WO2018090641A1 - Procédé, appareil et dispositif d'identification de numéro de police d'assurance, et support d'informations lisible par ordinateur - Google Patents

Procédé, appareil et dispositif d'identification de numéro de police d'assurance, et support d'informations lisible par ordinateur Download PDF

Info

Publication number
WO2018090641A1
WO2018090641A1 PCT/CN2017/091308 CN2017091308W WO2018090641A1 WO 2018090641 A1 WO2018090641 A1 WO 2018090641A1 CN 2017091308 W CN2017091308 W CN 2017091308W WO 2018090641 A1 WO2018090641 A1 WO 2018090641A1
Authority
WO
WIPO (PCT)
Prior art keywords
insurance policy
picture
insurance
sample
training
Prior art date
Application number
PCT/CN2017/091308
Other languages
English (en)
Chinese (zh)
Inventor
马进
王健宗
肖京
Original Assignee
平安科技（深圳）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 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2018090641A1 publication Critical patent/WO2018090641A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
    • G06V10/23—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition based on positionally close patterns or neighbourhood relationships
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/40—Document-oriented image-based pattern recognition
    • G06V30/41—Analysis of document content
    • G06V30/412—Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition

Definitions

  • the present invention relates to the field of image processing technologies, and in particular, to a method, an apparatus, a device, and a computer readable storage medium for identifying an insurance policy number.
  • Each insurance policy or insurance policy picture has a unique insurance policy number corresponding to it.
  • the insurance policy number is the key information, and the staff generally needs to perform the operation of searching or inquiring the insurance information according to the insurance policy number. If you want to get the picture information such as the insurance policy number from multiple insurance policy pictures, it is usually obtained by manual operation. When the number of insurance policy pictures is large, the staff cannot quickly get each insurance policy. The insurance policy number of the picture results in a very large workload and reduces work efficiency.
  • the object of the present invention is to provide a method, device, device and computer readable storage medium for identifying an insurance policy number, which aims to quickly obtain an insurance policy number from a large number of insurance policy pictures, reduce workload and improve work efficiency.
  • the present invention provides a method for identifying an insurance policy number, and the method for identifying an insurance policy number includes:
  • the present invention also provides an apparatus for identifying an insurance policy number, and the apparatus for identifying an insurance policy number includes:
  • a first extraction module configured to identify an insurance type corresponding to the insurance policy picture after receiving the insurance policy picture, and extract the insurance policy based on a predetermined insurance type and a positional relationship of the insurance policy number in the insurance policy picture The number is in the corresponding target line character area in the insurance policy picture;
  • a first identification module configured to invoke a first recognition model generated by the pre-training to perform character recognition on the target line character region, to identify an insurance policy number included in the target line character region, And identifying the insurance policy number and storing the insurance policy picture.
  • the present invention also provides an apparatus for identifying an insurance policy number, the apparatus comprising the memory, a processor, and an identification insurance number stored on the memory and operable on the processor
  • the program when the program for identifying the insurance policy number is executed by the processor, implements the following steps:
  • the present invention also provides a computer readable storage medium having stored thereon a program for identifying an insurance policy number, wherein the program for identifying an insurance policy number is executed by a processor to implement the following steps :
  • the invention has the beneficial effects that the invention firstly identifies the insurance type of the insurance policy picture, and through the positional relationship between the insurance type and the insurance policy number in the insurance policy picture, the target line character area corresponding to the insurance policy number can be extracted, and then called.
  • the first recognition model generated by the pre-training is used to identify the insurance policy number in the target line character area, and the entire operation process requires almost no manual participation, and the insurance policy number can be quickly obtained from a large number of insurance policy pictures, thereby greatly reducing the workload. ,Improve work efficiency.
  • FIG. 1 is a schematic flow chart of a first embodiment of a method for identifying an insurance policy number according to the present invention
  • FIG. 2 is a schematic flow chart of a second embodiment of a method for identifying an insurance policy number according to the present invention
  • FIG. 3 is a schematic flow chart of a third embodiment of a method for identifying an insurance policy number according to the present invention.
  • FIG. 4 is a schematic flow chart of a fourth embodiment of a method for identifying an insurance policy number according to the present invention.
  • FIG. 5 is a schematic flowchart diagram of a fifth embodiment of a method for identifying an insurance policy number according to the present invention.
  • FIG. 6 is a schematic structural diagram of a first embodiment of an apparatus for identifying an insurance policy number according to the present invention.
  • FIG. 7 is a schematic structural diagram of a second embodiment of an apparatus for identifying an insurance policy number according to the present invention.
  • FIG. 8 is a schematic structural diagram of a third embodiment of an apparatus for identifying an insurance policy number according to the present invention.
  • FIG. 9 is a block diagram showing the hardware structure of an apparatus for identifying an insurance policy number according to the present invention.
  • FIG. 1 is a schematic flowchart of an embodiment of a method for identifying an insurance policy number according to the present invention.
  • the method for identifying an insurance policy number includes the following steps:
  • Step S1 after receiving the insurance policy picture, identifying the insurance type corresponding to the insurance policy picture, and extracting the insurance policy number based on the predetermined insurance type and the positional relationship of the insurance policy number in the insurance policy picture.
  • each insurance policy is an insurance type.
  • the insurance policy numbers are not in the same position. For example, some insurance policy numbers are located above the upper right corner of the insurance policy, and some insurance policy numbers are located to the left of the upper right corner of the insurance policy.
  • different types of insurance policies are stored in association with the location of the insurance policy number.
  • the insurance type of the insurance policy image is first identified. The specific identification process is: The size, color and content layout of the single are comprehensively identified to determine the type of insurance to which the insurance policy picture belongs.
  • the insurance type to which the insurance policy picture belongs may be identified by other methods, for example, by identifying the content of the insurance picture. Information to determine the type of insurance to which it belongs.
  • Step S2 calling the first recognition model generated by the pre-training to perform character recognition on the target line character region to identify the insurance policy number included in the target line character region, and identify the insurance policy number and the insurance A single picture is stored in association.
  • the first recognition model is generated in advance, and the first recognition model may be one of a plurality of models related to image processing.
  • the first recognition model is a time recurrent neural network model. Calling the first recognition model to perform character recognition on the target line character region to identify each character in the target line character region.
  • the insurance policy number is a number, and when all the characters are recognized, an insurance policy can be obtained. number.
  • the insurance policy number is stored in association with the insurance policy picture, so that when the employee searches or retrieves the insurance policy number, the insurance policy number can be used to query or retrieve the associated insurance number. Insurance policy picture.
  • the embodiment first identifies the insurance type of the insurance policy picture, and by using the positional relationship between the insurance type and the insurance policy number in the insurance policy picture, the target line character area corresponding to the insurance policy number can be extracted, and then Calling the first recognition model generated by the pre-training to identify the insurance policy number in the target line character area, the whole operation process requires almost no human participation, and can quickly obtain the insurance policy number from a large number of insurance policy pictures, thereby greatly reducing the work. Quantity, improve work efficiency.
  • step S1 is replaced by:
  • step S0 after receiving the insurance policy picture, the second recognition model generated by the pre-training is invoked to identify the target line character area in which the insurance policy number is located in the insurance policy picture.
  • the second recognition model is generated in advance, and the second recognition model may be one of a plurality of models related to image processing.
  • the second recognition model is a convolutional neural network model.
  • the second identification model is called to locate and identify the insurance policy picture to identify the target line character area where the insurance policy number is located.
  • the present embodiment identifies the target line character area where the insurance policy number is located in the insurance policy picture by calling the second identification model. Since the second recognition model is obtained by training a large amount of data, it is possible to Accurately identify the target line character area.
  • the method further includes:
  • Step S01 obtaining a preset number of insurance policy sample pictures, using the insurance policy sample picture including the insurance policy number as the first picture set, and using the insurance policy sample picture not including the insurance policy number as the second picture set;
  • Step S02 extracting, from the first picture set and the second picture set, a first preset proportion of the insurance order sample picture as the sample picture to be trained, and collecting the remaining insurance policies in the first picture set and the second picture set.
  • the sample image is used as a sample image to be verified;
  • Step S03 performing model training using each sample picture to be trained to generate the convolutional neural network model, and verifying the generated convolutional neural network model by using each sample picture to be verified;
  • step S04 if the verification pass rate is greater than or equal to the preset threshold, the training is completed, otherwise the number of the insurance ticket sample pictures is increased to re-train and verify.
  • the second recognition model is a convolutional neural network model. Before using the convolutional neural network model to identify the target line character region, the convolutional neural network model is first trained to generate:
  • Obtain a preset number of insurance policy sample images for example, obtain 100,000 insurance policy sample images, where the insurance policy sample image including the insurance policy number is used as the first photo collection, and the insurance policy sample image not including the insurance policy number As a second collection of pictures.
  • the first picture set and the second picture set respectively extract the first preset proportion of the insurance order sample picture as the sample picture to be trained, for example, the first picture set and the second picture set respectively extract 80% of the insurance sample picture as the to-be-trained
  • the sample picture, the first picture set and the remaining insurance order sample pictures in the second picture set are used as sample pictures to be verified.
  • the parameters of the convolutional neural network model are trained by default parameters, and the parameters are continuously adjusted during the training process.
  • each of the to-be-verified models is used.
  • the sample picture is used to verify the generated convolutional neural network model. If the verification pass rate is greater than or equal to a preset threshold, for example, the pass rate is greater than or equal to 98%, the training ends, and the convolutional neural network model obtained by the training is used to identify the target.
  • the model of the line character area if If the verification pass rate is less than the preset threshold, for example, less than 98%, the number of the insurance ticket sample pictures is increased, and the above steps S01, S02, S03, and S04 are performed again until the verification pass rate is greater than or equal to the preset threshold.
  • the preset threshold for example, less than 98%
  • the method further includes:
  • S21 Obtain a preset number of insurance ticket number sample images, extract a second preset proportion of insurance policy number sample images as a training set, and test the remaining insurance policy number sample images in the preset number of insurance policy number sample images as a test set;
  • S22 Input the insurance ticket number sample image in the training set to the time recurrent neural network model for model training, and use the insurance ticket number sample image in the test set to perform the trained time recurrent neural network model every preset time. Testing to assess the recognition effect of the trained time recurrent neural network model;
  • the first recognition model is a time recurrent neural network model
  • the time recurrent neural network model is first trained to identify the characters in the target line character region by using the time recurrent neural network model:
  • Obtain a preset number of insurance ticket number sample images for example, obtain 100,000 insurance policy number sample images, wherein the insurance policy number sample image only contains one line of numbers, the line number is the insurance policy number, the font is black, and the background is white.
  • the name of each insurance policy number sample picture can be named as the insurance policy number included.
  • the number of insurance ticket number sample pictures in the test set for example, 80% of the insurance order number sample pictures in the insurance order number sample picture is used as the training set, and the remaining 20% of the insurance number number sample pictures are used as the test set.
  • the parameters of the time recurrent neural network model are trained by default parameters, and the insurance order number sample images in the training set are input into the time recurrent neural network model for training,
  • the preset time uses the insurance policy number sample picture in the test set to test the trained time recurrent neural network model. For example, the training set is tested with the test set after every 1000 iterations to evaluate the trained time recurrent neural network model. Identify the effect.
  • the trained model is used to identify the insurance policy number of the insurance policy number sample image in the test set, and compare the identification result with the used name of the insurance ticket number sample picture (the insurance policy number sample picture utilizes the The insurance policy number is named to evaluate the recognition effect of the trained time recurrent neural network model.
  • the recognition error of the trained time recurrent neural network model is calculated, and the identification error is the identified insurance policy number and the name of the insurance ticket number sample picture.
  • the edit distance of the insurance policy number used if the recognition error converges, the training is completed, and the trained time recurrent neural network model is used as a model for identifying characters in the target line character region; if the recognition error is divergent, the time recursive nerve is adjusted.
  • the model parameters of the network model are re-executed in steps S21, S22, and S23 described above until the recognition error converges.
  • the method further includes:
  • the user when searching or viewing the information in the insurance policy, the user first sends a search request carrying the insurance policy number to the device where the insurance policy number is located, and after receiving the search request, the device according to the search request The insurance policy number matches the stored insurance policy number. After matching the consistent insurance policy number, the insurance policy image associated with the matched insurance policy number is fed back to the terminal, so that the terminal user can view the insurance policy picture. details.
  • FIG. 6 is a schematic structural diagram of an apparatus for identifying an insurance policy number according to an embodiment of the present invention.
  • the apparatus for identifying an insurance policy number includes:
  • the first extraction module 101 is configured to identify an insurance type corresponding to the insurance policy picture after receiving the insurance policy picture, and extract the insurance based on a predetermined insurance type and a positional relationship of the insurance policy number in the insurance policy picture.
  • the single number is in the corresponding target line character area in the insurance policy picture;
  • each insurance policy is an insurance type.
  • the insurance policy numbers are not in the same position. For example, some insurance policy numbers are located above the upper right corner of the insurance policy, and some insurance policy numbers are located to the left of the upper right corner of the insurance policy.
  • different types of insurance policies are stored in association with the location of the insurance policy number.
  • the insurance type of the insurance policy image is first identified. The specific identification process is: The size, color and content layout of the single are comprehensively identified to determine the type of insurance to which the insurance policy picture belongs.
  • the insurance type to which the insurance policy picture belongs may be identified by other methods, for example, by identifying the content of the insurance picture. Information to determine the type of insurance to which it belongs.
  • the first identification module 102 is configured to call a first recognition model generated by the pre-training to perform character recognition on the target line character region to identify an insurance policy number included in the target line character region, and identify an insurance policy The number is stored in association with the insurance policy picture.
  • the first recognition model is generated in advance, and the first recognition model may be one of a plurality of models related to image processing.
  • the first recognition model is a time recurrent neural network module. type. Calling the first recognition model to perform character recognition on the target line character region to identify each character in the target line character region.
  • the insurance policy number is a number, and when all the characters are recognized, an insurance policy can be obtained. number.
  • the insurance policy number is stored in association with the insurance policy picture, so that when the employee searches or retrieves the insurance policy number, the insurance policy number can be used to query or retrieve the associated insurance number. Insurance policy picture.
  • the first extraction module 101 is replaced with: a second identification module 100, after receiving the insurance policy picture, The second recognition model generated by the pre-training is invoked to identify the target line character region in which the insurance policy number is located in the insurance policy picture.
  • the second recognition model is generated in advance, and the second recognition model may be one of a plurality of models related to image processing.
  • the second recognition model is a convolutional neural network model.
  • the second identification model is called to locate and identify the insurance policy picture to identify the target line character area where the insurance policy number is located.
  • the present embodiment identifies the target line character area where the insurance policy number is located in the insurance policy picture by calling the second identification model. Since the second recognition model is obtained by training a large amount of data, it is possible to Accurately identify the target line character area.
  • the second identification model is a convolutional neural network model
  • the insurance number identification device further includes:
  • An acquiring module configured to obtain a preset number of insurance ticket sample pictures, use an insurance policy sample picture including an insurance policy number as a first picture set, and use an insurance policy sample picture that does not include an insurance policy number as a second picture set;
  • a second extraction module configured to separately extract a first preset ratio of the insurance ticket sample image from the first image set and the second image set as a sample image to be trained, and focus the first image set and the second image set The remaining insurance policy sample image is used as the sample image to be verified;
  • a first training module configured to perform model training by using each sample image to be trained, to generate the convolutional neural network model, and verify the generated convolutional neural network model by using each sample image to be verified;
  • the first processing module is configured to: if the verification pass rate is greater than or equal to the preset threshold, the training is completed, otherwise the number of the insurance ticket sample pictures is increased to perform training and verification again.
  • the second recognition model is a convolutional neural network model. Before using the convolutional neural network model to identify the target line character region, the convolutional neural network model is first trained to generate:
  • Obtain a preset number of insurance policy sample images for example, obtain 100,000 insurance policy sample images, where the insurance policy sample image including the insurance policy number is used as the first photo collection, and the insurance policy sample image not including the insurance policy number As a second collection of pictures.
  • the first picture set and the second picture set respectively extract the first preset proportion of the insurance order sample picture as the sample picture to be trained, for example, the first picture set and the second picture set respectively extract 80% of the insurance sample picture as the to-be-trained Sample map
  • the slice, the first picture set and the remaining insurance order sample pictures in the second picture set are used as sample pictures to be verified.
  • the parameters of the convolutional neural network model are trained by default parameters, and the parameters are continuously adjusted during the training process.
  • each of the to-be-verified models is used.
  • the sample picture is used to verify the generated convolutional neural network model. If the verification pass rate is greater than or equal to a preset threshold, for example, the pass rate is greater than or equal to 98%, the training ends, and the convolutional neural network model obtained by the training is used to identify the target.
  • the identification device of the insurance policy number further includes:
  • the third extraction module is configured to obtain a preset number of insurance ticket number sample images, extract a second preset ratio insurance ticket number sample image as a training set, and replace the remaining insurance policy in the preset number of insurance policy number sample images
  • the number sample picture is used as a test set;
  • a second training module configured to input the insurance ticket number sample picture in the training set to a time recurrent neural network model for model training, and use the insurance ticket number sample picture in the test set to train the time every preset time
  • the recurrent neural network model is tested to evaluate the recognition effect of the trained time recurrent neural network model
  • a second processing module configured to calculate a recognition error of the trained time recurrent neural network model after each test, if the recognition error converges, the training is completed, otherwise the model parameters of the time recurrent neural network model are adjusted, To re-train and test.
  • the first recognition model is a time recurrent neural network model
  • the time recurrent neural network model is first trained to identify the characters in the target line character region by using the time recurrent neural network model:
  • Obtain a preset number of insurance ticket number sample images for example, obtain 100,000 insurance policy number sample images, wherein the insurance policy number sample image only contains one line of numbers, the line number is the insurance policy number, the font is black, and the background is white.
  • the name of each insurance policy number sample picture can be named as the insurance policy number included.
  • the number of insurance ticket number sample pictures in the test set for example, 80% of the insurance order number sample pictures in the insurance order number sample picture is used as the training set, and the remaining 20% of the insurance number number sample pictures are used as the test set.
  • the parameters of the time recurrent neural network model are trained by default parameters, and the insurance order number sample images in the training set are input into the time recurrent neural network model for training,
  • the preset time uses the insurance policy number sample picture in the test set to test the trained time recurrent neural network model. For example, the training set is tested with the test set after every 1000 iterations to evaluate the trained time recurrent neural network model. Identify the effect.
  • the trained model is used to identify the insurance policy number of the insurance policy number sample image in the test set, and compare the identification result with the used name of the insurance ticket number sample picture (the insurance policy number sample picture utilizes the The insurance policy number is named to evaluate the recognition effect of the trained time recurrent neural network model.
  • the recognition error of the trained time recurrent neural network model is calculated, and the identification error is an edit of the identified insurance policy number and the insurance policy number used for naming the insurance ticket number sample picture.
  • Distance if the recognition error converges, the training is completed, and the trained time recurrent neural network model is used as a model for identifying characters in the target line character region; if the recognition error is diverged, the model parameters of the time recurrent neural network model are adjusted until identification The error converges.
  • the identification device for the insurance policy number further includes:
  • the search module is configured to: after receiving the search request for carrying the insurance policy number issued by the terminal, search for an insurance policy picture associated with the insurance policy number, and send the found insurance policy picture to the terminal.
  • the user when searching or viewing the information in the insurance policy, the user first sends a search request carrying the insurance policy number to the device where the insurance policy number is located, and after receiving the search request, the device according to the search request The insurance policy number matches the stored insurance policy number. After matching the consistent insurance policy number, the insurance policy image associated with the matched insurance policy number is fed back to the terminal, so that the terminal user can view the insurance policy picture. details.
  • FIG. 9 a block diagram of a hardware structure of an apparatus for identifying an insurance policy number according to the present invention is shown.
  • the device for identifying the insurance policy number may be a PC (Personal Computer), or may be a portable terminal device having a display function such as a smart phone, a tablet computer, an e-book reader, or a portable computer.
  • PC Personal Computer
  • portable terminal device having a display function such as a smart phone, a tablet computer, an e-book reader, or a portable computer.
  • the device for identifying the policy number includes a memory 11, a processor 12, a communication bus 13, and a network interface 14.
  • the memory 11 includes at least one type of readable storage medium including a flash memory, a hard disk, a multimedia card, a card type memory (for example, an SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, and the like.
  • the memory 11 may in some embodiments be an internal storage unit of the device that identifies the insurance policy number, such as the hard disk of the device that identifies the insurance policy number.
  • the memory 11 may also be an external storage device of the device that identifies the insurance policy number in other embodiments, such as a plug-in hard disk equipped on a device that identifies the insurance policy number, a smart memory card (SMC), a secure digital number. (Secure Digital, SD) card, flash card, etc.
  • the memory 11 may also include both an internal storage unit of the device that identifies the insurance policy number and an external storage device.
  • the memory 11 can be used not only for storing application software and various types of data installed in a device for identifying an insurance policy number, such as a code for a program for identifying an insurance policy number, but also for temporarily storing data that has been output or is to be output.
  • the processor 12 may be a central processing unit (Central Processing Unit, in some embodiments).
  • Communication bus 13 is used to implement connection communication between these components.
  • the network interface 14 can optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), and is typically used to establish a communication connection between the device and other electronic devices.
  • a standard wired interface such as a WI-FI interface
  • Figure 9 shows only the device for identifying the policy number with the components 11-14 and the program identifying the policy number, but it should be understood that not all of the illustrated components are required to be implemented, alternative implementations may be more or more Less components.
  • the device may further include a user interface
  • the user interface may include a display
  • an input unit such as a keyboard
  • the optional user interface may further include a standard wired interface and a wireless interface.
  • the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch sensor, or the like.
  • the display may also be referred to as a display or display unit, as appropriate, for displaying information processed in the device identifying the insurance policy number and a user interface for displaying the visualization.
  • a program for identifying an insurance policy number is stored in the memory 11, the network interface 14 is mainly used to connect to a server, and performs data communication with the server; the processor 12 executes the storage in the memory 11.
  • the procedure for identifying the policy number is as follows:
  • step S2 can be replaced by:
  • the second recognition model generated by the pre-training is invoked to identify the target line character area in which the insurance policy number is located in the insurance policy picture.
  • the second identification model is a convolutional neural network model
  • the processor is further configured to execute the program for identifying the insurance policy number, so as to implement the following steps before step S0:
  • S01 Obtain a preset number of insurance policy sample pictures, use an insurance policy sample picture including an insurance policy number as a first picture set, and use an insurance policy sample picture that does not include an insurance policy number as a second picture set;
  • S03 performing model training by using each sample image to be trained to generate the convolutional neural network model, and verifying the generated convolutional neural network model by using each sample image to be verified;
  • verification pass rate is greater than or equal to the preset threshold, the training is completed, otherwise the protection is increased.
  • the first recognition model is a time recurrent neural network model
  • the processor is further configured to execute the program for identifying the insurance policy number, so as to implement the following steps before step S2:
  • S21 Obtain a preset number of insurance ticket number sample images, extract a second preset proportion of insurance policy number sample images as a training set, and test the remaining insurance policy number sample images in the preset number of insurance policy number sample images as a test set;
  • S22 Input the insurance ticket number sample image in the training set to the time recurrent neural network model for model training, and use the insurance ticket number sample image in the test set to perform the trained time recurrent neural network model every preset time. Testing to assess the recognition effect of the trained time recurrent neural network model;
  • processor is further configured to execute the program for identifying the insurance policy number, so after step S2, further implementing the following steps:
  • the present invention also provides a computer readable storage medium having stored thereon a program for identifying an insurance policy number, the program for identifying the insurance policy number being executed by the processor to:
  • step S2 can be replaced by:
  • the second recognition model generated by the pre-training is invoked to identify the target line character area in which the insurance policy number is located in the insurance policy picture.
  • step S0 when the program for identifying the insurance policy number is executed by the processor, the following steps are further implemented before step S0:
  • S01 Obtain a preset number of insurance policy sample pictures, use an insurance policy sample picture including an insurance policy number as a first picture set, and use an insurance policy sample picture that does not include an insurance policy number as a second picture set;
  • step S2 when the program for identifying the insurance policy number is executed by the processor, the following steps are further implemented before step S2:
  • S21 Obtain a preset number of insurance ticket number sample images, extract a second preset proportion of insurance policy number sample images as a training set, and test the remaining insurance policy number sample images in the preset number of insurance policy number sample images as a test set;
  • S22 Input the insurance ticket number sample image in the training set to the time recurrent neural network model for model training, and use the insurance ticket number sample image in the test set to perform the trained time recurrent neural network model every preset time. Testing to assess the recognition effect of the trained time recurrent neural network model;
  • step S2 when the program for identifying the insurance policy number is executed by the processor, the following steps are further implemented after step S2:
  • the specific embodiment of the computer readable storage medium of the present invention is substantially the same as the method embodiment for identifying the insurance policy number, and is not described herein.
  • the technical solution of the present invention which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a storage medium (such as ROM/RAM, disk,
  • a storage medium such as ROM/RAM, disk,
  • the optical disc includes a number of instructions for causing a terminal device (which may be a cell phone, a computer, a server, or a network device, etc.) to perform the methods described in various embodiments of the present invention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)

Abstract

L'invention concerne un procédé, un appareil et un dispositif d'identification d'un numéro de police d'assurance, et un support d'informations lisible par ordinateur. Le procédé d'identification d'un numéro de police d'assurance comprend les étapes suivantes : après la réception d'une image de police d'assurance, l'identification d'un type d'assurance correspondant à l'image de police d'assurance, et l'extraction d'une région de caractère de ligne cible correspondant au numéro de police d'assurance dans l'image de police d'assurance sur la base d'une relation de position prédéterminée entre un type d'assurance et un numéro de police d'assurance dans l'image de police d'assurance (S1) ; et à invoquer un premier modèle d'identification qui est généré par apprentissage à l'avance pour effectuer une identification de caractère sur la région de caractère de rangée cible, de façon à identifier un numéro de police d'assurance inclus dans la région de caractère de ligne cible, et à stocker le numéro de police d'assurance identifié et l'image de police d'assurance d'une manière associée (S2). Au cours de l'ensemble du processus de fonctionnement, presque aucune intervention manuelle n'est nécessaire, et un numéro de police d'assurance peut être rapidement acquis à partir d'un lot d'images de police d'assurance, de telle sorte que la charge de travail est sensiblement réduite et l'efficacité de travail est améliorée.
PCT/CN2017/091308 2016-11-15 2017-06-30 Procédé, appareil et dispositif d'identification de numéro de police d'assurance, et support d'informations lisible par ordinateur WO2018090641A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201611005112.1A CN106557747B (zh) 2016-11-15 2016-11-15 识别保险单号码的方法及装置
CN201611005112.1 2016-11-15

Publications (1)

Publication Number Publication Date
WO2018090641A1 true WO2018090641A1 (fr) 2018-05-24

Family

ID=58444147

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/091308 WO2018090641A1 (fr) 2016-11-15 2017-06-30 Procédé, appareil et dispositif d'identification de numéro de police d'assurance, et support d'informations lisible par ordinateur

Country Status (2)

Country Link
CN (1) CN106557747B (fr)
WO (1) WO2018090641A1 (fr)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110231939A (zh) * 2019-05-16 2019-09-13 平安科技（深圳）有限公司 模型生成方法、系统、计算机设备和存储介质
CN111275039A (zh) * 2020-01-17 2020-06-12 深圳信息职业技术学院 水尺字符定位方法、装置、计算设备及存储介质
CN111382297A (zh) * 2018-12-29 2020-07-07 杭州海康存储科技有限公司 一种用户侧用户数据的上报方法及装置

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106557747B (zh) * 2016-11-15 2018-06-22 平安科技（深圳）有限公司 识别保险单号码的方法及装置
CN107220648B (zh) * 2017-04-11 2018-06-22 平安科技（深圳）有限公司 理赔单据的字符识别方法及服务器
CN107766809B (zh) * 2017-10-09 2020-05-19 平安科技（深圳）有限公司 电子装置、票据信息识别方法和计算机可读存储介质
CN108564035B (zh) 2018-04-13 2020-09-25 杭州睿琪软件有限公司 识别单据上记载的信息的方法及系统
CN110619252B (zh) * 2018-06-19 2022-11-04 百度在线网络技术（北京）有限公司 识别图片中表单数据的方法、装置、设备及存储介质
CN109918984A (zh) * 2018-12-15 2019-06-21 深圳壹账通智能科技有限公司 保险单号码识别方法、装置、电子设备及存储介质
CN109829444A (zh) * 2019-02-28 2019-05-31 广州达安临床检验中心有限公司 单据录入方法、装置、计算机设备和存储介质
CN109903174B (zh) * 2019-03-22 2023-11-24 成都肯定科技有限公司 一种基于移动终端摄像的保险单录入系统及方法
CN110110726A (zh) * 2019-05-15 2019-08-09 深圳供电局有限公司 电力设备铭牌识别方法、装置、计算机设备和存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120197826A1 (en) * 2011-01-28 2012-08-02 Fujitsu Limited Information matching apparatus, method of matching information, and computer readable storage medium having stored information matching program
CN105005793A (zh) * 2015-07-15 2015-10-28 广州敦和信息技术有限公司 一种发票字条自动识别录入的方法及装置
CN105426356A (zh) * 2015-10-29 2016-03-23 杭州九言科技股份有限公司 一种目标信息识别方法和装置
CN105825211A (zh) * 2016-03-17 2016-08-03 世纪龙信息网络有限责任公司 名片识别方法、装置及系统
CN106557747A (zh) * 2016-11-15 2017-04-05 平安科技（深圳）有限公司 识别保险单号码的方法及装置

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8150161B2 (en) * 2008-09-22 2012-04-03 Intuit Inc. Technique for correcting character-recognition errors
CN102567764B (zh) * 2012-01-13 2016-03-02 中国工商银行股份有限公司 一种提高电子影像识别效率的票据凭证及系统
CN103793846A (zh) * 2014-01-20 2014-05-14 中国建设银行股份有限公司 业务凭证的要素信息的处理方法及装置
CN105095842B (zh) * 2014-05-22 2018-12-11 口碑控股有限公司 一种单据的信息识别的方法和装置
CN104077577A (zh) * 2014-07-03 2014-10-01 浙江大学 一种基于卷积神经网络的商标检测方法
CN104298976B (zh) * 2014-10-16 2017-09-26 电子科技大学 基于卷积神经网络的车牌检测方法
CN204576535U (zh) * 2014-12-22 2015-08-19 深圳中兴网信科技有限公司 一种票据识别装置
CN105184312B (zh) * 2015-08-24 2018-09-25 中国科学院自动化研究所 一种基于深度学习的文字检测方法及装置
CN105678612A (zh) * 2015-12-30 2016-06-15 远光软件股份有限公司 移动端原始凭证电子化智能填单系统及方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120197826A1 (en) * 2011-01-28 2012-08-02 Fujitsu Limited Information matching apparatus, method of matching information, and computer readable storage medium having stored information matching program
CN105005793A (zh) * 2015-07-15 2015-10-28 广州敦和信息技术有限公司 一种发票字条自动识别录入的方法及装置
CN105426356A (zh) * 2015-10-29 2016-03-23 杭州九言科技股份有限公司 一种目标信息识别方法和装置
CN105825211A (zh) * 2016-03-17 2016-08-03 世纪龙信息网络有限责任公司 名片识别方法、装置及系统
CN106557747A (zh) * 2016-11-15 2017-04-05 平安科技（深圳）有限公司 识别保险单号码的方法及装置

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111382297A (zh) * 2018-12-29 2020-07-07 杭州海康存储科技有限公司 一种用户侧用户数据的上报方法及装置
CN111382297B (zh) * 2018-12-29 2024-05-17 杭州海康存储科技有限公司 一种用户侧用户数据的上报方法及装置
CN110231939A (zh) * 2019-05-16 2019-09-13 平安科技（深圳）有限公司 模型生成方法、系统、计算机设备和存储介质
CN110231939B (zh) * 2019-05-16 2024-05-28 平安科技（深圳）有限公司 模型生成方法、系统、计算机设备和存储介质
CN111275039A (zh) * 2020-01-17 2020-06-12 深圳信息职业技术学院 水尺字符定位方法、装置、计算设备及存储介质
CN111275039B (zh) * 2020-01-17 2023-05-16 深圳信息职业技术学院 水尺字符定位方法、装置、计算设备及存储介质

Also Published As

Publication number Publication date
CN106557747B (zh) 2018-06-22
CN106557747A (zh) 2017-04-05

Similar Documents

Publication Publication Date Title
WO2018090641A1 (fr) Procédé, appareil et dispositif d'identification de numéro de police d'assurance, et support d'informations lisible par ordinateur
JP6710483B2 (ja) 損害賠償請求書類の文字認識方法、装置、サーバ及び記憶媒体
WO2019174130A1 (fr) Procédé de reconnaissance de facture, serveur et support de stockage lisible par ordinateur
WO2019109526A1 (fr) Procédé et dispositif de reconnaissance de l'âge de l'image d'un visage, et support de stockage
WO2019071662A1 (fr) Dispositif électronique, procédé d'identification d'informations de facture et support d'informations lisible par ordinateur
WO2020134991A1 (fr) Procédé de saisie automatique pour formulaire papier, appareil et dispositif informatique et support d'informations
WO2019196303A1 (fr) Procédé d'authentification d'identité d'utilisateur, serveur et support de stockage
WO2019062080A1 (fr) Procédé de reconnaissance d'identité, dispositif électronique et support d'informations lisible par ordinateur
WO2018166116A1 (fr) Procédé de reconnaissance de dégâts causés à une voiture, appareil électronique et support d'informations lisible par ordinateur
WO2021012494A1 (fr) Procédé et appareil de reconnaissance faciale basée sur l'apprentissage profond, et support de stockage lisible par ordinateur
CN112861648B (zh) 文字识别方法、装置、电子设备及存储介质
WO2021042747A1 (fr) Procédé et système de reconnaissance et de vérification d'image de facture, dispositif, et support de stockage lisible
US20210174135A1 (en) Method of matching image and apparatus thereof, device, medium and program product
CN108491866B (zh) 色情图片鉴定方法、电子装置及可读存储介质
WO2021017272A1 (fr) Procédé et dispositif d'annotation d'une image de pathologie, appareil informatique et support d'informations
WO2019062191A1 (fr) Dispositif électronique, procédé et système d'extraction de données dans une table de données, et support de stockage
US10339373B1 (en) Optical character recognition utilizing hashed templates
WO2019062081A1 (fr) Procédé de formation de profil de vendeur, dispositif électronique et support d'informations lisible par ordinateur
CN109194689B (zh) 异常行为识别方法、装置、服务器及存储介质
WO2019085466A1 (fr) Procédé et système d'essai d'association, serveur d'application et support d'informations lisible par ordinateur
CN110610127A (zh) 人脸识别方法、装置、存储介质及电子设备
CN110288755A (zh) 基于文本识别的发票检验方法、服务器及存储介质
WO2019085338A1 (fr) Appareil électronique, procédé et système de classement par âge basés sur des images, et support de stockage
WO2019223146A1 (fr) Procédé d'identification du cancer de l'estomac, dispositif et support d'informations
CN111177113A (zh) 数据迁移方法、装置、计算机设备和存储介质

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: 17872723

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 10/09/2019)

122 Ep: pct application non-entry in european phase

Ref document number: 17872723

Country of ref document: EP

Kind code of ref document: A1