CN111898618A - Method, device and program storage medium for identifying ancient graphics and characters - Google Patents
Method, device and program storage medium for identifying ancient graphics and characters Download PDFInfo
- Publication number
- CN111898618A CN111898618A CN202010644573.3A CN202010644573A CN111898618A CN 111898618 A CN111898618 A CN 111898618A CN 202010644573 A CN202010644573 A CN 202010644573A CN 111898618 A CN111898618 A CN 111898618A
- Authority
- CN
- China
- Prior art keywords
- ancient
- visual
- images
- image
- inscriptions
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000012545 processing Methods 0.000 claims abstract description 19
- 230000000007 visual effect Effects 0.000 claims description 235
- 238000012163 sequencing technique Methods 0.000 claims description 16
- 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 13
- 230000009466 transformation Effects 0.000 claims description 13
- 238000013507 mapping Methods 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000013519 translation Methods 0.000 claims description 4
- 238000004220 aggregation Methods 0.000 claims description 3
- 230000002776 aggregation Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 2
- 230000002207 retinal effect Effects 0.000 claims 1
- 238000000844 transformation Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 210000001525 retina Anatomy 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000004984 smart glass Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 239000011232 storage material Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/5854—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using shape and object relationship
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Library & Information Science (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Character Discrimination (AREA)
Abstract
The invention discloses a method, a device and a program storage medium for identifying ancient graphics and characters, wherein the method comprises the following steps: extracting feature data of images to be processed belonging to different ancient inscriptions, handwriting and graphic characters from image data of the ancient inscriptions, handwriting and graphic characters of an original image; responding to data matching in a database of various ancient inscriptions, handwriting, graphic, character and image characteristic data subjected to digital processing; and determining the image data of the recognized ancient inscription, handwriting and graph characters and outputting the image data by using modern standard Chinese characters or phrases. According to the invention, the accuracy of image recognition is improved and false recognition is reduced through secondary matching.
Description
Technical Field
The present invention relates to a method and an apparatus for recognizing characters, and more particularly, to a method and an apparatus for recognizing ancient patterns and characters, and a program storage medium.
Background
Ancient writing and hand-written graph characters cannot be converted into modern standard Chinese characters through a modern dictionary because relevant characteristics such as radicals, voice, strokes and the like cannot be determined. The recognition of ancient inscriptions and handwritten pattern characters can not be realized through the characteristics, and the difficulty in practical use exists. In the prior art, a chinese patent with publication number CN104794455B discloses a dongba pictograph recognition method, which includes the steps: carrying out feature extraction on the Dongba pictograph by adopting a projection method; according to the extracted characteristics of the Dongba pictographs, the Dongba pictographs are identified by combining a similarity method and a network feedback method, and the Dongba pictographs which are simple in structural strokes, different in morphological structures and easy to identify are identified by adopting the similarity method; and identifying the Dongba pictographs with complex structural strokes and similar morphological structures by adopting a network feedback method. The method distinguishes the Dongba pictographs, and a similarity method is adopted for the simple and easily-recognized Dongba pictographs so as to realize quick recognition, but the similarity method is simple in process and is not suitable for recognition of more complex ancient inscriptions and handwritten patterns and characters, and the recognition error rate is increased.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide a method for identifying ancient image characters, which is used for quickly and accurately classifying and identifying the ancient image characters based on a similarity method so as to facilitate practical application. It is another object of the present invention to provide an apparatus and a program storage medium for implementing the method for recognizing ancient graphics and texts.
The technical scheme of the invention is as follows:
a method for recognizing ancient graphics and texts comprises the following steps:
acquiring ancient inscription and handwritten pattern character images to be inquired;
performing perspective processing on the image data stroke gaps of the ancient inscriptions and handwritten pattern characters to be inquired, and obtaining corresponding image data of a first visual characteristic;
classifying the image data of the first visual characteristic into a perspective image data set with uniform size and pixels to form second visual characteristic data;
converting the stroke number or the percentage interval range of the shadow coverage rate in the image of the second visual feature data into different classified characters for distinguishing and classifying to form a third visual feature classification and sequencing;
performing plane translation, plane rotation and scaling on the images of the third characteristic visual characteristic pair, matching the images of the reference ancient inscriptions and handwritten pattern characters in an image database of the third characteristic visual characteristic pair, determining visual characteristic distances between the images of the query ancient inscriptions and handwritten pattern characters and the images of the candidate ancient inscriptions and handwritten pattern characters, sequencing the images of the candidate ancient inscriptions and handwritten pattern characters according to the visual characteristic distances, and determining image sets of the similar ancient inscriptions and handwritten pattern characters from the images of the candidate ancient inscriptions and handwritten pattern characters according to the sequencing result;
removing the matched pairs of visual features that exceed the threshold from the plurality of sets of pairs of visual features in the fourth classification and ordering of visual feature data to form a fifth set of visual feature data.
Checking local visual feature pairs of feature images of a plurality of groups of images of similar ancient inscriptions and handwritten pattern characters of the fifth visual feature data set to obtain visual feature distances among the images, and sequencing the images of the candidate ancient inscriptions and handwritten pattern characters according to the visual feature distances; determining an image set of similar ancient inscriptions and handwritten pattern characters from images of the candidate ancient inscriptions and handwritten pattern characters according to the sequencing result to form a sixth visual characteristic data set;
removing the visual feature pairs which are mismatched beyond the threshold value from the plurality of groups of visual feature pairs in the sixth visual feature data classification and sorting to form a seventh visual feature data set;
and calculating affine mapping transformation and errors between the images of the similar ancient inscriptions and the handwritten pattern characters and the images of the query ancient inscriptions and the handwritten pattern characters according to the seventh visual characteristic data set, indicating successful matching when the calculation result is within a preset range, and outputting modern standard Chinese characters or phrases corresponding to the successfully matched ancient inscriptions and handwritten pattern character images.
Further, the step of calculating affine mapping transformation and errors between the images of the similar ancient inscriptions and handwritten pattern characters and the images of the query ancient inscriptions and handwritten pattern characters according to the seventh visual feature data set includes the steps of:
according to the seventh visual characteristic data set, affine mapping transformation between the images of the similar ancient inscriptions and the handwritten pattern characters and the images of the query ancient inscriptions and the handwritten pattern characters is calculated, the number of the inner group points is obtained according to the result of the radial mapping, after all the number of the inner group points is calculated, if the number of the inner group points is less than a certain threshold value, the matching failure of the query ancient inscriptions and the handwritten pattern character images with the similar ancient inscriptions and the handwritten pattern character images is indicated, otherwise, the matching success is indicated, and the next step is carried out;
and calculating the error of affine transformation between the query image and the similar image according to the error of affine transformation between the image of the similar ancient inscription and the handwritten pattern character and the image of the query ancient inscription and the handwritten pattern character, wherein if the error is larger than a certain threshold value, the matching failure of the image of the query ancient inscription and the handwritten pattern character with the similar ancient inscription and the handwritten pattern character is represented, and otherwise, the matching success is represented.
Further, the following steps are performed after the multiple groups of visual feature pairs in the classification and sorting of the third visual features are removed of the matched visual feature pairs exceeding the threshold before the plane translation, the plane rotation and the scaling are performed on the images of the third visual feature pairs.
Further, local visual characteristic pairs of characteristic images of a plurality of groups of similar ancient inscriptions and handwritten pattern characters of the fifth visual characteristic data set are verified in a Hough transform algorithm mode.
Further, the fifth visual feature is a global visual feature describing the full-image visual content of the image and a local visual feature describing the local visual content of the image.
Further, the global visual features include any one of bag-of-words model features, local feature aggregation descriptors and Fisher vector features, and the local visual features include any one of ORBs and fast retina key point features.
Further, the fourth visual feature comprises at least one of a scale-invariant feature, an acceleration robust feature, or any combination thereof.
An apparatus for recognizing ancient graphic and text is used for realizing a method for recognizing ancient graphic and text, and comprises:
the image input module is used for acquiring ancient inscription and handwritten pattern character images to be inquired;
the processing module is used for carrying out gap perspective processing and normalized data processing on the ancient inscription and handwritten pattern character images obtained by the image input module;
the matching module is used for matching second visual characteristic data, third visual characteristic data, fourth visual characteristic data, fifth visual characteristic data, sixth visual characteristic data and seventh visual characteristic data of the inquired ancient inscription and handwritten pattern character images with visual characteristics of similar images to form a plurality of groups of visual characteristic pairs, wherein the visual characteristics comprise local visual characteristics; the similar images are retrieved according to second visual features of the ancient inscriptions and the handwritten pattern character images to be inquired, and the second visual features comprise global visual features and/or local visual features of the images;
the removing module is used for removing the visual feature pairs which are mismatched in error from the plurality of groups of visual feature pairs and screening out proper visual feature data pairs;
the confirming module is used for confirming whether the ancient inscription and handwritten pattern character image to be inquired is successfully matched with the similar image or not according to the proper visual characteristic data pair obtained by the removing module;
and the Chinese character output module is used for outputting standard Chinese characters and related evolution annotations according to the determined similar images.
A program storage medium storing a computer program which, when executed by a processor, implements a method of identifying ancient graphic texts.
Compared with the prior art, the invention has the advantages that: through the stroke gap perspective processing of the image, the influence on image characteristic data caused by different skills, storage environments, image acquisition light, character storage materials and materials in the production process of ancient inscriptions, handwriting and graphic characters is well solved. The method can perform secondary matching on the similar images obtained by the first image retrieval, and can remove the visual feature pairs of the mismatching by using a more refined mismatching removal method than the first retrieval in the secondary matching, thereby improving the accuracy of image identification and reducing the misidentification. Under the conditions that the illumination condition changes, the mobile terminal moves rapidly to cause image blurring, the screen occupation ratio of an effective target in the query image is small and the like, whether an image matched with the query image exists or not can be accurately identified.
Drawings
FIG. 1 is a schematic block diagram of a process for identifying ancient graphic texts.
FIG. 2 is a schematic view of the structure of an apparatus for recognizing ancient graphics and texts.
Detailed Description
The present invention is further illustrated by the following examples, which are not to be construed as limiting the invention thereto.
Please refer to fig. 1, the method for identifying ancient graphics and texts according to the present embodiment includes the following steps:
101. acquiring image data of ancient inscriptions and handwritten pattern characters to be inquired, carrying out perspective processing on pen drawing gaps of the image data of the ancient inscriptions and handwritten pattern characters, and obtaining corresponding image data of a first visual characteristic.
102. Combining the obtained image data of the first visual characteristics of the ancient inscription and handwritten graphic characters to be inquired, classifying the image data into a perspective image data set with uniform size and pixels to form second visual characteristic data, and facilitating calculation processing in subsequent classification processing.
103. And according to the second visual characteristic data of the query image, calculating the stroke number and the shadow coverage rate in the image in percentage, converting the stroke number or the percentage interval range of the shadow coverage rate into different classified characters for distinguishing and classifying, and forming third visual characteristic classification and sequencing.
104. Removing the matched pairs of visual features that exceed the threshold from the plurality of sets of pairs of visual features in the third classification and ordering of visual feature data.
105. And matching the images of the ancient inscriptions, the handwritten pattern characters and the images of the reference ancient inscriptions and the handwritten pattern characters in an image database of the handwritten pattern characters by using the residual third characteristic visual characteristics after the mismatching visual characteristic pair is removed, and determining the visual characteristic distances between the images of the ancient inscriptions, the handwritten pattern characters and the images of the candidate ancient inscriptions and the handwritten pattern characters, wherein the visual characteristic distances are used for representing the similarity, so that whether the images of the ancient inscriptions, the handwritten pattern characters and the images of the similar ancient inscriptions and the handwritten pattern characters are successfully matched or not is determined. Sequencing the images of the candidate ancient inscriptions and handwritten pattern characters according to the visual characteristic distance between the image of the inquired ancient inscription and handwritten pattern character and the image of each candidate ancient inscription and handwritten pattern character; and determining an image set of similar ancient inscriptions and handwritten pattern characters from the images of the candidate ancient inscriptions and handwritten pattern characters according to the sequencing result to form a fourth visual characteristic data set.
106. Removing the matched pairs of visual features that exceed the threshold from the plurality of sets of pairs of visual features in the fourth classification and ordering of visual feature data to form a fifth set of visual feature data.
107. Checking local visual feature pairs of feature images of a plurality of groups of images of similar ancient inscriptions and handwritten pattern characters of the fifth visual feature data set in a Hough transform algorithm mode to obtain visual feature distances among the images, and sequencing the images of the candidate ancient inscriptions and handwritten pattern characters according to the visual feature distances; and determining an image set of similar ancient inscriptions and handwritten pattern characters from the images of the candidate ancient inscriptions and handwritten pattern characters according to the sequencing result to form a sixth visual characteristic data set.
108. Visual feature pairs that exceed the threshold mismatch are removed from the sets of visual feature pairs in the sixth visual feature data classification and ordering to form a seventh visual feature data set.
109. And according to the seventh visual characteristic data set, calculating affine mapping transformation between the images of the similar ancient inscriptions and the handwritten pattern characters and the images of the query ancient inscriptions and the handwritten pattern characters, obtaining the number of the inner group points according to the result of the radial mapping, calculating the number of all the inner group points, and if the number of the inner group points is less than a certain threshold value, indicating that the matching of the query ancient inscriptions and the handwritten pattern character images with the similar ancient inscriptions and the handwritten pattern character images fails, otherwise, indicating that the matching succeeds and entering the next step.
110. And calculating the error of affine transformation between the query image and the similar image according to the error of affine transformation between the image of the similar ancient inscription and handwritten pattern character and the image of the query ancient inscription and handwritten pattern character. If the error is larger than a certain threshold value, the matching of the query ancient inscription, the handwritten pattern character image and the similar ancient inscription and the handwritten pattern character image is failed, otherwise, the matching is successful, and the modern standard Chinese characters or phrases corresponding to the successfully matched ancient inscription and handwritten pattern character image are output.
In the embodiment of the present invention, an image search technique may also be used to first search one or more similar images matching the first visual feature of the query image from each reference image in the reference image database. Then, a second matching is performed with the query image for each similar image. In the second matching process, for each similar image, the distance between the fourth visual feature of the query image and the fourth visual feature of the similar image can be calculated, and a plurality of groups of visual feature pairs are formed by using the similar features with the shorter distances. And then removing the mismatched visual feature pairs from the plurality of sets of visual feature pairs. And then determining whether the query image and the similar image are successfully matched or not according to the remaining visual feature pairs. Finally, it can be determined which similar images match the query image successfully.
In the above step, the fifth visual feature may be a global visual feature that describes the full-image visual content of the image, or may be a local visual feature that can describe the local visual content of the image. The global visual features of the fifth visual Feature include, but are not limited to, any one of a BOW (Bag of Words Bag model) Feature, a local Feature aggregation Descriptor (Vector of locally Aggregated Descriptor), and a fisher Vector Feature, and the fourth visual Feature includes at least one of a Scale-invariant Feature (Scale-invariant Feature Transform), an accelerated Robust Feature (Speeded Up Robust Feature), or any combination thereof.
The fifth visual characteristic may comprise a local visual characteristic. The local visual features related to the fifth visual feature may be more lightweight binarized visual features, such as any one of orb (organized FAST and Rotated brief) features, FREAK (FAST Retina Keypoint) features.
The distance between the two visual features can be calculated through a certain measurement mode, and the closer the distance between the two visual features is, the higher the visual similarity of the two images can be represented.
In the image recognition process of this embodiment, the step of retrieving the similar images according to the fifth visual feature of the query image may be performed by a server in the cloud or may be performed at a terminal.
Terminals involved in the implementation of the present invention may include, but are not limited to, mobile phones, Personal Digital Assistants (PDAs), wireless handheld devices, Tablet computers, personal computers, wearable devices such as: smart glasses, smart watches, smart bracelets, and the like.
In another example, the step of performing secondary matching on the fifth visual feature of the query image and the fifth visual features of the similar images may be performed in a server in the cloud or in a terminal. Compared with the number of reference images in the reference image database, the number of similar images is small, the requirement on the processing capacity of the equipment is not high, and therefore the terminal can process the images quickly.
In another example, if the similar images are retrieved at the cloud according to the fifth visual features of the query image, the cloud may send the retrieved similar images to the terminal, and perform secondary matching on the fifth visual features of the query image and the fifth visual features of the similar images at the terminal. Therefore, the part with high requirement on the processing capacity can be placed at the cloud end, the part with low requirement on the processing capacity of the equipment can be placed at the terminal, and the computing resources can be utilized more reasonably.
The method for identifying ancient graphic and text in another embodiment of the invention is different from the previous embodiment in that the steps can include: establishing an image feature index library according to second visual feature data, third visual feature data, fourth visual feature data, fifth visual feature data, sixth visual feature data and seventh visual feature data corresponding to each reference image in the reference image database; and searching in the image characteristic index library according to the visual characteristic of the query image to obtain similar images similar to the visual characteristic of the query image.
The ancient inscription and handwritten pattern character image visual characteristic index database data is prepared and annotated by ancient inscription and handwritten pattern character research experts and scholars on recognized ancient Chinese character data, the ancient inscription and handwritten pattern character are gradually changed due to the difference of the ages, data matching and input arrangement are carried out according to the requirements of the whole section of characters or other patterns through a neural network algorithm, a deep learning technology and an artificial intelligent image recognition technology in the process of establishing database data, samples are repeatedly added and modified to carry out artificial intelligent image recognition training, and the data of the ancient inscription and handwritten pattern character visual image characteristic index database is perfected.
The images of the second visual image feature index library of the ancient inscriptions and handwritten pattern characters can comprise a second visual feature and a set of structural information thereof, wherein the second visual feature and the set of structural information are formed by organizing the second visual features of a plurality of ancient inscriptions and handwritten pattern character images according to a specific mode, and are used for accelerating the retrieval process of the ancient inscriptions and the handwritten pattern character images. For example, the index structure of the second visual image feature index library of ancient inscriptions and handwritten pattern characters can be in an inverted index mode. The second visual characteristics of the reference images in the reference image database are indexed in an inverted index mode, so that the images containing the same visual words can be organized into the same index, and one image can appear in multiple indexes. The way of the inverted index is only an example, and other ways, such as a hierarchical inverted index, may also be adopted to establish the image feature index library, which is not limited in this embodiment.
Extracting a second visual feature from the query image; after determining the visual words included in the query image according to the local visual features in the second visual features, the indexes including the visual words of the query image can be searched in the image feature index library according to the inverted index structure of the image feature index library, and each index has a corresponding relationship with the image. The image corresponding to the index comprising the visual word of the query image belongs to the set of candidate images. Then, according to the first visual feature of the query image and the first visual feature of each candidate image in the candidate image set, determining the visual feature distance between the query image and each candidate image, and thus determining the similarity between the query image and each candidate image. And finally, according to the visual characteristic distance between the query image and each candidate image, sequencing each candidate image according to a certain sequence, such as the sequence of the distance from small to large. From the candidate images, one or several candidate images, for example, highest ranked or top ranked ones, are selected as similar images having the highest similarity to the query image.
In one possible implementation, the second visual feature of the query image includes a plurality of local visual features. And calculating the distance between each local visual feature of the query image and each local visual feature of the similar image, and forming a plurality of groups of local visual feature pairs according to the distances.
The distance between a local visual feature of the query image and each local visual feature of the similar images can also be calculated, the local visual feature of the similar image with the closest distance is selected, and a local visual feature pair is formed with the local visual feature of the query image.
In another approach, the distance of a local second visual feature of the query image from each local visual feature of the second visual features of the similar images may be calculated. And calculating the ratio of the nearest distance to the next nearest distance, if the ratio is smaller than a set threshold, selecting the local visual features of the similar images with the nearest distance to form local visual feature pairs with the local visual features of the query graphic text images. If the ratio is greater than or equal to the set threshold, then there is no feature in the similar image that can form a local visual feature pair with the local visual feature of the query graphic text image.
Referring to fig. 2, the device for identifying ancient graphic characters comprises an image input module 21 for obtaining an image of ancient inscription and handwritten graphic characters to be queried; the processing module 22 is used for carrying out gap perspective processing and normalized data processing on the ancient inscription and handwritten pattern character images obtained by the image input module; the matching module 23 is configured to match the second visual feature data, the third visual feature data, the fourth visual feature data, the fifth visual feature data, the sixth visual feature data, and the seventh visual feature data of the ancient inscription and handwritten pattern character image to the visual features of the similar image to form a plurality of groups of visual feature pairs, where the visual features include local visual features; the similar images are retrieved according to second visual features of the ancient inscriptions and the handwritten pattern character images to be inquired, and the second visual features comprise global visual features and/or local visual features of the images; a removing module 24, configured to remove a visual feature pair subjected to mismatching from the multiple sets of visual feature pairs, and screen out a suitable visual feature data pair; a confirming module 25, configured to determine whether the ancient inscription and handwritten pattern text image to be queried is successfully matched with the similar image according to the appropriate pair of visual feature data obtained by the removing module 24; and a Chinese character output module 26 for outputting normalized Chinese characters and associated evolution annotations according to the determined similar images.
Claims (9)
1. A method for identifying ancient graphics and characters is characterized by comprising the following steps:
acquiring ancient inscription and handwritten pattern character images to be inquired;
performing perspective processing on the image data stroke gaps of the ancient inscriptions and handwritten pattern characters to be inquired, and obtaining corresponding image data of a first visual characteristic;
classifying the image data of the first visual characteristic into a perspective image data set with uniform size and pixels to form second visual characteristic data;
converting the stroke number or the percentage interval range of the shadow coverage rate in the image of the second visual feature data into different classified characters for distinguishing and classifying to form a third visual feature classification and sequencing;
performing plane translation, plane rotation and scaling on the images of the third characteristic visual characteristic pair, matching the images of the reference ancient inscriptions and handwritten pattern characters in an image database of the third characteristic visual characteristic pair, determining visual characteristic distances between the images of the query ancient inscriptions and handwritten pattern characters and the images of the candidate ancient inscriptions and handwritten pattern characters, sequencing the images of the candidate ancient inscriptions and handwritten pattern characters according to the visual characteristic distances, and determining image sets of the similar ancient inscriptions and handwritten pattern characters from the images of the candidate ancient inscriptions and handwritten pattern characters according to the sequencing result;
removing the matched pairs of visual features that exceed the threshold from the plurality of sets of pairs of visual features in the fourth classification and ordering of visual feature data to form a fifth set of visual feature data.
Checking local visual feature pairs of feature images of a plurality of groups of images of similar ancient inscriptions and handwritten pattern characters of the fifth visual feature data set to obtain visual feature distances among the images, and sequencing the images of the candidate ancient inscriptions and handwritten pattern characters according to the visual feature distances; determining an image set of similar ancient inscriptions and handwritten pattern characters from images of the candidate ancient inscriptions and handwritten pattern characters according to the sequencing result to form a sixth visual characteristic data set;
removing the visual feature pairs which are mismatched beyond the threshold value from the plurality of groups of visual feature pairs in the sixth visual feature data classification and sorting to form a seventh visual feature data set;
and calculating affine mapping transformation and errors between the images of the similar ancient inscriptions and the handwritten pattern characters and the images of the query ancient inscriptions and the handwritten pattern characters according to the seventh visual characteristic data set, indicating successful matching when the calculation result is within a preset range, and outputting modern standard Chinese characters or phrases corresponding to the successfully matched ancient inscriptions and handwritten pattern character images.
2. The method of claim 1, wherein said step of calculating affine mapping transformations and errors between images of similar ancient inscriptions, handwritten pattern words and images of query ancient inscriptions, handwritten pattern words from a seventh visual feature data set comprises the steps of:
according to the seventh visual characteristic data set, affine mapping transformation between the images of the similar ancient inscriptions and the handwritten pattern characters and the images of the query ancient inscriptions and the handwritten pattern characters is calculated, the number of the inner group points is obtained according to the result of the radial mapping, after all the number of the inner group points is calculated, if the number of the inner group points is less than a certain threshold value, the matching failure of the query ancient inscriptions and the handwritten pattern character images with the similar ancient inscriptions and the handwritten pattern character images is indicated, otherwise, the matching success is indicated, and the next step is carried out;
and calculating the error of affine transformation between the query image and the similar image according to the error of affine transformation between the image of the similar ancient inscription and the handwritten pattern character and the image of the query ancient inscription and the handwritten pattern character, wherein if the error is larger than a certain threshold value, the matching failure of the image of the query ancient inscription and the handwritten pattern character with the similar ancient inscription and the handwritten pattern character is represented, and otherwise, the matching success is represented.
3. The method of claim 1, wherein the steps of removing the matched pairs of visual features exceeding a threshold from the plurality of groups of pairs of visual features in the classification and ranking of the third visual features before performing plane translation, plane rotation and scaling on the images of the pairs of visual features of the third feature are performed.
4. The method for identifying ancient graphic and text according to claim 1, wherein local visual feature pairs of feature images of a plurality of groups of similar ancient inscriptions and handwritten graphic and text images of a fifth visual feature data set are verified by means of Hough transform algorithm
5. The method of claim 1, wherein the fifth visual characteristic comprises a global visual characteristic describing a global visual content of the image and a local visual characteristic describing a local visual content of the image.
6. The method of claim 5, wherein the global visual features comprise any one of bag of words model features, local feature aggregation descriptors, and Fisher vector features, and the local visual features comprise any one of ORBs and fast retinal keypoint features.
7. The method of claim 1, wherein the fourth visual feature comprises at least one of a scale-invariant feature, an accelerated robust feature, or any combination thereof.
8. An apparatus for recognizing ancient graphic characters, which is used for realizing the method for recognizing ancient graphic characters according to any one of claims 1 to 7, and comprises:
the image input module is used for acquiring ancient inscription and handwritten pattern character images to be inquired;
the processing module is used for carrying out gap perspective processing and normalized data processing on the ancient inscription and handwritten pattern character images obtained by the image input module;
the matching module is used for matching second visual characteristic data, third visual characteristic data, fourth visual characteristic data, fifth visual characteristic data, sixth visual characteristic data and seventh visual characteristic data of the inquired ancient inscription and handwritten pattern character images with visual characteristics of similar images to form a plurality of groups of visual characteristic pairs, wherein the visual characteristics comprise local visual characteristics; the similar images are retrieved according to second visual features of the ancient inscriptions and the handwritten pattern character images to be inquired, and the second visual features comprise global visual features and/or local visual features of the images;
the removing module is used for removing the visual feature pairs which are mismatched in error from the plurality of groups of visual feature pairs and screening out proper visual feature data pairs;
the confirming module is used for confirming whether the ancient inscription and handwritten pattern character image to be inquired is successfully matched with the similar image or not according to the proper visual characteristic data pair obtained by the removing module;
and the Chinese character output module is used for outputting standard Chinese characters and related evolution annotations according to the determined similar images.
9. A program storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method of identifying ancient graphic texts according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010644573.3A CN111898618B (en) | 2020-07-07 | 2020-07-07 | Method, device and program storage medium for identifying ancient graphic characters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010644573.3A CN111898618B (en) | 2020-07-07 | 2020-07-07 | Method, device and program storage medium for identifying ancient graphic characters |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111898618A true CN111898618A (en) | 2020-11-06 |
| CN111898618B CN111898618B (en) | 2024-02-27 |
Family
ID=73191852
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010644573.3A Active CN111898618B (en) | 2020-07-07 | 2020-07-07 | Method, device and program storage medium for identifying ancient graphic characters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111898618B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114972506A (en) * | 2022-05-05 | 2022-08-30 | 武汉大学 | Image positioning method based on deep learning and street view image |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108334644A (en) * | 2018-03-30 | 2018-07-27 | 百度在线网络技术（北京）有限公司 | Image-recognizing method and device |
| CN111241329A (en) * | 2020-01-06 | 2020-06-05 | 北京邮电大学 | Image retrieval-based ancient character interpretation method and device |
-
2020
- 2020-07-07 CN CN202010644573.3A patent/CN111898618B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108334644A (en) * | 2018-03-30 | 2018-07-27 | 百度在线网络技术（北京）有限公司 | Image-recognizing method and device |
| US20190303700A1 (en) * | 2018-03-30 | 2019-10-03 | Baidu Online Network Technology (Beijing) Co., Ltd . | Image recognition method and device |
| CN111241329A (en) * | 2020-01-06 | 2020-06-05 | 北京邮电大学 | Image retrieval-based ancient character interpretation method and device |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114972506A (en) * | 2022-05-05 | 2022-08-30 | 武汉大学 | Image positioning method based on deep learning and street view image |
| CN114972506B (en) * | 2022-05-05 | 2024-04-30 | 武汉大学 | Image positioning method based on deep learning and street view image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111898618B (en) | 2024-02-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108664996B (en) | Ancient character recognition method and system based on deep learning | |
| Wilkinson et al. | Neural Ctrl-F: segmentation-free query-by-string word spotting in handwritten manuscript collections | |
| CN108664975B (en) | Uyghur handwritten letter recognition method and system and electronic equipment | |
| Obaidullah et al. | Handwritten Indic script identification in multi-script document images: a survey | |
| WO2021042505A1 (en) | Note generation method and apparatus based on character recognition technology, and computer device | |
| CN109784146A (en) | A kind of font type recognition methods, electronic equipment, storage medium | |
| Shi et al. | Stroke detector and structure based models for character recognition: a comparative study | |
| CN107526721B (en) | Ambiguity elimination method and device for comment vocabularies of e-commerce products | |
| Abuzaraida et al. | Online handwriting Arabic recognition system using k-nearest neighbors classifier and DCT features | |
| Izakian et al. | Multi-font Farsi/Arabic isolated character recognition using chain codes | |
| CN110414622B (en) | Classifier training method and device based on semi-supervised learning | |
| Dhanikonda et al. | An efficient deep learning model with interrelated tagging prototype with segmentation for telugu optical character recognition | |
| CN111898618B (en) | Method, device and program storage medium for identifying ancient graphic characters | |
| CN108268883B (en) | Mobile terminal information template self-construction system based on open data | |
| Zhang et al. | OCR with the Deep CNN Model for Ligature Script‐Based Languages like Manchu | |
| Zhang et al. | Deep learning based tangut character recognition | |
| Susanto et al. | Javanese script recognition based on metric, eccentricity and local binary pattern | |
| Salamah et al. | Towards the machine reading of arabic calligraphy: a letters dataset and corresponding corpus of text | |
| CN112560849B (en) | Neural network algorithm-based grammar segmentation method and system | |
| CN115147846A (en) | Multi-language bill identification method, device, equipment and storage medium | |
| Bashir et al. | Script identification: a review | |
| Shafi et al. | Urdu character recognition: A systematic literature review | |
| Kumar et al. | Cross domain descriptor for sketch based image retrieval using siamese network | |
| CN113920291A (en) | Error correction method and device based on picture recognition result, electronic equipment and medium | |
| Jain | Unconstrained Arabic & Urdu text recognition using deep CNN-RNN hybrid networks |
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 |