WO2022116524A1 - 图片识别方法、装置、电子设备及介质 - Google Patents

图片识别方法、装置、电子设备及介质 Download PDF

Info

Publication number
WO2022116524A1
WO2022116524A1 PCT/CN2021/103280 CN2021103280W WO2022116524A1 WO 2022116524 A1 WO2022116524 A1 WO 2022116524A1 CN 2021103280 W CN2021103280 W CN 2021103280W WO 2022116524 A1 WO2022116524 A1 WO 2022116524A1
Authority
WO
WIPO (PCT)
Prior art keywords
character
character string
tree structure
current
string
Prior art date
Application number
PCT/CN2021/103280
Other languages
English (en)
French (fr)
Inventor
王从涛
阳家俊
魏远明
陈伟
韦涛
吴军
龚力
朱伟基
Original Assignee
北京搜狗科技发展有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 北京搜狗科技发展有限公司 filed Critical 北京搜狗科技发展有限公司
Publication of WO2022116524A1 publication Critical patent/WO2022116524A1/zh
Priority to US18/137,884 priority Critical patent/US20230290167A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/31—Indexing; Data structures therefor; Storage structures
    • G06F16/316—Indexing structures
    • G06F16/322—Trees
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/19—Recognition using electronic means
    • G06V30/19007—Matching; Proximity measures
    • G06V30/19073—Comparing statistics of pixel or of feature values, e.g. histogram matching
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/31—Indexing; Data structures therefor; Storage structures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/31—Indexing; Data structures therefor; Storage structures
    • G06F16/316—Indexing structures
    • G06F16/325—Hash tables
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3346—Query execution using probabilistic model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2323—Non-hierarchical techniques based on graph theory, e.g. minimum spanning trees [MST] or graph cuts
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/10—Text processing
    • G06F40/12—Use of codes for handling textual entities
    • G06F40/151—Transformation
    • G06F40/154—Tree transformation for tree-structured or markup documents, e.g. XSLT, XSL-FO or stylesheets
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/26—Techniques for post-processing, e.g. correcting the recognition result
    • G06V30/262—Techniques for post-processing, e.g. correcting the recognition result using context analysis, e.g. lexical, syntactic or semantic context
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
    • H04L9/32—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols including means for verifying the identity or authority of a user of the system or for message authentication, e.g. authorization, entity authentication, data integrity or data verification, non-repudiation, key authentication or verification of credentials
    • H04L9/3236—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols including means for verifying the identity or authority of a user of the system or for message authentication, e.g. authorization, entity authentication, data integrity or data verification, non-repudiation, key authentication or verification of credentials using cryptographic hash functions

Definitions

  • the present disclosure relates to the technical field of image processing, and in particular, to a picture recognition method, apparatus, electronic device, and medium.
  • the comparison complexity of the entire collection is O(n ⁇ 2);
  • the amount of computation required to store the intermediate character string is extremely large, which makes the storage efficiency of the intermediate character string low.
  • the purpose of the present disclosure is, at least in part, to provide a picture recognition method, device, electronic device and medium, which can effectively reduce the amount of calculation required for storing intermediate character strings in the process of character recognition for images, and can effectively improve the storage of intermediate character strings. s efficiency.
  • a first aspect of the present disclosure provides a picture recognition method, including: in the process of performing text recognition on a collected image by an optical character recognition model, acquiring the recognized current character string and its hash value, and using the The current string and its hash value are stored in the first preset tree structure; the new probability value of the current string at the next moment is predicted, and the hash value of the current string is stored in the second preset tree In the shape structure; expand the current character string through the target character set identified at the next moment, obtain the expanded character string set, and store the probability value and hash value of each expanded character string in the expanded character string set into the first preset tree structure, and store the hash value of each extended string in the second preset tree structure;
  • N character strings with the highest probability values are obtained from the first preset tree structure and retained, wherein N is an integer not less than 1; and Taking the N character strings as the current character string, the above steps are repeated until the identification of all the collected images is completed, and the character string with the highest probability value is obtained as the final identification result.
  • obtaining the recognized current character string and its hash value in the process of performing character recognition on the collected image by using the optical character recognition model includes: performing text recognition on the collected image by using the optical character recognition model.
  • the current character string composed of characters with a recognition probability greater than a preset probability is obtained, and a hash value of the current character string is obtained.
  • the predicting a new probability of the current character string at the next moment, and storing the hash value of the current character string in a second preset tree structure includes: predicting the current character The predicted character of the string at the next moment; the combined probability of the current character string and the predicted character is obtained as the new probability value, and the hash value of the current character string is stored in the second preset tree structure wherein, the second preset tree structure is a set structure.
  • expanding the current character string through the target character set identified at the next moment to obtain the expanded character string set includes: obtaining a recognition probability identified at the next moment that is greater than the preset The probabilistic characters form the target character set; the current character string is combined with each character in the target character set, and all the combined character strings are obtained as the expanded character string set.
  • the N character strings with the highest probability values are obtained from the first preset tree structure and retained, including: According to the new probability value of the current string and the probability value of each expanded string, determine the N hash values with the highest probability value from the second preset tree structure; according to the N hash values , and determine the N character strings corresponding to the N hash values from the first preset tree structure.
  • the first preset tree structure is a set structure.
  • a second aspect of the present disclosure provides a picture recognition device, comprising: a character recognition unit for acquiring a recognized current character string and its hash value during the process of character recognition on a collected image by an optical character recognition model , and store the current character string and its hash value in the first preset tree structure; the prediction unit is used to predict the new probability value of the current character string at the next moment, The hash value is stored in the second preset tree structure; the expansion unit is used to expand the current character string through the target character set identified at the next moment, obtain the expanded character string set, and use the expanded character set.
  • the probability value and hash value of each extended string in the string set are stored in the first preset tree structure, and the hash value of each extended string is stored in the second preset tree structure ;
  • the loop execution unit is used to obtain N character strings with the highest probability values from the first preset tree structure according to the hash value stored in the second preset tree structure and retain them, wherein N An integer not less than 1; taking the N character strings as the current character string, repeating the above steps until the identification of all collected images is completed, and obtaining a character string with the highest probability value as the final identification result.
  • the character recognition unit is configured to obtain the current character string composed of characters whose recognition probability is greater than a preset probability during the process of character recognition of the collected image by the optical character recognition model, and obtain all the characters.
  • the hash value of the current string is configured to obtain the current character string composed of characters whose recognition probability is greater than a preset probability during the process of character recognition of the collected image by the optical character recognition model, and obtain all the characters. The hash value of the current string.
  • the prediction unit is configured to predict the predicted character of the current character string at the next moment; obtain the combined probability of the current character string and the predicted character as the new probability value, and use the The hash value of the current character string is stored in the second preset tree structure, wherein the second preset tree structure is a set structure.
  • the expansion unit is configured to acquire characters with a recognition probability greater than the preset probability recognized at the next moment to form the target character set; combine the current character string with each character set in the target character set The characters are combined, and all the combined character strings are obtained as the expanded character string set.
  • the loop execution unit is configured to, according to the new probability value of the current character string and the probability value of each expanded character string, determine the one with the highest probability value from the second preset tree structure N hash values; according to the N hash values, the N character strings corresponding to the N hash values are determined from the first preset tree structure.
  • the first preset tree structure is a set structure.
  • a third aspect of the present disclosure provides an apparatus for data processing, comprising a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be processed by the one or more programs
  • the execution of the one or more programs by a processor includes steps for a picture recognition method as described above.
  • a fourth aspect of the present disclosure provides a machine-readable medium having instructions stored thereon that, when executed by one or more processors, cause an apparatus to perform the steps of the above-described picture recognition method.
  • the present invention adopts the mapping relationship between probability values, hash values and strings, and can determine which string to store only by storing the hash value, which can be extremely It greatly reduces the amount of calculation required for storing strings, which can effectively improve the efficiency of storing strings; and on the basis of improving the efficiency of storing strings, since the stored strings are used for character recognition, the OCR model recognizes The efficiency of writing is also improved.
  • FIG. 1 shows a method flowchart of a picture recognition method according to one or more embodiments of the present disclosure
  • FIG. 2 shows a schematic structural diagram of a picture recognition apparatus according to one or more embodiments of the present disclosure
  • FIG. 3 shows a structural block diagram of a picture recognition apparatus as a device according to one or more embodiments of the present disclosure
  • FIG. 4 shows a structural block diagram of a server according to one or more embodiments of the present disclosure.
  • the present disclosure provides a picture recognition scheme, which is used to obtain a picture in the process of character recognition of a collected image by an optical character recognition model.
  • the identified current string and its hash value, and the current string and its hash value are stored in the first preset tree structure; the new probability value of the current string at the next moment is predicted, and the The hash value of the current character string is stored in the second preset tree structure; the current character string is expanded by the target character set identified at the next moment, the expanded character string set is obtained, and the expanded character set is obtained.
  • the probability value and hash value of each extended string in the string set are stored in the first preset tree structure, and the hash value of each extended string is stored in the second preset tree structure in; according to the hash value stored in the second preset tree structure, obtain N character strings with the highest probability value from the first preset tree structure and keep them, wherein N is not less than an integer of 1 and take the N character strings as the current character string, repeat the above steps until the identification of all collected images is completed, and obtain a character string with the highest probability value as the final identification result.
  • the OCR model when judging that a certain string needs to be stored, it is preferred to determine the hash value of the N characters that need to be stored through the mapping relationship between the probability value and the hash value, and then According to the mapping relationship between the hash value and the string, and according to the hash value of N characters, N characters to be stored are determined;
  • the present invention adopts the mapping relationship between the probability value, the hash value and the character string, which can determine which character string to store only by storing the hash value, which can greatly reduce the
  • the amount of calculation required to store strings can effectively improve the efficiency of storing strings; and on the basis of improving the efficiency of storing strings, since the stored strings are used for character recognition, the OCR model can recognize characters efficiently. also increased.
  • an optical character recognition (Optical Character Recognition, OCR for short) model is usually used to recognize text in an image, and when using the OCR model to recognize text in an image, it is necessary to Use the OCR model to predict the probability of each word for beam search decoding, so that the text in the image can be predicted through the OCR model.
  • OCR Optical Character Recognition
  • the characters may be words, punctuation, symbols, etc., which are not specifically limited in this specification.
  • the present disclosure provides a picture recognition method, which includes the following steps:
  • S103 Expand the current character string through the target character set identified at the next moment, obtain an expanded character string set, and store the probability value and hash value of each expanded character string in the expanded character string set in the first preset tree structure, and the hash value of each extended string is stored in the second preset tree structure;
  • N is an integer not less than 1 and take the N character strings as the current character string, repeat the above steps until the identification of all collected images is completed, and obtain a character string with the highest probability value as the final identification result.
  • the OCR model is usually deployed on an image recognition device
  • the image recognition device can be served by a server
  • the image recognition device can be, for example, a dictionary pen and a reading pen
  • the server can be, for example, a laptop computer, a desktop Computers and All-in-Ones, etc.
  • step S101 in the process of character recognition of the collected image by the OCR model, the current character string is predicted by the OCR model, and the predicted current character string is performed by the hash algorithm.
  • Hash get the hash value of the current string.
  • the hash algorithm may be, for example, MD5, SHA-1, SHA-2, etc., which is not specifically limited in this specification.
  • a current character string composed of characters with a recognition probability greater than a preset probability can be obtained, and a hash value of the current character string can be obtained. , and then store the current string and its hash value in the first preset tree structure.
  • a preset probability can be preset, and then according to the preset probability, each character in the image predicted by the OCR model at the current moment is filtered. , in this way, the current string is obtained by filtering the preset probability, wherein the probability of each word in the current string is greater than the preset probability; then the current string obtained by filtering is hashed by using a hash algorithm to obtain the current string
  • the hash value of the current string and its hash value are stored in a first preset tree structure, wherein the first preset tree structure can be a set structure in C language, or a binary tree structure.
  • a continuous picture set is obtained; then the OCR model is used to identify the picture set.
  • the probability of the word being A1 is 15%, the probability of being A2 is 35%, the probability of being A3 is 65%, the probability of being A4 is 85%, the probability of being A5 is 45%, and the preset probability is 50%, Then filter out A1, A2 and A5, keep A3 and A4, then determine that the current string is A3 and A4; if the preset probability is 70%, then determine that the current string is A4; and after obtaining the current string as A4 , Hash A4 through SHA-2, the obtained hash value H1 is used as the hash value of A4, and then A4 and H1 are stored in the first set structure.
  • step S102 is executed, in this step, the new probability value of the current character string at the next moment is predicted, and the hash value of the current character string is stored into the second preset tree structure.
  • the predicted character of the current character string at the next moment is predicted; the combined probability of the current character string and the predicted character is obtained as a new probability value, and the hash value of the current character string is stored in the second preset value;
  • the second preset tree structure may be a set structure and a binary tree structure.
  • the last character in the current character string can be used as the predicted character, and the space can also be used as the predicted character, so that the predicted character can include Any one or more of the last character and a space in the current string.
  • obtaining the combined probability of the current character string and the predicted character is to obtain the sum of the combined probabilities of the current character and each predicted character as the new probability value, or the sum of the weights of the combined probabilities can be the new probability value , or the product of the combined probability may be a new probability value, which is not specifically limited in this specification.
  • the current character string is additionally stored in the second preset tree structure, wherein, there are no repeated numbers in the second preset tree structure, and the corresponding Also, there are no repeated numbers in the first preset tree structure.
  • the probability value of abc+space is predicted by the OCR model, and the probability value of abc+c is predicted to be H2. If the probability value is H3, the new probability value of abc is determined to be H2+H3, and the hash value X of abc is stored in the second set structure.
  • step S103 is executed.
  • the current character string is expanded by the target character set identified at the next moment to obtain the expanded character string set, and the probability value and hash value of each expanded character string in the expanded character string set are obtained.
  • the hash value is stored in the first preset tree structure, and the hash value of each extended string is stored in the second preset tree structure.
  • character recognition is performed on the image collected at the next moment by the OCR model, and according to a preset probability, each character in the image predicted by the OCR model at the next moment is filtered, and the recognition probability is obtained.
  • the target character set of all word group words with a preset probability is greater than the preset probability, and then the current string is combined with each character in the target character set, and all the combined character strings are obtained as the expanded string set.
  • abcd For example, also taking the current character string as abc as an example, if the image collected at time T+1 is recognized by the OCR model, and the characters whose recognition probability is greater than the preset probability are identified as d, e and f, respectively, abcd, abce are obtained.
  • step S104 storing the probability value and hash value of each extended string in the extended string set in the first preset tree structure, and storing the hash value of each extended string in the second preset tree structure After that, step S104 is performed.
  • N hash values with the highest probability values may be determined from the second preset tree structure according to the new probability value of the current character string and the probability value of each expanded character string; according to the N hash values , determine N character strings corresponding to N hash values from the first preset tree structure; reserve N character strings, and keep them as an integer, where N is not less than 1; and use N character strings as For the current character string, the above steps are repeated until the recognition of all the collected images is completed, and a character string with the highest probability value is obtained from the first preset tree structure as the final recognition result.
  • the value of N can be the maximum set number in the beam search in the OCR model, or can be set manually or by equipment, for example, it can be 1, 2, 3, and 5, etc., This specification does not make specific restrictions.
  • the desired values are X2 and X3; and according to X2 and X3, the string corresponding to X2 obtained from the first set structure is abce, and the string corresponding to X3 is abcf, then the N strings are determined to be abce and abcf; At this time, the abce and abcf in the first set structure are used as the current strings and retained, and only the hash values X2 and X3 corresponding to abce and abcf may be retained in the second set structure; and then the next step is processed.
  • the new probability of abce at time T+2 is predicted to be the probability of abce+space and abce+e. and predict that the new probability of abcf at time T+2 is the sum of the probability of abcf+space and abcf+f; if time T+2 is used as the current time, the real-time images collected are identified by OCR mode, If the characters whose recognition probability is greater than the preset probability are identified as p and k, abcep, abcek, abcfp and abcfk are obtained as the expanded string set, and then the probability values of abcep, abcek, abcfp and abcfk are predicted by the OCR model as D4, D5, D6 and D7, and perform hash calculation on abcep, abcek, abcfp and abcfk through the hash algorithm, and the obtained hash values are X4, X5, X6 and
  • the probability value the mapping relationship between the hash value and the character string, it is determined that the N character strings are: abcek and abcep, at this time, use abce and abcf in the first set structure as the current string and keep it, and also keep only the hash values X2 and X3 corresponding to abce and abcf in the second set structure; then use the same
  • the next step is processed in the manner of , until the recognition of all the collected images is completed, and a character string with the highest probability value is obtained as the final recognition result.
  • the hash value of the N characters that need to be stored determines the N characters that need to be stored; compared with the prior art, not only does it not need to compare the lengths of the stored fields, but also does not need to compare the lengths of the fields one by one before storing them.
  • the present invention adopts the mapping relationship between probability value, hash value and character string, which can determine which character string to store only by storing the hash value, which can greatly reduce the amount of calculation required for storing character strings, and can effectively improve the The efficiency of storing character strings; and on the basis of improving the efficiency of storing character strings, since the stored character strings are used for character recognition, the efficiency of character recognition by the OCR model is also improved.
  • the present disclosure also provides a picture recognition device, including:
  • the character recognition unit 201 is used to obtain the recognized current character string and its hash value in the process of character recognition of the collected image by the optical character recognition model, and store the current character string and its hash value in a in the first preset tree structure;
  • the prediction unit 202 is used to predict the new probability value of the current character string at the next moment, and store the hash value of the current character string in the second preset tree structure;
  • the expansion unit 203 is used to expand the current character string through the target character set identified at the next moment, obtain the expanded character string set, and combine the probability value and hash of each expanded character string in the expanded character string set.
  • the value is stored in the first preset tree structure, and the hash value of each extended string is stored in the second preset tree structure;
  • the loop execution unit 204 is configured to obtain N character strings with the highest probability values from the first preset tree structure according to the hash value stored in the second preset tree structure and retain them, wherein N An integer not less than 1; taking the N character strings as the current character string, repeating the above steps until the identification of all collected images is completed, and obtaining a character string with the highest probability value as the final identification result.
  • the character recognition unit 201 is configured to obtain the current character string composed of characters whose recognition probability is greater than a preset probability during the process of character recognition of the collected image by an optical character recognition model, and obtain the The hash value of the current string.
  • the prediction unit 202 is configured to predict the predicted character of the current character string at the next moment; obtain the combined probability of the current character string and the predicted character as the new probability value, and use the The hash value of the current string is stored in the second preset tree structure, where the second preset tree structure is a set structure.
  • the expansion unit 203 is configured to obtain characters whose recognition probability is greater than the preset probability recognized at the next moment to form the target character set; The characters are combined to obtain all the combined strings as the expanded string set.
  • the loop execution unit 204 is configured to determine N with the highest probability value from the second preset tree structure according to the new probability value of the current character string and the probability value of each expanded character string and determining the N character strings corresponding to the N hash values from the first preset tree structure according to the N hash values.
  • the first preset tree structure is a set structure.
  • FIG. 3 is a structural block diagram of a picture recognition apparatus provided as a device according to an embodiment of the present disclosure.
  • the apparatus 900 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, or the like.
  • the apparatus 900 may include one or more of the following components: a processing component 902, a memory 904, a power supply component 906, a multimedia component 908, an audio component 910, an input/output (I/O) interface 912, a sensor component 914, And the communication component 916 .
  • the processing component 902 generally controls the overall operation of the device 900, such as operations associated with display, incoming calls, data communications, camera operations, and recording operations.
  • the processing element 902 may include one or more processors 920 to execute instructions to perform all or part of the steps of the methods described above.
  • processing component 902 may include one or more modules to facilitate interaction between processing component 902 and other components.
  • processing component 902 may include a multimedia module to facilitate interaction between multimedia component 908 and processing component 902.
  • Memory 904 is configured to store various types of data to support operation at device 900 . Examples of such data include instructions for any application or method operating on device 900, contact data, caller book data, messages, pictures, videos, and the like. Memory 904 may be implemented by any type of volatile or non-volatile storage device or combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read only memory
  • EPROM erasable Programmable Read Only Memory
  • PROM Programmable Read Only Memory
  • ROM Read Only Memory
  • Magnetic Memory Flash Memory
  • Magnetic or Optical Disk Magnetic Disk
  • Power supply assembly 906 provides power to various components of device 900 .
  • Power supply components 906 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to device 900 .
  • Multimedia component 908 includes a screen that provides an output interface between the device 900 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user.
  • the touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundaries of a touch or swipe motion action, but also detect the duration and pressure associated with the touch or swipe action.
  • the multimedia component 908 includes a front-facing camera and/or a rear-facing camera. When the device 900 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data. Each of the front and rear cameras can be a fixed optical lens system or have focal length and optical zoom capability.
  • Audio component 910 is configured to output and/or input audio signals.
  • audio component 910 includes a microphone (MIC) that is configured to receive external audio signals when device 900 is in operating modes, such as call mode, recording mode, and voice recognition mode. The received audio signal may be further stored in memory 904 or transmitted via communication component 916 .
  • audio component 910 also includes a speaker for outputting audio signals.
  • the I/O interface 912 provides an interface between the processing component 902 and a peripheral interface module, which may be a keyboard, a click wheel, a button, or the like. These buttons may include, but are not limited to: home button, volume buttons, start button, and lock button.
  • Sensor assembly 914 includes one or more sensors for providing status assessment of various aspects of device 900 .
  • the sensor assembly 914 can detect the open/closed state of the device 900, the relative positioning of components, such as the display and keypad of the device 900, and the sensor assembly 914 can also detect changes in the position of the device 900 or a component of the device 900 , the presence or absence of user contact with the device 900 , the orientation or acceleration/deceleration of the device 900 and the temperature change of the device 900 .
  • Sensor assembly 914 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact.
  • Sensor assembly 914 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor assembly 914 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • Communication component 916 is configured to facilitate wired or wireless communication between apparatus 900 and other devices.
  • the device 900 may access a wireless network according to a communication standard, such as WiFi, 2G or 3G, or a combination thereof.
  • the communication component 916 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communication component 916 also includes a near field communication (NFC) module to facilitate short-range communication.
  • NFC near field communication
  • the NFC module may be implemented according to radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
  • RFID radio frequency identification
  • IrDA infrared data association
  • UWB ultra-wideband
  • Bluetooth Bluetooth
  • apparatus 900 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gates An array (FPGA), controller, microcontroller, microprocessor, or other electronic component implementation for performing the above method.
  • ASICs application specific integrated circuits
  • DSPs digital signal processors
  • DSPDs digital signal processing devices
  • PLDs programmable logic devices
  • FPGA field programmable gates
  • controller microcontroller, microprocessor, or other electronic component implementation for performing the above method.
  • a non-transitory computer-readable storage medium including instructions such as a memory 904 including instructions, executable by the processor 920 of the apparatus 900 to perform the method described above is also provided.
  • the non-transitory computer-readable storage medium may be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like.
  • FIG. 4 is a structural block diagram of a server in some embodiments of the present disclosure.
  • the server 1900 may vary widely depending on configuration or performance, and may include one or more central processing units (CPUs) 1922 (eg, one or more processors) and memory 1932, one or more A storage medium 1930 (eg, one or more mass storage devices) that stores applications 1942 or data 1944 above.
  • the memory 1932 and the storage medium 1930 may be short-term storage or persistent storage.
  • the program stored in the storage medium 1930 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server.
  • the central processing unit 1922 may be configured to communicate with the storage medium 1930 to execute a series of instruction operations in the storage medium 1930 on the server 1900 .
  • Server 1900 may also include one or more power supplies 1926, one or more wired or wireless network interfaces 1950, one or more input/output interfaces 1958, one or more keyboards 1956, and/or, one or more operating systems 1941 , such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM and so on.
  • a non-transitory computer-readable storage medium when instructions in the storage medium are executed by a processor of an apparatus (device or server), the apparatus enables the apparatus to execute a picture recognition method, the method comprising: In the process of character recognition of the collected image by the character recognition model, the recognized current character string and its hash value are obtained, and the current character string and its hash value are stored in the first preset tree structure; The new probability value of the current character string at the next moment, the hash value of the current character string is stored in the second preset tree structure; Carry out expansion, obtain an expanded string set, and store the probability value and hash value of each expanded string in the expanded string set in the first preset tree structure, and store the probability value of each expanded string in the first preset tree structure.
  • the hash value is stored in the second preset tree structure; according to the hash value stored in the second preset tree structure, the N with the highest probability value is obtained from the first preset tree structure and keep the number of strings, where N is an integer not less than 1; take the N strings as the current string, repeat the above steps until the identification of all collected images is completed, and obtain the string with the highest probability value as the final string Identify the results.
  • These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory result in an article of manufacture comprising the instruction apparatus, the instructions
  • the device implements the functions specified in the flow or flows of the flowcharts and/or the block or blocks of the block diagrams.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Computational Linguistics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computer Security & Cryptography (AREA)
  • Probability & Statistics with Applications (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • General Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Discrete Mathematics (AREA)
  • Character Discrimination (AREA)
  • Image Analysis (AREA)

Abstract

一种图片识别方法、装置、电子设备及介质，在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将当前字符串及其哈希值存储到第一预设树形结构及第二预设树形结构中中，预测当前字符串在下一时刻的新概率值，获取扩充字符串集，根据第二预设树形结构中存储的哈希值，获取概率值最高的N个字符串并保留，并将N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。能够有效降低针对图像进行文字识别过程存储中间字符串所需的计算量，有效提高存储中间字符串的效率。

Description

图片识别方法、装置、电子设备及介质
相关申请的交叉引用
本申请要求于2020年12月04日提交、申请号为202011409602.4且名称为“图片识别方法、装置、电子设备及介质”的中国专利申请的优先权，其全部内容通过引用合并于此。
技术领域
本公开内容涉及图像处理技术领域，尤其涉及一种图片识别方法、装置、电子设备及介质。
背景技术
本公开内容随着互联网技术的飞速发展，用户可以使用图像识别装置例如词典笔或点读笔对非母语言的文本进行翻译或者注释，提高用户的学习效率。
但是，现有技术中在针对图像进行文字识别时，需要存储中间字符串及其对应的概率值，例如需要存储a、ab、abc等字符串及概率值，如果字段相同则需要累加起来，如果字段不同则需要和其他字段一起存储起来；通常的做法是先比较两个字段的长度，如果长度不同，则这两个字段肯定不同；如果长度相同，则一个一个字符的比较，遇到不同的则说明两个字段不同，此时如果集合存储了n个字段，每个字段的比较需要O(n)的复杂度，则整个集合的比较复杂度为O(n^2)；使得在针对图像进行文字识别过程中存储中间字符串所需的计算量极大，使得中间字符串存储的效率较低。
发明内容
本公开内容的目的至少部分在于，提供了一种图片识别方法、装置、电子设备及介质，能够有效降低针对图像进行文字识别过程存储中间字符串所需的计算量，能够有效提高存储中间字符串的效率。
本公开内容的第一方面提供了一种图片识别方法，包括：在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将所述当前字符串及其哈希值存储到第一预设树形结构中；预测所述当前字符串在下 一时刻的新概率值，将所述当前字符串的哈希值存储到第二预设树形结构中；通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，并将所述扩充字符串集中每个扩充字符串的概率值及哈希值存储到所述第一预设树形结构中，以及将每个扩充字符串的哈希值存储到所述第二预设树形结构中；
根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，其中，N不小于1的整数；并将所述N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。
在一些实施例中，所述在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，包括：在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别概率大于预设概率的字符组成的所述当前字符串，并获取所述当前字符串的哈希值。
在一些实施例中，所述预测所述当前字符串在下一时刻的新概率，将所述当前字符串的哈希值存储到第二预设树形结构中，包括：预测出所述当前字符串在下一时刻的预测字符；获取所述当前字符串与所述预测字符的组合概率为所述新概率值，将所述当前字符串的哈希值存储到所述第二预设树形结构中，其中，所述第二预设树形结构为set结构。
在一些实施例中，所述通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，法包括：获取下一时刻识别出的识别概率大于所述预设概率的字符组成所述目标字符集；将所述当前字符串与所述目标字符集中每个字符进行组合，得到组合后的所有字符串为所述扩充字符串集。
在一些实施例中，所述根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，包括：根据所述当前字符串的新概率值和每个扩充字符串的概率值，从所述第二预设树形结构中确定概率值最高的N个哈希值；根据所述N个哈希值，从所述第一预设树形结构中确定与所述N个哈希值对应的所述N个字符串。
在一些实施例中，所述第一预设树形结构为set结构。
本公开内容第二方面提供了一种图片识别装置，包括：字符识别单元，用于在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将所述当前字符串及其哈希值存储到第一预设树形结构中；预测单元，用于预测所述当前字符串在下一时刻的新概率值，将所述当前字符串的哈希值存储到第二预设树形结构中；扩充单元，用于通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，并将所述扩充字符串集中每个扩充字符串的概率值及哈希值存储到所述第一预设树形结构中，以及将每个扩充字符串的哈希值存储到所述第二预设树形结构中；循环执行单元，用于根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，其中，N不小于1的整数；并将所述N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。
在一些实施例中，所述字符识别单元，用于在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别概率大于预设概率的字符组成的所述当前字符串，并获取所述当前字符串的哈希值。
在一些实施例中，所述预测单元，用于预测出所述当前字符串在下一时刻的预测字符；获取所述当前字符串与所述预测字符的组合概率为所述新概率值，将所述当前字符串的哈希值存储到所述第二预设树形结构中，其中，所述第二预设树形结构为set结构。
在一些实施例中，所述扩充单元，用于获取下一时刻识别出的识别概率大于所述预设概率的字符组成所述目标字符集；将所述当前字符串与所述目标字符集中每个字符进行组合，得到组合后的所有字符串为所述扩充字符串集。
在一些实施例中，所述循环执行单元，用于根据所述当前字符串的新概率值和每个扩充字符串的概率值，从所述第二预设树形结构中确定概率值最高的N个哈希值；根据所述N个哈希值，从所述第一预设树形结构中确定与所述N个哈希值对应的所述N个字符串。
在一些实施例中，所述第一预设树形结构为set结构。
本公开内容第三方面提供了一种用于数据处理的装置，包括有存储器，以及 一个或者一个以上的程序，其中一个或者一个以上程序存储于存储器中，且经配置以由一个或者一个以上处理器执行所述一个或者一个以上程序包含用于如上述图片识别方法的步骤。
本公开内容第四方面提供了一种机器可读介质，其上存储有指令，当由一个或多个处理器执行时，使得装置执行如上述图片识别方法的步骤。
根据上述技术方案，在通过OCR模型对图像进行文字识别过程中，在判断某个字符串需要存储时，首选通过概率值与哈希值的映射关系，确定出需要存储的N个字符的哈希值，再根据哈希值与字符串的映射关系，根据N个字符的哈希值，确定出需要存储的N个字符；与现有技术相比，不仅不需要比较存储的字段之间的长度，也不需要在字段长度相同时逐个字符比较之后再存储，而本发明采用概率值、哈希值与字符串的映射关系，可以仅通过存储哈希值即可确定存储哪个字符串，能够极大的降低进行存储字符串所需的计算量，能够有效提高存储字符串的效率；以及在提高存储字符串的效率基础上，由于存储的字符串是用于进行文字识别的，使得OCR模型识别文字的效率也随之提高。
附图说明
图1示出了依据本公开内容的一个或多个实施例的图片识别方法的方法流程图；
图2示出了依据本公开内容的一个或多个实施例的图片识别装置的结构示意图；
图3示出了依据本公开内容的一个或多个实施例的用于图片识别装置的作为设备时的结构框图；
图4示出了依据本公开内容的一个或多个实施例的服务端的结构框图。
具体实施方式
为了更好的理解上述技术方案，下面通过附图以及具体实施例对本公开内容的技术方案做详细的说明，应当理解本公开内容以及实施例中的具体特征是对本公开内容技术方案的详细的说明，而不是对本公开内容技术方案的限定，在不冲突的情况 下，本公开内容以及实施例中的技术特征可以相互组合。
针对图像进行文字识别过程中存储中间字符串的效率低的技术问题，本公开内容提供了一种图片识别方案，该方案用于在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将所述当前字符串及其哈希值存储到第一预设树形结构中；预测所述当前字符串在下一时刻的新概率值，将所述当前字符串的哈希值存储到第二预设树形结构中；通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，并将所述扩充字符串集中每个扩充字符串的概率值及哈希值存储到所述第一预设树形结构中，以及将每个扩充字符串的哈希值存储到所述第二预设树形结构中；根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，其中，N不小于1的整数；并将所述N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。
如此，在通过OCR模型对图像进行文字识别过程中，在判断某个字符串需要存储时，首选通过概率值与哈希值的映射关系，确定出需要存储的N个字符的哈希值，再根据哈希值与字符串的映射关系，根据N个字符的哈希值，确定出需要存储的N个字符；与现有技术相比，不仅不需要比较存储的字段之间的长度，也不需要在字段长度相同时逐个字符比较之后再存储，而本发明采用概率值、哈希值与字符串的映射关系，可以仅通过存储哈希值即可确定存储哪个字符串，能够极大的降低进行存储字符串所需的计算量，能够有效提高存储字符串的效率；以及在提高存储字符串的效率基础上，由于存储的字符串是用于进行文字识别的，使得OCR模型识别文字的效率也随之提高。
依据本公开内容的一些实施例，在图文识别领域，通常使用光学字符识别(Optical Character Recognition，简称OCR)模型来识别图像中的文字，以及在使用OCR模型来识别图像中的文字时，需要使用OCR模型预测出每个字的概率进行束搜索解码，如此，通过OCR模型能够预测出图像中的文字。
依据本公开内容的一些实施例，字符可以是文字、标点和符号等，本说明书不作具体限制。
如图1所示，本公开内容提供了一种图片识别方法，包括以下步骤：
S101、在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将所述当前字符串及其哈希值存储到第一预设树形结构中；
S102、预测所述当前字符串在下一时刻的新概率值，将所述当前字符串的哈希值存储到第二预设树形结构中；
S103、通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，并将所述扩充字符串集中每个扩充字符串的概率值及哈希值存储到所述第一预设树形结构中，以及将每个扩充字符串的哈希值存储到所述第二预设树形结构中；
S104、根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，其中，N不小于1的整数；并将所述N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。
依据本公开内容的一些实施例，OCR模型通常部署在图像识别装置上，图像识别装置可以由服务器提供服务，图像识别装置例如可以是词典笔和点读笔等，服务器例如可以是笔记本电脑，台式电脑和一体机等。
依据本公开内容的一些实施例，在步骤S101中，在通过OCR模型对采集的图像进行文字识别过程中，通过OCR模型预测出当前字符串，并使用哈希算法对预测出的当前字符串进行哈希，得到当前字符串的哈希值。其中，哈希算法例如可以是MD5、SHA-1和SHA-2等，本说明书不作具体限制。
依据本公开内容的一些实施例，可以在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别概率大于预设概率的字符组成的当前字符串，并获取当前字符串的哈希值，然后将当前字符串及其哈希值存储到第一预设树形结构中。
依据本公开内容的一些实施例，在通过OCR模型预测出当前字符串时，可以预先设定预设概率，再根据预设概率，对当前时刻通过OCR模型预测出图像中的每个字符进行过滤，如此，通过预设概率过滤得到当前字符串，其中，当前字符串中的每个字的概率大于预设概率；再使用哈希算法对过滤得到的当前字符串进行哈希，得 到当前字符串的哈希值，将当前字符串及其哈希值存储在第一预设树形结构中，其中，第一预设树形结构可以为C语言中set结构，也可以是二叉树结构，本说明书不作具体限制。
例如，以文本为例，在通过图像扫描装置持续扫描文本的过程中，获取到一段连续的图片集；然后使用OCR模型对图片集进行识别，在识别过程中，若当前时刻识别出的图片中的字为A1的概率为15％，为A2的概率为35％，为A3的概率为65％,为A4的概率为85％，为A5的概率为45％，而预设概率为50％，则将A1、A2和A5过滤掉，保留A3和A4，则确定当前字符串为A3和A4；若预设概率为70％，则确定当前字符串为A4；以及在获取当前字符串为A4之后，通过SHA-2对A4进行哈希，得到的哈希值H1作为A4的哈希值，然后将A4及H1存储到第一set结构中。
在将当前字符串及哈希值存储到第一预设树形结构之后，执行步骤S102，在该步骤中，预测当前字符串在下一时刻的新概率值，将当前字符串的哈希值存储到第二预设树形结构中。
依据本公开内容的一些实施例，预测出当前字符串在下一时刻的预测字符；获取当前字符串与预测字符的组合概率为新概率值，将当前字符串的哈希值存储到第二预设树形结构中，其中，第二预设树形结构可以为set结构和二叉树结构。
依据本公开内容的一些实施例，在预测出当前字符串在下一时刻的预测字符时，可以将当前字符串中的最后一个字符作为预测字符，还可以将空格作为预测字符，使得预测字符可以包括当前字符串中的最后一个字符和空格中的任意一个或多个。在预测字符为多个时，获取当前字符串与预测字符的组合概率为获取当前字符与每个预测字符的组合概率之和作为新概率值，也可以是组合概率的权重之和为新概率值，还可以是组合概率的乘积为新概率值，本说明书不作具体限制。
依据本公开内容的一些实施例，在获取新概率值之和，将当前字符串另存储到第二预设树形结构中，其中，第二预设树形结构中不存在重复的数字，相应地，第一预设树形结构中也不存在重复的数字。
例如，假设T-1时刻当前字符串为abc，则预测出当前字符串在T时刻为abc+ 空格和abc+c，通过OCR模型预测出abc+空格的概率值为H2,以及预测出abc+c的概率值为H3，则确定出abc的新概率值为H2+H3,并将abc的哈希值X存储在第二set结构中。
接下来执行步骤S103，在该步骤中，通过下一时刻识别出的目标字符集对当前字符串进行扩充，获取扩充字符串集，并将扩充字符串集中每个扩充字符串的概率值及哈希值存储到第一预设树形结构中，以及将每个扩充字符串的哈希值存储到第二预设树形结构中。
依据本公开内容的一些实施例，获取下一时刻识别出的识别概率大于预设概率的字符组成目标字符集；将当前字符串与目标字符集中每个字符进行组合，得到组合后的所有字符串为扩充字符串集。
依据本公开内容的一些实施例，通过OCR模型对下一时刻采集的图像进行文字识别，根据预设概率，对下一时刻通过OCR模型预测出图像中的每个字符进行过滤，获取到识别概率大于预设概率的所有字组词目标字符集，然后再当前字符串与目标字符集中每个字符进行组合，得到组合后的所有字符串为扩充字符串集。
例如，同样以当前字符串为abc为例，若通过OCR模型对T+1时刻采集的图像进行识别，识别出识别概率大于预设概率的字符分别为d、e和f，则获取abcd，abce和abcf作为扩充字符串集，然后通过OCR模型预测出abcd，abce和abcf的概率值依次为D1，D2和D3，并通过哈希算法对abcd，abce和abcf进行哈希计算，得到的哈希值依次为X1,X2和X3；如此，将abcd及X1，abce及X2和abcf及X3存储在第一set结构中，并将X1,X2和X3存储在第二set结构中。
在将扩充字符串集中每个扩充字符串的概率值及哈希值存储到第一预设树形结构中，以及将每个扩充字符串的哈希值存储到第二预设树形结构中之后，执行步骤S104。
在步骤S104中，可以根据当前字符串的新概率值和每个扩充字符串的概率值，从第二预设树形结构中确定概率值最高的N个哈希值；根据N个哈希值，从第一预设树形结构中确定与N个哈希值对应的N个字符串；保留N个字符串，并以保留，其中，N不小于1的整数；并将N个字符串作为当前字符串，重复上述步骤直至完 成所有采集的图像的识别，从第一预设树形结构中获取概率值最高的一个字符串作为最终识别结果。
依据本公开内容的一些实施例，N的值可以为OCR模型中的束搜索中的最大设定个数，也可以由人工或设备自行设定，例如可以为1，2，3和5等，本说明书不作具体限制。
例如，同样以当前字符串为abc为例，获取到T+1时刻的扩充字符串集为abcd，abce和abcf及对应的概率值为D1，D2和D3，以及获取abc的新概率值为H2+H3；若D2>D3>D1>(H2+H3),且N＝2；此时，可以根据概率值与哈希值的映射关系，从第二set结构中获取到N个字符串的哈希值为X2和X3；以及根据X2和X3，从第一set结构中获取到X2对应的字符串为abce，以及X3对应的字符串为abcf，则确定出N个字符串为abce和abcf；此时，将第一set结构中的abce和abcf作为当前字符串并保留，并也可以在第二set结构中仅保留abce和abcf对应的哈希值X2和X3；然后进行下一步处理。
依据本公开内容的一些实施例，首先需要预测N个字符串在T+2时刻时的新概率，具体地，预测出abce在T+2时刻时的新概率为abce+空格和abce+e的概率之和；以及预测出abcf在T+2时刻时的新概率为abcf+空格和abcf+f的概率之和；若在T+2时刻作为当前时刻时，通过OCR模式对采集的实时图片进行识别，识别出识别概率大于预设概率的字符为p和k，则获取abcep，abcek，abcfp和abcfk作为扩充字符串集，然后通过OCR模型预测出abcep，abcek，abcfp和abcfk的概率值依次为D4，D5,D6和D7，并通过哈希算法对abcep，abcek，abcfp和abcfk进行哈希计算，得到的哈希值依次为X4,X5，X6和X7；如此，将abcep及X4,，abcek及X5,abcfp及X6和abcfk及X7存储在第一set结构中，并将X4,X5，X6和X7存储在第二set结构中。
依据本公开内容的一些实施例，若D5>D4>D7>D6>abce的新概率>abcf的新概率，则根据概率值，哈希值与字符串的映射关系，确定出N个字符串为abcek和abcep，此时，将第一set结构中的abce和abcf作为当前字符串并保留，并也可以在第二set结构中仅保留abce和abcf对应的哈希值X2和X3；然后采用相同的方式进 行下一步处理，直至完成对采集的所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。
如此，本说明书实施例在识别当前字符串时，仅保留识别概率大于预设概率的字符符串，而将识别概率不大于预设概率的字符符串丢弃，从而能够整体提高OCR模型识别的效率。
以及在判断某个字符串需要存储时，首选通过概率值与哈希值的映射关系，确定出需要存储的N个字符的哈希值，再根据哈希值与字符串的映射关系，根据N个字符的哈希值，确定出需要存储的N个字符；与现有技术相比，不仅不需要比较存储的字段之间的长度，也不需要在字段长度相同时逐个字符比较之后再存储，而本发明采用概率值、哈希值与字符串的映射关系，可以仅通过存储哈希值即可确定存储哪个字符串，能够极大的降低进行存储字符串所需的计算量，能够有效提高存储字符串的效率；以及在提高存储字符串的效率基础上，由于存储的字符串是用于进行文字识别的，使得OCR模型识别文字的效率也随之提高。
如图2所示，本公开内容还提供了一种图片识别装置，包括：
字符识别单元201，用于在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将所述当前字符串及其哈希值存储到第一预设树形结构中；
预测单元202，用于预测所述当前字符串在下一时刻的新概率值，将所述当前字符串的哈希值存储到第二预设树形结构中；
扩充单元203，用于通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，并将所述扩充字符串集中每个扩充字符串的概率值及哈希值存储到所述第一预设树形结构中，以及将每个扩充字符串的哈希值存储到所述第二预设树形结构中；
循环执行单元204，用于根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，其中，N不小于1的整数；并将所述N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。
在一些实施例中，字符识别单元201，用于在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别概率大于预设概率的字符组成的所述当前字符串，并获取所述当前字符串的哈希值。
在一些实施例中，预测单元202，用于预测出所述当前字符串在下一时刻的预测字符；获取所述当前字符串与所述预测字符的组合概率为所述新概率值，将所述当前字符串的哈希值存储到所述第二预设树形结构中，其中，所述第二预设树形结构为set结构。
在一些实施例中，扩充单元203，用于获取下一时刻识别出的识别概率大于所述预设概率的字符组成所述目标字符集；将所述当前字符串与所述目标字符集中每个字符进行组合，得到组合后的所有字符串为所述扩充字符串集。
在一些实施例中，循环执行单元204，用于根据所述当前字符串的新概率值和每个扩充字符串的概率值，从所述第二预设树形结构中确定概率值最高的N个哈希值；根据所述N个哈希值，从所述第一预设树形结构中确定与所述N个哈希值对应的所述N个字符串。
在一些实施例中，所述第一预设树形结构为set结构。
图3是根据本公开内容的一种实施例提供的一种图片识别装置作为设备时的结构框图。例如，装置900可以是移动来电，计算机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等。
参照图3，装置900可以包括以下一个或多个组件：处理组件902，存储器904，电源组件906，多媒体组件908，音频组件910，输入/输出(I/O)的接口912，传感器组件914，以及通信组件916。
处理组件902通常控制装置900的整体操作，诸如与显示，来电呼叫，数据通信，相机操作和记录操作相关联的操作。处理元件902可以包括一个或多个处理器920来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件902可以包括一个或多个模块，便于处理组件902和其他组件之间的交互。例如，处理组件902可以包括多媒体模块，以方便多媒体组件908和处理组件902之间的交互。
存储器904被配置为存储各种类型的数据以支持在设备900的操作。这些数 据的示例包括用于在装置900上操作的任何应用程序或方法的指令，联系人数据，来电簿数据，消息，图片，视频等。存储器904可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。
电源组件906为装置900的各种组件提供电力。电源组件906可以包括电源管理系统，一个或多个电源，及其他与为装置900生成、管理和分配电力相关联的组件。
多媒体组件908包括在所述装置900和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动运动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件908包括一个前置摄像头和/或后置摄像头。当设备900处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件910被配置为输出和/或输入音频信号。例如，音频组件910包括一个麦克风(MIC)，当装置900处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器904或经由通信组件916发送。在一些实施例中，音频组件910还包括一个扬声器，用于输出音频信号。
I/O接口912为处理组件902和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件914包括一个或多个传感器，用于为装置900提供各个方面的状态评估。例如，传感器组件914可以检测到设备900的打开/关闭状态，组件的相对 定位，例如所述组件为装置900的显示器和小键盘，传感器组件914还可以检测装置900或装置900一个组件的位置改变，用户与装置900接触的存在或不存在，装置900方位或加速/减速和装置900的温度变化。传感器组件914可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件914还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件914还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。
通信组件916被配置为便于装置900和其他设备之间有线或无线方式的通信。装置900可以接入根据通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信部件916经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信部件916还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可根据射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。
在一些实施例中，装置900可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述方法。
在一些实施例中，还提供了一种包括指令的非临时性计算机可读存储介质，例如包括指令的存储器904，上述指令可由装置900的处理器920执行以完成上述方法。例如，所述非临时性计算机可读存储介质可以是ROM、随机存取存储器(RAM)、CD-ROM、磁带、软盘和光数据存储设备等。
图4是本公开内容的一些实施例中服务器的结构框图。该服务器1900可因配置或性能不同而产生比较大的差异，可以包括一个或一个以上中央处理器(central processing units，CPU)1922(例如，一个或一个以上处理器)和存储器1932，一个或一个以上存储应用程序1942或数据1944的存储介质1930(例如一个或一个以上海量存储设备)。其中，存储器1932和存储介质1930可以是短暂存储或持久存储。 存储在存储介质1930的程序可以包括一个或一个以上模块(图示没标出)，每个模块可以包括对服务器中的一系列指令操作。更进一步地，中央处理器1922可以设置为与存储介质1930通信，在服务器1900上执行存储介质1930中的一系列指令操作。
服务器1900还可以包括一个或一个以上电源1926，一个或一个以上有线或无线网络接口1950，一个或一个以上输入输出接口1958，一个或一个以上键盘1956，和/或，一个或一个以上操作系统1941，例如Windows ServerTM，Mac OS XTM，UnixTM,LinuxTM，FreeBSDTM等等。
一种非临时性计算机可读存储介质，当所述存储介质中的指令由装置(设备或者服务器)的处理器执行时，使得装置能够执行一种图片识别方法，所述方法包括：在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将所述当前字符串及其哈希值存储到第一预设树形结构中；预测所述当前字符串在下一时刻的新概率值，将所述当前字符串的哈希值存储到第二预设树形结构中；通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，并将所述扩充字符串集中每个扩充字符串的概率值及哈希值存储到所述第一预设树形结构中，以及将每个扩充字符串的哈希值存储到所述第二预设树形结构中；根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，其中，N不小于1的整数；并将所述N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。
本发明是参照根据本公开内容的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的设备。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以 特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令设备的制造品，该指令设备实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
尽管已描述了本发明的优选实施例，但本领域内的技术人员一旦得知了基本创造性概念，则可对这些实施例作出另外的变更和修改。所以，所附权利要求意欲解释为包括优选实施例以及落入本发明范围的所有变更和修改。
显然，本领域的技术人员可以对本发明进行各种改动和变型而不脱离本发明的精神和范围。这样，倘若本发明的这些修改和变型属于本发明权利要求及其等同技术的范围之内，则本发明也意图包含这些改动和变型在内。

Claims (14)

  1. 一种图片识别方法，包括：
    在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将所述当前字符串及其哈希值存储到第一预设树形结构中；
    预测所述当前字符串在下一时刻的新概率值，将所述当前字符串的哈希值存储到第二预设树形结构中；
    通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，并将所述扩充字符串集中每个扩充字符串的概率值及哈希值存储到所述第一预设树形结构中，以及将每个扩充字符串的哈希值存储到所述第二预设树形结构中；
    根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，其中，N不小于1的整数；并将所述N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。
  2. 如权利要求1所述的方法，其中，所述在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，包括：
    在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别概率大于预设概率的字符组成的所述当前字符串，并获取所述当前字符串的哈希值。
  3. 如权利要求2所述的方法，其中，所述预测所述当前字符串在下一时刻的新概率，将所述当前字符串的哈希值存储到第二预设树形结构中，包括：
    预测出所述当前字符串在下一时刻的预测字符；
    获取所述当前字符串与所述预测字符的组合概率为所述新概率值，将所述当前字符串的哈希值存储到所述第二预设树形结构中，其中，所述第二预设树形结构为set结构。
  4. 如权利要求1所述的方法，其中，所述通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，法包括：
    获取下一时刻识别出的识别概率大于所述预设概率的字符组成所述目标字符集；
    将所述当前字符串与所述目标字符集中每个字符进行组合，得到组合后的所有字符串为所述扩充字符串集。
  5. 如权利要求4所述的方法，其中，所述根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，包括：
    根据所述当前字符串的新概率值和每个扩充字符串的概率值，从所述第二预设树形结构中确定概率值最高的N个哈希值；
    根据所述N个哈希值，从所述第一预设树形结构中确定与所述N个哈希值对应的所述N个字符串。
  6. 如权利要求1-5任一项所述的方法，其中，所述第一预设树形结构为set结构。
  7. 一种图片识别装置，包括：
    字符识别单元，用于在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将所述当前字符串及其哈希值存储到第一预设树形结构中；
    预测单元，用于预测所述当前字符串在下一时刻的新概率值，将所述当前字符串的哈希值存储到第二预设树形结构中；
    扩充单元，用于通过下一时刻识别出的目标字符集对所述当前字符串进行扩充，获取扩充字符串集，并将所述扩充字符串集中每个扩充字符串的概率值及哈希值存储到所述第一预设树形结构中，以及将每个扩充字符串的哈希值存储到所述第二预设树形结构中；
    循环执行单元，用于根据所述第二预设树形结构中存储的哈希值，从所述第一预设树形结构中获取概率值最高的N个字符串并保留，其中，N不小于1的整数；并将所述N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。
  8. 如权利要求7所述的装置，其中，所述字符识别单元，用于在通过光学 字符识别模型对采集的图像进行文字识别过程中，获取识别概率大于预设概率的字符组成的所述当前字符串，并获取所述当前字符串的哈希值。
  9. 如权利要求8所述的装置，其中，所述预测单元，用于预测出所述当前字符串在下一时刻的预测字符；获取所述当前字符串与所述预测字符的组合概率为所述新概率值，将所述当前字符串的哈希值存储到所述第二预设树形结构中，其中，所述第二预设树形结构为set结构。
  10. 如权利要求7所述的装置，其中，所述扩充单元，用于获取下一时刻识别出的识别概率大于所述预设概率的字符组成所述目标字符集；将所述当前字符串与所述目标字符集中每个字符进行组合，得到组合后的所有字符串为所述扩充字符串集。
  11. 如权利要求10所述的装置，其中，所述循环执行单元，用于根据所述当前字符串的新概率值和每个扩充字符串的概率值，从所述第二预设树形结构中确定概率值最高的N个哈希值；根据所述N个哈希值，从所述第一预设树形结构中确定与所述N个哈希值对应的所述N个字符串。
  12. 如权利要求7-11任一项所述的装置，其中，所述第一预设树形结构为set结构。
  13. 一种用于数据处理的装置，包括有存储器，以及一个或者一个以上的程序，其中一个或者一个以上程序存储于存储器中，且经配置以由一个或者一个以上处理器执行所述一个或者一个以上程序包含如权利要求1-6任一权项所述的方法步骤。
  14. 一种机器可读介质，其上存储有指令，当由一个或多个处理器执行时，使得装置执行如权利要求1至6中一个或多个所述的方法。
PCT/CN2021/103280 2020-12-04 2021-06-29 图片识别方法、装置、电子设备及介质 WO2022116524A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/137,884 US20230290167A1 (en) 2020-12-04 2023-04-21 Picture recognition method and apparatus, electronic device, and medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011409602.4A CN114612916A (zh) 2020-12-04 2020-12-04 图片识别方法、装置、电子设备及介质
CN202011409602.4 2020-12-04

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/137,884 Continuation US20230290167A1 (en) 2020-12-04 2023-04-21 Picture recognition method and apparatus, electronic device, and medium

Publications (1)

Publication Number Publication Date
WO2022116524A1 true WO2022116524A1 (zh) 2022-06-09

Family

ID=81852925

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/103280 WO2022116524A1 (zh) 2020-12-04 2021-06-29 图片识别方法、装置、电子设备及介质

Country Status (3)

Country Link
US (1) US20230290167A1 (zh)
CN (1) CN114612916A (zh)
WO (1) WO2022116524A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140093173A1 (en) * 2012-10-01 2014-04-03 Silverbrook Research Pty Ltd Classifying a string formed from hand-written characters
CN104156706A (zh) * 2014-08-12 2014-11-19 华北电力大学句容研究中心 一种基于光学字符识别技术的中文字符识别方法
CN109766893A (zh) * 2019-01-09 2019-05-17 北京数衍科技有限公司 适于购物小票的图片文字识别方法
CN111079768A (zh) * 2019-12-23 2020-04-28 北京爱医生智慧医疗科技有限公司 一种基于ocr的文字图像识别方法及装置
CN111914513A (zh) * 2019-05-08 2020-11-10 亿阳安全技术有限公司 一种rdp窗口标题文字识别的方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140093173A1 (en) * 2012-10-01 2014-04-03 Silverbrook Research Pty Ltd Classifying a string formed from hand-written characters
CN104156706A (zh) * 2014-08-12 2014-11-19 华北电力大学句容研究中心 一种基于光学字符识别技术的中文字符识别方法
CN109766893A (zh) * 2019-01-09 2019-05-17 北京数衍科技有限公司 适于购物小票的图片文字识别方法
CN111914513A (zh) * 2019-05-08 2020-11-10 亿阳安全技术有限公司 一种rdp窗口标题文字识别的方法及装置
CN111079768A (zh) * 2019-12-23 2020-04-28 北京爱医生智慧医疗科技有限公司 一种基于ocr的文字图像识别方法及装置

Also Published As

Publication number Publication date
CN114612916A (zh) 2022-06-10
US20230290167A1 (en) 2023-09-14

Similar Documents

Publication Publication Date Title
JP7153088B2 (ja) テキスト認識方法及びテキスト認識装置、電子機器、記憶媒体並びにコンピュータプログラム
WO2020029966A1 (zh) 视频处理方法及装置、电子设备和存储介质
WO2017114020A1 (zh) 语音输入方法和终端设备
WO2021128880A1 (zh) 一种语音识别方法、装置和用于语音识别的装置
US11455491B2 (en) Method and device for training image recognition model, and storage medium
WO2019165832A1 (zh) 文字信息处理方法、装置及终端
WO2021208666A1 (zh) 字符识别方法及装置、电子设备和存储介质
WO2017088247A1 (zh) 输入处理方法、装置及设备
WO2018090688A1 (zh) 一种输入方法、装置及电子设备
US20230386449A1 (en) Method and apparatus for training neural network, and method and apparatus for audio processing
CN109783244B (zh) 处理方法和装置、用于处理的装置
WO2022116524A1 (zh) 图片识别方法、装置、电子设备及介质
CN109901726B (zh) 一种候选词生成方法、装置及用于候选词生成的装置
CN109725736B (zh) 一种候选排序方法、装置及电子设备
CN109144286B (zh) 一种输入方法及装置
WO2019095682A1 (zh) 一种候选项联想方法和装置
CN110908523A (zh) 一种输入方法及装置
CN112987941B (zh) 生成候选词的方法及装置
WO2022198821A1 (zh) 人脸和人体匹配的方法、装置、电子设备、存储介质及程序
CN114168809A (zh) 基于相似度的文档字符串编码匹配方法及装置
CN109917927B (zh) 一种候选项确定方法和装置
CN107992893B (zh) 压缩图像特征空间的方法及装置
CN107977089B (zh) 一种输入方法和装置、一种用于输入的装置
WO2019214234A1 (zh) 一种输入预测方法及装置
CN111522448A (zh) 一种提供输入候选项的方法、装置和设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21899566

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21899566

Country of ref document: EP

Kind code of ref document: A1