CN111915540B - Rubbing oracle character image augmentation method, rubbing oracle character image augmentation system, computer equipment and medium - Google Patents
Rubbing oracle character image augmentation method, rubbing oracle character image augmentation system, computer equipment and medium Download PDFInfo
- Publication number
- CN111915540B CN111915540B CN202010553556.9A CN202010553556A CN111915540B CN 111915540 B CN111915540 B CN 111915540B CN 202010553556 A CN202010553556 A CN 202010553556A CN 111915540 B CN111915540 B CN 111915540B
- Authority
- CN
- China
- Prior art keywords
- oracle
- image
- character
- rubbing
- character image
- 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
- 230000003416 augmentation Effects 0.000 title claims abstract description 78
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000003190 augmentative effect Effects 0.000 claims abstract description 35
- 230000005012 migration Effects 0.000 claims abstract description 31
- 238000013508 migration Methods 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 22
- 210000000988 bone and bone Anatomy 0.000 claims abstract description 19
- 238000009826 distribution Methods 0.000 claims abstract description 18
- 230000000877 morphologic effect Effects 0.000 claims abstract description 17
- 238000013507 mapping Methods 0.000 claims abstract description 9
- 230000006870 function Effects 0.000 claims description 16
- 230000009466 transformation Effects 0.000 claims description 14
- 238000005260 corrosion Methods 0.000 claims description 13
- 230000007797 corrosion Effects 0.000 claims description 13
- 230000000694 effects Effects 0.000 claims description 13
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 125000004122 cyclic group Chemical group 0.000 claims description 12
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 230000008719 thickening Effects 0.000 claims description 5
- 238000009827 uniform distribution Methods 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000005530 etching Methods 0.000 claims 1
- 230000008961 swelling Effects 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 6
- 238000004590 computer program Methods 0.000 description 5
- 238000011160 research Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000021615 conjugation Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000012634 fragment Substances 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000008485 antagonism Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000009313 farming 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
- 239000002689 soil Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a rubbing oracle character image augmentation method, a rubbing oracle character image augmentation system, computer equipment and a rubbing oracle character image augmentation medium, wherein the rubbing oracle character image augmentation method comprises the following steps: inputting the character image set of the copying oracle character into a font augmentation module for random morphological processing to obtain the character image set of the copying oracle; constructing a style migration network to learn a mapping function between the distribution of the augmented copy oracle character image set to the distribution of the rubbing oracle character image set; inputting the amplified character image set of the oracle bone script into a style migration network for processing to obtain an amplified data set of the oracle bone script image; mixing the rubbing oracle character image set with the rubbing oracle image augmentation data set, and training the rubbing oracle character recognition network. The invention can obtain the augmentation data set with sufficient total number and balanced category so as to solve the problem of lack of training data in the character recognition task of the rubbing oracle, and the accuracy of the recognition model can be improved by training by using the augmentation data set.
Description
Technical Field
The invention relates to a rubbing oracle character image augmentation method, a rubbing oracle character image augmentation system, computer equipment and a storage medium, and belongs to the field of image processing and artificial intelligence.
Background
The oracle is the earliest literal system in China and is one of the earliest pictographs in the world, and is widely used for the preemption of various events in the commodity generation, including war, farming, medical treatment and other activities, and is valuable data for the current person to study ancient histories and cultures. Since the development of oracle, a historian has achieved a number of efforts to study oracle, including conjugation of oracle fragments, oracle word study, and the like. However, these works are seriously dependent on the expertise of the historian, and in front of massive historian, the expert needs to spend a lot of time to search and understand manually, which consumes a lot of effort and time and reduces the research efficiency. The development of random computer technology, how to utilize the automation technology to accelerate the oracle research so as to better inherit the culture and develop the culture treasures, has great practical significance.
The first stage of the oracle research automation is to convert oracle bone stock into a digital image by means of photographing/scanning, the second stage is to identify oracle bone in the image into a character coding database which can be stored and searched by a computer, and the third stage is to use the database to serve downstream tasks, including automatic conjugation of oracle bone fragments, oracle bone stock search and the like. At present, the automatization of oracle research is still in the second stage, and although the existing work can achieve better recognition effect on the manually copied sample, the recognition effect on the real rubbing picture is very poor, wherein the biggest obstacle is the lack of rubbing oracle character training data. The problem of lack of data is reflected in two planes: (1) The total sample amount is limited by the number of the unearthed cultural relics, because the rubbing oracle characters are taken from the real oracle pieces, the number of the rubbing oracle characters is limited on an objective level; (2) The data has unbalanced categories, similar to modern Chinese, and characters used by low frequency/high frequency exist in oracle, which leads to low/high character soil yield of the corresponding category, and taking the database adopted by the invention as an example, the category with the least sample size only comprises 1 sample, and the highest category comprises tens of thousands of samples. Because of the problems, the recognition model is difficult to learn effective recognition features from the existing real samples, particularly for the category lacking training samples, the existing method can hardly recognize the characteristics, and a large amount of noise interference such as character incomplete, breakage and the like exists in the rubbing, so that the model needs more data driving to learn the effective recognition features.
Disclosure of Invention
In view of the above, the invention provides a rubbing oracle character image augmentation method, a rubbing oracle character image augmentation system, a rubbing oracle character image augmentation computer system, a rubbing oracle character image storage medium, wherein the rubbing oracle character image is different in existing copying oracle character image conversion form and lifelike in effect, the generated rubbing oracle character image is used for augmenting the existing data set, the total number of the rubbing oracle character image is sufficient, the category of the rubbing oracle character image is balanced, the problem of lack of training data in a rubbing oracle character recognition task is solved, and the accuracy of a rubbing oracle character recognition network can be improved by training through the augmentation data set.
The first aim of the invention is to provide a rubbing oracle character image augmentation method.
The second object of the invention is to provide a rubbing oracle character image augmentation system.
A third object of the present invention is to provide a computer device.
A fourth object of the present invention is to provide a storage medium.
The first object of the present invention can be achieved by adopting the following technical scheme:
a method for augmenting an image of a character of an oracle of a rubbing, the method comprising:
inputting the character image set of the copying oracle character into a font augmentation module for random morphological processing to obtain the character image set of the copying oracle;
constructing a style migration network to learn a mapping function between the distribution of the augmented copy oracle character image set to the distribution of the rubbing oracle character image set;
inputting the amplified character image set of the oracle bone script into a style migration network for processing to obtain an amplified data set of the oracle bone script image;
mixing the rubbing oracle character image set with the rubbing oracle image augmentation data set, and training the rubbing oracle character recognition network.
Further, the step of inputting the character image set of the copying oracle into the font augmentation module for random morphological processing to obtain the character image set of the copying oracle comprises the following steps:
the method comprises the steps of defining endpoints, inflection points, intersection points and bifurcation points on the morphology of the oracle characters as character key points, detecting the key points in a copying oracle character image set by using a target detection algorithm, and representing the key points for a specific copying oracle character image as follows:
wherein ,representing the position coordinates of the ith key point in the image, and N represents the total number of character key points in the image;
random dithering is applied to key points P of character images of the copying oracle, so that key points of the character pattern-enhanced copying images are obtained, and the key points are as follows:
wherein , and />Sampling and self-uniform distribution;
decomposing the key points P and P' into a plurality of triangular areas by using a triangulation algorithm, carrying out affine transformation and image interpolation on each triangular area, and further converting the character image of the copying oracle into an augmented gesture copying image;
and carrying out random morphological corrosion/expansion on the gesture-enhanced copy image, and adding a stroke thinning/thickening effect to the character to obtain a font-enhanced copy image, thereby obtaining an enhanced copy oracle character image set.
Further, the affine transformation operates as follows:
wherein (x, y) and (x ', y') are pixel coordinates before and after transformation, respectively, transformation parameter a i,j Is solved by the simultaneous equations of the coordinates of the vertices of the triangles in P and P'.
Further, the corrosion/expansion operation is as follows:
wherein I and I' are an input image and an output image, respectively, W is a rectangular window of size k×k, for any 0.ltoreq.h.ltoreq.k, 0.ltoreq.w.ltoreq.k, W (W, h) =1, the parameter K being used to control the extent of corrosion/expansion.
Further, the style migration network is a style migration network based on a recurring antagonism generation network, which includes an image generator and an image discriminator.
Further, the number of layers of the residual convolution module in the image generator is set to four.
Further, the image generator is trained with a weighted cyclic loss function calculated as follows:
L cyc (G)＝W*‖F(G(x))-x‖ 1
wherein x is an input image, G is an image generator, F is an image discriminator, and W is a weight matrix, as follows:
wherein ,Sfg ,S bg The area of the stroke area and the area of the background area, respectively.
The second object of the invention can be achieved by adopting the following technical scheme:
a system for augmenting an image of an oracle character of a rubbing, the system comprising:
the first processing unit is used for inputting the character image set of the copying oracle into the font augmentation module for random morphological processing to obtain the character image set of the copying oracle;
the learning unit is used for constructing a style migration network so as to learn a mapping function between the distribution of the augmented copy oracle character image set and the distribution of the rubbing oracle character image set;
the second processing unit is used for inputting the amplified character image set of the oracle character into a style migration network for processing to obtain an amplified data set of the oracle image of the rubbing;
the training unit is used for mixing the rubbing oracle character image set with the rubbing oracle image augmentation data set and training the rubbing oracle character recognition network.
The third object of the present invention can be achieved by adopting the following technical scheme:
the computer equipment comprises a processor and a memory for storing a program executable by the processor, wherein the processor realizes the rubbing oracle character image augmentation method when executing the program stored by the memory.
The fourth object of the present invention can be achieved by adopting the following technical scheme:
a storage medium storing a program which, when executed by a processor, implements the rubbing oracle character image augmentation method described above.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention provides a font augmentation module for augmenting the fonts of the oracle characters, which can generate the augmented copy oracle character images with different morphologies from the copy oracle character images with single morphologies so as to increase the font richness of the oracle characters, simulate the effect of different morphologies in the oracle characters of a real rubbing, increase the diversity of the augmented copy oracle character images, convert the augmented copy oracle character images into augmented rubbing oracle character images by using a style migration network, and realize the vivid rubbing character effect.
2. The style migration network is based on the cyclic countermeasure generation network, wherein the image generator of the cyclic countermeasure generation network adopts a weighted cyclic loss function, so that the semantic importance of characters and the background can be better balanced in the training stage, and the generated image can better keep the oracle character structure.
3. The invention can fully utilize the existing character images of the copying oracle, expand the number and the category of the oracle data set of the rubbing, and improve the performance of the oracle character recognition network of the rubbing under the condition of not increasing extra labor cost.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to the structures shown in these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of the method for augmenting the character image of the bone and oracle of the rubbing according to embodiment 1 of the present invention.
Fig. 2 is a schematic diagram of an augmentation method of a character image of a rubbing oracle according to embodiment 1 of the present invention.
Fig. 3 is a schematic diagram of a character gesture augmentation sub-module of embodiment 1 of the present invention.
Fig. 4 is a schematic diagram showing the comparison between the generated and enlarged oracle character image of the rubbing and the actual oracle character image of the rubbing in embodiment 1 of the present invention.
Fig. 5 is a schematic diagram of an augmentation effect of a character image of a rubbing oracle according to embodiment 1 of the present invention.
Fig. 6 is a block diagram of the system for enhancing the character image of the bone and oracle of the rubbing according to embodiment 2 of the present invention.
Fig. 7 is a block diagram showing the structure of a computer device according to embodiment 3 of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments, and all other embodiments obtained by those skilled in the art without making any inventive effort based on the embodiments of the present invention are within the scope of protection of the present invention.
Example 1:
as shown in fig. 1 and 2, the present embodiment provides a rubbing oracle character image augmentation method, which can be implemented by using Python programming language and PyTorch deep learning framework, and includes the following steps:
s101, inputting the character image set S of the copying oracle characters into a font augmentation module for random morphological processing to obtain an augmented character image set S' of the copying oracle characters.
The character image set S of the copying oracle is an existing character image set of the copying oracle, the font augmentation module is realized by adopting a font augmentation algorithm and comprises a character gesture augmentation sub-module and a stroke weight augmentation sub-module, wherein the character gesture augmentation sub-module is realized by a key point detection algorithm and an image transformation algorithm, the stroke weight augmentation sub-module is realized by morphological corrosion expansion operation, and the input of the font augmentation module is a gray image matrix with the size of (H, W).
1) In this embodiment, the endpoints, inflection points, intersection points and bifurcation points on the morphology of the oracle characters are defined as character key points, the character gesture augmentation submodule is shown in fig. 3, the target detection algorithm, namely, the fast-RCNN, is used as a key point detection algorithm, the key points in the copying oracle character image set S are detected through the fast-RCNN, the coordinates of the key points are returned, and for a specific copying oracle character image, the key points are expressed as follows:
wherein ,representing the position coordinates of the ith keypoint in the image, N representing the total number of character keypoints in the image.
2) The character gesture augmentation sub-module applies random dithering on the key points P of the character image of the copying oracle to obtain the key points of the character gesture augmentation copying image, and the key points are as follows:
wherein , and />The samples are taken from a Uniform distribution R, in this example R-uniformity (-10, 10), representing random variables subject to Uniform distribution.
3) The character gesture augmentation sub-module uses a triangulation algorithm built in a scientific computing library scipy to decompose key points P and P' into a plurality of triangle areas, carries out affine transformation and image interpolation on each triangle area, and further transforms a character image of the copying oracle into a gesture augmentation copy image.
Further, the affine transformation operates as follows:
wherein (x, y) and (x ', y') are pixel coordinates before and after transformation, respectively, transformation parameter a i,j Is solved by the simultaneous equations of the coordinates of the vertices of the triangles in P and P'.
4) The stroke weight augmentation sub-module performs random morphological corrosion/expansion on the gesture augmentation copy image, adds stroke thinning/thickening effects to the characters, and obtains a font augmentation copy image, thereby obtaining an augmentation copy oracle character image set S' with more abundant fonts.
Further, the corrosion/expansion operation is as follows:
wherein I and I' are an input image and an output image, respectively, W is a rectangular window of size k×k, for any 0.ltoreq.h.ltoreq.k, 0.ltoreq.w.ltoreq.k, W (W, h) =1, the parameter K being used to control the extent of corrosion/expansion.
In the image augmentation process of the embodiment, the probability of using the expansion operation is 90%, the probability of using the corrosion operation is 10%, and the parameter K is also generated in a random manner, so that the effect of thinning/thickening random strokes is obtained, and specific numerical values are obtained according to the following empirical formula:
wherein the parameter K representing the corrosion parameter expansion and corrosion operation is derived from a random variable K, respectively d and Kc And (3) obtaining the N by sampling, wherein N is the number of character key points.
S102, constructing a style migration network to learn a mapping function F between the distribution of the augmented copy oracle character image set S' and the distribution of the rubbing oracle character image set T.
The style migration network is a style migration network based on a cyclic countermeasure generation network (Cycle-GAN), the cyclic countermeasure generation network invented by the university of california berkeley division artificial intelligence laboratory has been widely used in the field of image style migration, the cyclic countermeasure generation network is composed of two groups of symmetrical image generators and image discriminators, the embodiment builds the style migration network by referring to the structure of the cyclic countermeasure generation network, preferably, the number of residual convolution module layers in the image generator of the embodiment is set to be four, and the number of layers can better extract high-dimensional semantic information of images, and meanwhile, can also retain structural information of oracle characters.
The cyclic loss function is used in Cycle-GAN to guide the network to pay attention to semantic information in the input image, but the rubbing oracle image has more noise, if characters and noise are treated indiscriminately, the generated characters may not be recognized, in this embodiment, the image generator is trained by adopting the weighted cyclic loss function, so that the image after style migration still can keep a clearer character structure, and the weighted cyclic loss function is calculated as follows:
L cyc (G)＝W*‖F(G(x))-x‖ 1
wherein x is an input image, G is an image generator, F is an image discriminator, and W is a weight matrix, as follows:
wherein ,Sfg ,S bg The area of the stroke area and the area of the background area, respectively.
S103, inputting the augmented copy oracle character image set S 'into a style migration network for processing to obtain a rubbing oracle image augmented data set T'.
And (3) connecting the font augmentation module of the step (S101) and the style migration network of the step (S102) in series, regarding the category of less than 1000 samples in the original rubbing oracle character image set (T) as the category needing image augmentation, sampling from a copy character image library in the category needing image augmentation to obtain a copy oracle character image set (S), processing the copy oracle character image set (S ') by the input font augmentation module to obtain an augmented copy oracle character image set (S'), and then inputting the obtained augmented copy oracle character image set (S ') into the style migration network to process to obtain a rubbing oracle image augmentation data set (T') with sufficient quantity and rich fonts, wherein the comparison between the generated augmented oracle character image and the real rubbing oracle character image is shown in fig. 4.
S104, mixing the rubbing oracle character image set T with the rubbing oracle image augmentation data set T', and training the rubbing oracle character recognition network.
The rubbing oracle character recognition network is the rubbing oracle character image classification network, the rubbing oracle character image set T is mixed with the rubbing oracle image augmentation data set T', the rubbing oracle character recognition network is trained, and the purpose of improving the accuracy of the rubbing oracle character recognition network is achieved.
As shown in fig. 5, the "no added augmentation data" indicates the accuracy of training the rubbing oracle character recognition network by using only the real rubbing oracle character rubbing data set, and the "added augmentation data" indicates the accuracy of training the rubbing oracle character recognition network after adding the augmentation oracle character image generated in the embodiment, so that the recognition accuracy of the rubbing oracle character recognition network is obviously improved after adding the augmentation oracle character image.
It should be noted that although the method operations of the above embodiments are depicted in the drawings in a particular order, this does not require or imply that the operations must be performed in that particular order or that all illustrated operations be performed in order to achieve desirable results. Rather, the depicted steps may change the order of execution. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step to perform, and/or one step decomposed into multiple steps to perform.
Example 2:
as shown in fig. 6, the present embodiment provides a rubbing oracle character image augmentation system, which includes a first processing unit 601, a construction learning unit 602, and a second processing unit 603, where specific functions of each module are as follows:
the first processing unit 601 is configured to input the set of character images of the facsimile oracle into the font augmentation module for performing random morphological processing, so as to obtain the set of character images of the facsimile oracle.
The learning unit 602 is configured to construct a style migration network to learn a mapping function between the distribution of the augmented copy oracle character image set to the distribution of the rubbing oracle character image set.
The second processing unit 603 is configured to input the augmented copy oracle character image set into a style migration network for processing, so as to obtain a rubbing oracle image augmented data set.
The training unit 604 is configured to mix the rubbing oracle character image set with the rubbing oracle image augmentation data set, and train the rubbing oracle character recognition network.
Specific implementation of each module in this embodiment may be referred to embodiment 1 above, and will not be described in detail herein; it should be noted that, in the system provided in this embodiment, only the division of the above functional modules is used as an example, in practical application, the above functional allocation may be performed by different functional modules according to needs, that is, the internal structure is divided into different functional modules to perform all or part of the functions described above.
Example 3:
the present embodiment provides a computer device, which may be a computer, as shown in fig. 7, and includes a processor 702, a memory, an input device 703, a display 704 and a network interface 705 connected by a system bus 701, where the processor is configured to provide computing and control capabilities, the memory includes a nonvolatile storage medium 706 and an internal memory 707, where the nonvolatile storage medium 706 stores an operating system, a computer program and a database, and the internal memory 707 provides an environment for the operating system and the computer program in the nonvolatile storage medium, and when the processor 702 executes the computer program stored in the memory, the rubbing oracle character image augmentation method of the foregoing embodiment 1 is implemented as follows:
inputting the character image set of the copying oracle character into a font augmentation module for random morphological processing to obtain the character image set of the copying oracle;
constructing a style migration network to learn a mapping function between the distribution of the augmented copy oracle character image set to the distribution of the rubbing oracle character image set;
inputting the amplified character image set of the oracle bone script into a style migration network for processing to obtain an amplified data set of the oracle bone script image;
mixing the rubbing oracle character image set with the rubbing oracle image augmentation data set, and training the rubbing oracle character recognition network.
Example 4:
the present embodiment provides a storage medium, which is a computer readable storage medium storing a computer program, where the computer program when executed by a processor implements the rubbing oracle character image augmentation method of the embodiment 1, as follows:
inputting the character image set of the copying oracle character into a font augmentation module for random morphological processing to obtain the character image set of the copying oracle;
constructing a style migration network to learn a mapping function between the distribution of the augmented copy oracle character image set to the distribution of the rubbing oracle character image set;
inputting the amplified character image set of the oracle bone script into a style migration network for processing to obtain an amplified data set of the oracle bone script image;
mixing the rubbing oracle character image set with the rubbing oracle image augmentation data set, and training the rubbing oracle character recognition network.
The storage medium in this embodiment may be a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a usb disk, a removable hard disk, or the like.
In summary, the invention can convert the prior rubbing oracle character images with different forms and vivid effects, expand the prior data set by using the generated rubbing oracle character images, and obtain the amplified data set with sufficient total number and balanced categories so as to solve the problem of lack of training data in the rubbing oracle character recognition task, and the accuracy of the rubbing oracle character recognition network can be improved by training by using the amplified data set.
The above-mentioned embodiments are only preferred embodiments of the present invention, but the protection scope of the present invention is not limited thereto, and any person skilled in the art can make equivalent substitutions or modifications according to the technical solution and the inventive concept of the present invention within the scope of the present invention disclosed in the present invention patent, and all those skilled in the art belong to the protection scope of the present invention.
Claims (8)
1. An augmentation method for a character image of a rubbing oracle, which is characterized by comprising the following steps:
inputting the character image set of the copying oracle character into a font augmentation module for random morphological processing to obtain the character image set of the copying oracle;
constructing a style migration network to learn a mapping function between the distribution of the augmented copy oracle character image set and the distribution of the rubbing oracle character image set, wherein the style migration network is based on a circulating countermeasure generation network, and the circulating countermeasure generation network comprises an image generator and an image discriminator;
inputting the amplified character image set of the oracle bone script into a style migration network for processing to obtain an amplified data set of the oracle bone script image;
mixing the rubbing oracle character image set with the rubbing oracle image augmentation data set, and training a rubbing oracle character recognition network;
inputting the character image set of the copying oracle into a font augmentation module for random morphological processing to obtain the character image set of the copying oracle, which specifically comprises the following steps:
the method comprises the steps of defining endpoints, inflection points, intersection points and bifurcation points on the morphology of the oracle characters as character key points, detecting the key points in a copying oracle character image set by using a target detection algorithm, and representing the key points for a specific copying oracle character image as follows:
wherein ,representing the position coordinates of the ith key point in the image, and N represents the total number of character key points in the image;
random dithering is applied to key points P of character images of the copying oracle, so that key points of the character pattern-enhanced copying images are obtained, and the key points are as follows:
wherein , and />Sampling and self-uniform distribution;
decomposing the key points P and P' into a plurality of triangular areas by using a triangulation algorithm, carrying out affine transformation and image interpolation on each triangular area, and further converting the character image of the copying oracle into an augmented gesture copying image;
and carrying out random morphological corrosion/expansion on the gesture-enhanced copy image, and adding a stroke thinning/thickening effect to the character to obtain a font-enhanced copy image, thereby obtaining an enhanced copy oracle character image set.
2. The method for augmenting an image of an oracle character of a rubbing according to claim 1, wherein the affine transformation is performed as follows:
wherein (x, y) and (x ', y') are pixel coordinates before and after transformation, respectively, transformation parameter a i,j By co-ordinating the coordinates of the vertices of triangles in P and PThe equation is solved.
3. The method for augmenting an image of a character of the bone and oracle of a rubbing according to claim 1, characterized in that said etching/swelling is performed as follows:
wherein I and I' are an input image and an output image, respectively, W is a rectangular window of size k×k, for any 0.ltoreq.h.ltoreq.k, 0.ltoreq.w.ltoreq.k, W (W, h) =1, the parameter K being used to control the extent of corrosion/expansion.
4. A method of augmenting an image of a character of the bone-script of a rubbing as claimed in any one of claims 1-3, characterized in that the number of layers of the residual convolution module in the image generator is set to four.
5. A method of augmenting an image of a character of the bone-script of a rubbing as claimed in any one of claims 1-3, wherein the image generator is trained with a weighted cyclic loss function calculated as follows:
L cyc (G)＝W*‖F(G(x))-x‖ 1
wherein x is an input image, G is an image generator, F is an image discriminator, and W is a weight matrix, as follows:
wherein ,Sfg ,S bg The area of the stroke area and the area of the background area, respectively.
6. An augmentation system for an image of an oracle character of a rubbing, the system comprising:
the first processing unit is used for inputting the character image set of the copying oracle into the font augmentation module for random morphological processing to obtain the character image set of the copying oracle;
the system comprises a learning unit, a learning unit and a control unit, wherein the learning unit is used for constructing a style migration network for learning a mapping function between the distribution of the augmented copy oracle character image set and the distribution of the rubbing oracle character image set, the style migration network is based on a circulating countermeasure generation network, and the circulating countermeasure generation network comprises an image generator and an image discriminator;
the second processing unit is used for inputting the amplified character image set of the oracle character into a style migration network for processing to obtain an amplified data set of the oracle image of the rubbing;
the training unit is used for mixing the rubbing oracle character image set with the rubbing oracle image augmentation data set and training a rubbing oracle character recognition network;
inputting the character image set of the copying oracle into a font augmentation module for random morphological processing to obtain the character image set of the copying oracle, which specifically comprises the following steps:
the method comprises the steps of defining endpoints, inflection points, intersection points and bifurcation points on the morphology of the oracle characters as character key points, detecting the key points in a copying oracle character image set by using a target detection algorithm, and representing the key points for a specific copying oracle character image as follows:
wherein ,representing the position coordinates of the ith key point in the image, and N represents the total number of character key points in the image;
random dithering is applied to key points P of character images of the copying oracle, so that key points of the character pattern-enhanced copying images are obtained, and the key points are as follows:
wherein , and />Sampling and self-uniform distribution;
decomposing the key points P and P' into a plurality of triangular areas by using a triangulation algorithm, carrying out affine transformation and image interpolation on each triangular area, and further converting the character image of the copying oracle into an augmented gesture copying image;
and carrying out random morphological corrosion/expansion on the gesture-enhanced copy image, and adding a stroke thinning/thickening effect to the character to obtain a font-enhanced copy image, thereby obtaining an enhanced copy oracle character image set.
7. A computer device comprising a processor and a memory for storing a program executable by the processor, wherein the processor, when executing the program stored in the memory, implements the method for augmenting a character image of a bone inscription of a rubbing according to any one of claims 1-5.
8. A storage medium storing a program which, when executed by a processor, implements the method for augmenting an image of a character of a bone and oracle of a rubbing according to any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010553556.9A CN111915540B (en) | 2020-06-17 | 2020-06-17 | Rubbing oracle character image augmentation method, rubbing oracle character image augmentation system, computer equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010553556.9A CN111915540B (en) | 2020-06-17 | 2020-06-17 | Rubbing oracle character image augmentation method, rubbing oracle character image augmentation system, computer equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111915540A CN111915540A (en) | 2020-11-10 |
| CN111915540B true CN111915540B (en) | 2023-08-18 |
Family
ID=73237797
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010553556.9A Active CN111915540B (en) | 2020-06-17 | 2020-06-17 | Rubbing oracle character image augmentation method, rubbing oracle character image augmentation system, computer equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111915540B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112613474B (en) * | 2020-12-30 | 2022-01-18 | 珠海大横琴科技发展有限公司 | Pedestrian re-identification method and device |
| CN113033567B (en) * | 2021-03-22 | 2022-11-01 | 安阳师范学院 | Oracle bone rubbing image character training method fusing segmentation network and generation network |
| CN113111803B (en) * | 2021-04-20 | 2022-03-22 | 复旦大学 | Small sample character and hand-drawn sketch identification method and device |
| CN113362361B (en) * | 2021-07-20 | 2023-09-05 | 辽宁师范大学 | Image data set construction method for oracle text detection under morphological priori constraint |
| CN113537229A (en) * | 2021-08-27 | 2021-10-22 | 广州广电运通金融电子股份有限公司 | Bill image generation method and device, computer equipment and storage medium |
| CN114092700B (en) * | 2021-11-25 | 2022-09-20 | 吉林大学 | Ancient character recognition method based on target detection and knowledge graph |
| CN114067331B (en) * | 2021-11-25 | 2022-09-06 | 吉林大学 | Oracle-bone inscription font construction method |
| CN114782961B (en) * | 2022-03-23 | 2023-04-18 | 华南理工大学 | Character image augmentation method based on shape transformation |
| CN114495118B (en) * | 2022-04-15 | 2022-08-09 | 华南理工大学 | Personalized handwritten character generation method based on countermeasure decoupling |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101881999A (en) * | 2010-06-21 | 2010-11-10 | 安阳师范学院 | Oracle video input system and implementation method |
| CN102592149A (en) * | 2012-01-04 | 2012-07-18 | 徐州师范大学 | Computer aided periodization method of oracle bone rubbings |
| CN109871904A (en) * | 2019-03-11 | 2019-06-11 | 广东工业大学 | Inscriptions on bones or tortoise shells word identification model and training method, system, equipment, computer media |
| CN111291750A (en) * | 2020-01-21 | 2020-06-16 | 河南大学 | Oracle automatic labeling method based on spatial neighbor relation |
-
2020
- 2020-06-17 CN CN202010553556.9A patent/CN111915540B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101881999A (en) * | 2010-06-21 | 2010-11-10 | 安阳师范学院 | Oracle video input system and implementation method |
| CN102592149A (en) * | 2012-01-04 | 2012-07-18 | 徐州师范大学 | Computer aided periodization method of oracle bone rubbings |
| CN109871904A (en) * | 2019-03-11 | 2019-06-11 | 广东工业大学 | Inscriptions on bones or tortoise shells word identification model and training method, system, equipment, computer media |
| CN111291750A (en) * | 2020-01-21 | 2020-06-16 | 河南大学 | Oracle automatic labeling method based on spatial neighbor relation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111915540A (en) | 2020-11-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111915540B (en) | Rubbing oracle character image augmentation method, rubbing oracle character image augmentation system, computer equipment and medium | |
| CN110765966B (en) | One-stage automatic recognition and translation method for handwritten characters | |
| CN111898696B (en) | Pseudo tag and tag prediction model generation method, device, medium and equipment | |
| CN108170649B (en) | Chinese character library generation method and device based on DCGAN deep network | |
| Li et al. | Computer vision imaging based on artificial intelligence | |
| CN110033054B (en) | Personalized handwriting migration method and system based on collaborative stroke optimization | |
| Ding et al. | Open-vocabulary universal image segmentation with maskclip | |
| Huang et al. | Glance and focus networks for dynamic visual recognition | |
| CN112381082A (en) | Table structure reconstruction method based on deep learning | |
| CN111401156A (en) | Image identification method based on Gabor convolution neural network | |
| CN111858878B (en) | Method, system and storage medium for automatically extracting answer from natural language text | |
| CN111242840A (en) | Handwritten character generation method, apparatus, computer device and storage medium | |
| CN115713632A (en) | Feature extraction method and device based on multi-scale attention mechanism | |
| CN111597816A (en) | Self-attention named entity recognition method, device, equipment and storage medium | |
| CN115170403A (en) | Font repairing method and system based on deep meta learning and generation countermeasure network | |
| Dhawan et al. | Deep Learning Based Sugarcane Downy Mildew Disease Detection Using CNN-LSTM Ensemble Model for Severity Level Classification | |
| CN117333881A (en) | Oracle auxiliary decoding method based on conditional diffusion model | |
| CN111831805A (en) | Model creation method and device, electronic equipment and readable storage device | |
| CN111860465A (en) | Remote sensing image extraction method, device, equipment and storage medium based on super pixels | |
| CN111832610A (en) | 3D printing organization prediction method, system, medium and terminal equipment | |
| CN116860985A (en) | Natural language interaction software framework based on knowledge graph and construction method thereof | |
| CN114299252B (en) | Method and device for reconstructing universal three-dimensional model, storage medium and electronic equipment | |
| Yu et al. | Artificial intelligence in paleontology | |
| CN114116779A (en) | Deep learning-based power grid regulation and control field information retrieval method, system and medium | |
| CN113569080A (en) | Word stock processing method, device, equipment and storage medium based on artificial intelligence |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |