims description 2
- 230000007797 corrosion Effects 0.000 claims description 2
- 238000013518 transcription Methods 0.000 abstract description 4
- 230000035897 transcription Effects 0.000 abstract description 4
- 238000011176 pooling Methods 0.000 description 8
- 238000012360 testing method Methods 0.000 description 6
- 230000004913 activation Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 238000012015 optical character recognition Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000002087 whitening effect Effects 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000012795 verification 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
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06V30/412—Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables
-
- 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
- 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
- G06V30/414—Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06V30/10—Character recognition
- G06V30/28—Character recognition specially adapted to the type of the alphabet, e.g. Latin alphabet
- G06V30/293—Character recognition specially adapted to the type of the alphabet, e.g. Latin alphabet of characters other than Kanji, Hiragana or Katakana
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- Computer Graphics (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for digitizing writings in antiquity. Its specific steps include
collecting data and utilizing collected data to train single character detection model so as to
obtain single character output results; meanwhile, training single character classification model
to get classification result of detected single character and the final document recognition
results can be obtained by combining results of single character detection and recognition;
extracting lines of document layout with the help of graphic morphology and designing
algorithm to solve the problem of biserial interlinear notes, which provides conditions for the
structured output of the document. And finally, outputting the digital results of the document
corresponding to the images to complete the document digitization work. The method of the
invention solves the problems of single character detection in ancient books and documents
with complex layout and dense documents and the existence of stain interference in large
document background. It has the advantages of simple, efficient and high recognition accuracy.
With combination of modem computer information technology and traditional humanistic
culture, the invention plays an important and positive role in digital heritage protection,
information discovery, paper document transcription, etc.
-1/4
|Data acquisition. Collecting
and labeling writings in
antiquity
Training single Training single
character character
detection model classification model Extraction of
and performing a and perfonning a layout lines
single character single character
detection classification.
Document
structuring
Outputting the final
digitized document content
Figure 1 The flow chart of the digitization method of the ancient book documents in the
invention
I c I
a d
Ftl ig ue i t
t~ 4
Cd e
Figure 2An introduction to dataset sampling used in the invention
Description
-1/4
|Data acquisition. Collecting and labeling writings in antiquity
Training single Training single character character detection model classification model Extraction of and performing a and perfonning a layout lines single character single character detection classification.
Document structuring
Outputting the final digitized document content
Figure 1 The flow chart of the digitization method of the ancient book documents in the
invention
Ic I a d
Ftl ig ue i t t~ 4
Cd e
Figure 2An introduction to dataset sampling used in the invention
PATENTS ACT 1990
A method for digitizing writings in antiquity
The invention is described in the following statement:-
A method for digitizing writings in antiquity
The invention relates to the technical field of image precise positioning and classification,
in particular to a method for digitizing writings in antiquity.
With a long history, Chinese culture is broad and profound. Ancient books and
documents contain all the wisdom essence of China's 5000 years history. They are not
only the traditional proof of China's long-standing culture, but also the foundation of the
Chinese nation. More importantly, they are our indispensable spiritual strength. The
historical relics, academic materials and artistic representativeness of ancient books and
documents play an extremely important role in the study of the social style and
development of production, science and culture in ancient China. China has tens of
thousands of ancient books and documents, which record China's long history and
culture, and are very valuable intangible cultural heritage. In order to avoid the aging or
disappearing of ancient books and documents due to the passage of time, and to excavate
and utilize the rich knowledge contained in ancient books and documents, the digitization
of ancient books and documents is particularly important. Optical character recognition
(OCR) technology is closely related to the digitization of ancient books, that is, the
characters on paper can be read out by optical technology and computer technology, so as
to obtain the corresponding text output results.
In recent years, with the development of deep neural network, OCR technology based on
deep learning has achieved remarkable results in fixed format, such as ID card
verification and license plate recognition, which not only reduces the labour cost, but also greatly improves people's work efficiency. However, the research on the transcription of ancient books and documents is developing slowly. The main technical difficulties include the complexity of ancient books and documents layout, the difficulty of extracting structured output information, the fuzzy image, low resolution and serious background interference, which seriously affect the detection and recognition of characters.
Therefore, there is an urgent need for a simple and efficient method of digitizing writings
in antiquity so as to protect ancient books through timely paper document transcription.
The purpose of the invention is to provide a method for digitizing writings in antiquity, so
as to solve the problems existing in the prior art and make the ancient books and
documents be accurately transcribed.
To achieve the above purpose, the invention provides the following scheme:
The invention provides a method for digitizing writings in antiquity, including the
following contents.
Si. Data acquisition. Collecting the image data of ancient books and labelling the image
data with single characters and text lines at the space level to obtain the training dataset.
S2. Training and detection of single character detection model. Pre-processing the
training dataset. Based on the general target detection framework YOLO-v3, setting
anchors with different scales and then training the pre-processed training dataset under
the YOLO-v3 detection framework to obtain the single character detection model. Using
the trained single character detection model to input the whole image directly to detect
and getting the single character detection result.
S3. Training and classification of single character classification model. In step 1, the
single character image will be obtained from the labelled single character, and the single
character classification model will be constructed by using convolution neural network.
Then using the single character image to train the single character classification model,
thereby obtaining the single character classification model. The trained single character
classification model is used to input single character images to get the classification
results.
S4. Extraction of layout lines. Detecting the straight-line position in the ancient book
document and extracting the different area blocks based on the content of ancient book to
obtain the position relationship between each area block.
S5. Structured output of document. The digitized ancient book document content is
output through the results of single word detection and classification and the position
relationship between each block obtained in step 4.
Preferably, the ancient books collected in step 1 include simple layout picture TKH and
complex layout images MTH1000 and MTH1200.
Preferably, the content of the single character annotation in step 1 includes the position of
the single character and the classification category corresponding to the single character;
the text line annotation is to annotate the coordinates and the corresponding sequence
content of the text line from right to left and from top to bottom according to the reading
order of ancient books.
Preferably, the data pre-processing in step 3 includes adaptive threshold binarization of
the image data in step Sl, as well as Gaussian noise addition and random white filling or
cutting off some pixel regions.
Preferably, in step S3, according to the morphological expansion corrosion method and
combined with projection method, the invention can extract the straight lines of ancient
book document layout so as to get the position relationship between each block.
Preferably, in step 5, the characters under the double columns are sorted first according to
the coordinates of the character detection and the position extracted from the layout and
then are output.
The invention discloses the following technical effects:
The invention solves the problems of single character detection in ancient writings with
complex layout and dense documents and the existence of stain interference in large
document background.
It can identify the content of ancient books simply and efficiently and combine modem
computer information technology with traditional humanistic culture subtly, which plays
an important role in digital heritage protection, information discovery, paper document
transcription and so on.
In order to explain the embodiments of the present invention or the technical scheme in
the prior art more clearly, the figures needed in the embodiments will be briefly
introduced below. Obviously, the figures in the following description are only some
embodiments of the present invention, and for ordinary technicians in the field, other
figures can be obtained according to these figures without paying creative labour.
Figure 1 is the flow chart of the digitization method of the ancient book documents in the
invention.
Figure 2 is an introduction to dataset sampling used in the invention.
Figure 3 is a schematic diagram of the single character classification model of the
invention.
Figure 4 is an example schematic diagram of the detection result of the invention.
Figure 5 is an example schematic diagram of a layout extraction result of the present
invention;
Figure 6 is an example schematic diagram of a structured output result of the present
invention;
Figure 7 is an example schematic diagram of the final result obtained by digitizing
writings in antiquity in the present invention.
Figure 8 is a partial enlarged view of the picture labelled C in Figure 2.
The technical scheme in the embodiments of the present invention will be described
clearly and completely with reference to the figures in the embodiments of the present
invention. Obviously, the described embodiments are only parts of the embodiments of
the present invention, not all of them. Based on the embodiments of the present
invention, all other embodiments obtained by ordinary technicians in the field without
creative labour should belong to the protection scope of the present invention.
In order to make the above objects, features and advantages of the present invention more
obvious and easier to understand, the present invention will be further explained in detail
with reference to the figures and specific embodiments.
As shown in Figs. 1-8, the invention provides a method for digitizing writings in
antiquity and documents, and the specific contents are as follows.
Fig. 1 is the flow chart of the digitization method for ancient books. First of all, obtaining
the ancient book dataset to be digitized. Wherein, the ancient book dataset of the
embodiment is composed of images with simple and complex layouts, named TKH,
MTH1000 and MTH1200 respectively. With a total of 3200 image data, there are 1000,
1000, and 1200 images in turn. Then, the 3200 images data are annotated at the space
level, including text line level and single character level based on the reading order. The
images sampled from the ancient book dataset are shown in Fig. 2. Fig. 8 is an enlarged
view of the picture labelled C in Fig. 2. Characters are divided into common characters
and rare characters. Wherein, the frequency of rare characters is low, and only some
common characters reach the highest frequency. The largest single character in a dataset
has a category of 1000 images. MTH1200 has the largest categories, while TKH has the
least. The specific statistics are shown in Table 1.
Table 1. Statistics of distribution of ancient datasets
TKH MTH1000 MTH1200 Total pages 1000 1000 1200 Total text lines 23468 27559 21416 Total characters 323501 420548 337613 Categories of character 1487 5341 5292 Proportion of double column text 0 9.0% 27% Training single character detection model:
Randomly dividing all 3200 images in the ancient book datasets into training dataset and
test dataset according to the ratio of 4:1, that is, there are 2560 images in training dataset
and 640 images in test dataset. Based on the YOLO-v3 detection model, the detection
results are analysed by comparing the full input method with the slice input method. In
the training process, all 2560 images of the training dataset are scaled to a fixed size of
2048*2048, and then the anchor size is set by K-means clustering method. After training the single character detection model by using the image data in the training dataset, using
640 images in the test dataset to test the trained single character detection model, and the
test results as shown in Table 2. As can be seen from the Table 2, the slice input can
reduce the number of text box in a single image, and significantly improve the index
under high IoU. As a data pre-processing operation, slice input has significant and
general effect in facing the detection of dense objects and high-resolution images. The
single character detection result of the embodiment is shown in Fig. 4.
Table 2. Comparison test results of single character detection
IoU=0.5 IoU=0.6 IoU=0.7 IoU=0.8 Full input 98.32% 97.36% 93.55% 73.28% Slice input 99.22% 98.61% 96.40% 86.66% Training single character classification network model:
After data pre-processing and data enhancement by rotation transformation, the single
character classification network model is shown in Fig. 3. It specifically includes
convolution layer (with convolution kernel size of 3*3, input channel number of 1, output
channel number of 32), regularization layer + Relu activation layer + pooling layer (with
pooling kernel size of 2*2), convolution layer (with convolution kernel size of 3*3, input
channel number of 32, output channel number of 64); regularization layer + Relu
activation layer + pooling layer (with pooling kernel size of 2*2); convolution layer (with
convolution kernel size of 3*3, input channel number of 64, output channel number of
128); regularization layer + Relu activation layer + pooling layer (with pooling kernel
size of 2*2); convolution layer (with convolution kernel size of 3*3, input channel
number of 128, output channel number of 256), regularization layer + Relu activation
layer + pooling layer (with pooling kernel size of 2*2); full connection layer (with 512 output nodes); regularization layer + Relu activation layer + dropout layer (with dropout ratio of 0.3 to prevent over fitting), full connection layer(with 512 input nodes and a single word category as the output node). The accuracy of Top Iand Top-5 in the training of single character classification network is 97.111% and 98.87%, respectively.
The data pre-processing operation includes adaptive threshold binarization of the image
data in step 1, as well as adding Gaussian noise, random whitening, or cutting off part of
the pixel area. Adaptive threshold binarization of the image data can avoid the
interference caused by different image backgrounds. Because binarization often
introduces noise, adding Gaussian noise can increase the generalization ability of the
model. Because the single character detection model cannot guarantee the single
character can be regressed particularly accurate, random whitening can improve the
robustness of the single character classification network model.
Extraction of layout straight line:
Through image processing method, combined with projection method to detect the
location of the straight line in the document, the ancient documents will be extracted
from different areas of the document, and finally get the location of each area of the
block relationship, the effect is shown in Fig.5.
Structured output of documents
The structured output of the ancient documents needs to restore the position of the text
and the content of the document. In particular, a technical problem that needs to be
solved for the structured output of antique documents is how to solve the problem of
double-column annotations in documents. Solving this problem requires outputting the
single column in top-to-bottom order, and then outputting the contents of the double column in right-to-left order. The algorithm shown in the following table pseudocode is designed to solve this problem.
Algorithm- structured output post-processing algorithm
Input- the output result R identified by each detection frame
Output-sorted identification output result 0
1. Sorting the recognition output R of each layout by word width.
2.For i in R.
3.Adding box i to the set A.
4.Forj in R except i.
5. If the left edge of the current box j is close to the left edge of set A, or the right edge of
the box is close, adding box j to the same set A.
6. Updating the left and right edges of the set A.
7. Repeating the operation recursively for the set obtained above, until there are no two
columns in the same set
The final result obtained by entering an image of an ancient document and going through
the document digitization method is shown in Fig. 6.
By analyzing the shortcomings of traditional methods and deep learning methods, this
invention puts forward some new ideas for the digitization of ancient documents, mainly
including the use of sliding window method to improve the accuracy of text detection,
and through the morphological method to get the results of the page extraction faster. As
a result, through the designed recognition network and data enhancement techniques, the
structured output of the double-column text content has the advantages of simple
implementation, high recognition accuracy and recognition speed.
The description of the invention, it needs to be understood that the orientation or position
relationship indicated by the terms "longitudinal", "transverse", "upper", "lower",
"front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inside" and
"outside" are based on the orientation or position relationship shown in the attached
figures, which is only for the convenience of describing the invention, instead of
indicating or implying that the device or element in question must have a specific
orientation, be constructed and operated in a specific orientation, and therefore it cannot
be understood as a limitation of the invention.
The above embodiments only describe the preferred mode of the invention, but do not
limit the scope of the invention. On the premise of not departing from the design spirit of
the invention, various modifications and improvements made by ordinary technicians in
the field to the technical scheme of the invention shall fall within the protection scope
determined by the claims of the invention.
Claims (6)
1. A method for digitizing writings in antiquity is characterized by comprising following
steps.
Si. Data acquisition. Collecting the image data of ancient books and labelling the image
data with single characters and text lines at the space level to obtain the training dataset.
S2. Training and detection of single character detection model. Pre-processing the
training dataset. Based on the general target detection framework YOLO-v3, setting
anchors with different scales and then training the pre-processed training dataset under
the YOLO-v3 detection framework to obtain the single character detection model. Using
the trained single character detection model to input the whole image directly to detect
and getting the single character detection result.
S3. Training and classification of single character classification model. In step 1, the
single character image will be obtained from the labelled single character, and the single
character classification model will be constructed by using convolution neural network.
Then using the single character image to train the single character classification model,
thereby obtaining the single character classification model. The trained single character
classification model is used to input single character images to get the classification
results.
S4. Extraction of layout lines. Detecting the straight-line position in the ancient book
document and extracting the different area blocks based on the content of ancient book to
obtain the position relationship between each area block.
S5. Structured output of document
2. The method for digitizing writings in antiquity according to claim 1 is characterized in
that the ancient books collected in step 1 include simple layout picture TKH and complex
layout images MTH1000 and MTH1200.
3. The method for digitizing writings in antiquity according to claim 1 is characterized in
that the content of the single character annotation in step 1 includes the position of the
single character and the classification category corresponding to the single character; the
text line annotation is to annotate the coordinates and the corresponding sequence content
of the text line from right to left and from top to bottom according to the reading order of
ancient books.
4. The method for digitizing writings in antiquity according to claim 1 is characterized in
that the data pre-processing in step 3 includes adaptive threshold binarization of the
image data in step 1, as well as Gaussian noise addition and random white filling or
cutting off some pixel regions.
5. The method for digitizing writings in antiquity according to claim 1 is characterized in
that in step S3, according to the morphological expansion corrosion method and
combined with projection method, the invention can extract the straight lines of ancient
book document layout so as to get the position relationship between each block.
6. The method for digitizing writings in antiquity according to claim 1 is characterized in
that in step 5, the digitized ancient book document content is output through the results of
single word detection and classification and the position relationship between each block
obtained in step 4.
-1/4- 09 Nov 2020 2020103315
Figure 1 The flow chart of the digitization method of the ancient book documents in the
invention
Figure 2 An introduction to dataset sampling used in the invention
-2/4- 09 Nov 2020 2020103315
Figure 3 A schematic diagram of the single character classification model of the
invention
Figure 4 An example schematic diagram of the detection result of the invention
-3/4- 09 Nov 2020 2020103315
Figure 5 An example schematic diagram of a layout extraction result of the present
invention
Figure 6 An example schematic diagram of a structured output result of the present
invention
-4/4- 09 Nov 2020 2020103315
Figure 7 An example schematic diagram of the final result obtained by digitizing writings
in antiquity in the present invention
Figure 8 A partial enlarged view of the picture labelled C in Figure 2
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| AU2020103315A AU2020103315A4 (en) | 2020-11-09 | 2020-11-09 | A method for digitizing writings in antiquity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| AU2020103315A AU2020103315A4 (en) | 2020-11-09 | 2020-11-09 | A method for digitizing writings in antiquity |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| AU2020103315A4 true AU2020103315A4 (en) | 2021-01-14 |
Family
ID=74103433
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| AU2020103315A Active AU2020103315A4 (en) | 2020-11-09 | 2020-11-09 | A method for digitizing writings in antiquity |
Country Status (1)
| Country | Link |
|---|---|
| AU (1) | AU2020103315A4 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114120333A (en) * | 2021-11-29 | 2022-03-01 | 武汉大学 | Natural scene ancient Chinese character recognition method and system based on deep learning |
| CN114359894A (en) * | 2022-01-13 | 2022-04-15 | 浙大城市学院 | Buddhist image cultural relic three-dimensional model identification and classification method |
-
2020
- 2020-11-09 AU AU2020103315A patent/AU2020103315A4/en active Active
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114120333A (en) * | 2021-11-29 | 2022-03-01 | 武汉大学 | Natural scene ancient Chinese character recognition method and system based on deep learning |
| CN114120333B (en) * | 2021-11-29 | 2024-08-23 | 武汉大学 | Deep learning-based natural scene ancient Chinese character recognition method and system |
| CN114359894A (en) * | 2022-01-13 | 2022-04-15 | 浙大城市学院 | Buddhist image cultural relic three-dimensional model identification and classification method |
| CN114359894B (en) * | 2022-01-13 | 2024-04-30 | 浙大城市学院 | Buddhism image cultural relic three-dimensional model identification and classification method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101719142B (en) | Method for detecting picture characters by sparse representation based on classifying dictionary | |
| Arif et al. | Table detection in document images using foreground and background features | |
| CN111507351B (en) | Ancient book document digitizing method | |
| CN112016481B (en) | OCR-based financial statement information detection and recognition method | |
| Clausner et al. | Icfhr 2018 competition on recognition of historical arabic scientific manuscripts–rasm2018 | |
| EP1894144A2 (en) | Grammatical parsing of document visual structures | |
| Chamchong et al. | Character segmentation from ancient palm leaf manuscripts in Thailand | |
| AU2020103315A4 (en) | A method for digitizing writings in antiquity | |
| CN110826393B (en) | Automatic extraction method of drilling histogram information | |
| Pantke et al. | An historical handwritten arabic dataset for segmentation-free word spotting-hadara80p | |
| US7729541B2 (en) | Comparative and analytic apparatus method for converting two-dimensional bit map data into three-dimensional data | |
| CN115311666A (en) | Image-text recognition method and device, computer equipment and storage medium | |
| CN112036330A (en) | Text recognition method, text recognition device and readable storage medium | |
| CN115019310B (en) | Image-text identification method and equipment | |
| Lin et al. | Multilingual corpus construction based on printed and handwritten character separation | |
| Azmi et al. | Digital paleography: Using the digital representation of Jawi manuscripts to support paleographic analysis | |
| Xin et al. | Comic text detection and recognition based on deep learning | |
| AU2021104475A4 (en) | Methods of digitizing ancient documents | |
| Guan et al. | An open dataset for the evolution of oracle bone characters: EVOBC | |
| Kleber et al. | Matching table structures of historical register books using association graphs | |
| CN115393865A (en) | Character retrieval method, character retrieval equipment and computer-readable storage medium | |
| Samonte et al. | Senior citizen social pension management system using optical character recognition | |
| Omar et al. | Toward a Robust Segmentation Module Based on Deep Learning Approaches Resolving Historical Cursive Fonts Challenges | |
| Graef et al. | A novel hybrid optical character recognition approach for digitizing text in forms | |
| Tamatjita et al. | A Lightweight Chinese Character Recognition Model for Elementary Level Hanzi Learning Application |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FGI | Letters patent sealed or granted (innovation patent) |