WO2021184718A1 - 卡片边框识别方法、装置、设备和计算机存储介质 - Google Patents

卡片边框识别方法、装置、设备和计算机存储介质 Download PDF

Info

Publication number
WO2021184718A1
WO2021184718A1 PCT/CN2020/117773 CN2020117773W WO2021184718A1 WO 2021184718 A1 WO2021184718 A1 WO 2021184718A1 CN 2020117773 W CN2020117773 W CN 2020117773W WO 2021184718 A1 WO2021184718 A1 WO 2021184718A1
Authority
WO
WIPO (PCT)
Prior art keywords
card
image
recognition
target image
straight line
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2020/117773
Other languages
English (en)
French (fr)
Inventor
张国辉
雷晨雨
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
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 Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Publication of WO2021184718A1 publication Critical patent/WO2021184718A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components

Definitions

  • This application relates to the field of image processing, and in particular to a method, device, equipment and computer storage medium for recognizing a card frame.
  • Card recognition is the first to bear the brunt of card border recognition.
  • the inventor realized that the existing card border recognition algorithms mainly use neural networks or traditional edge detection algorithms to find all the edge information in the picture, and then set various conditions to filter out some edges. Information, get the card frame.
  • Such a card recognition algorithm is relatively complicated, and the efficiency of card recognition is low, and if the background of the card shooting is complicated or the edge of the target image obtained by shooting is blurred, recognition errors are prone to occur, which affects the subsequent extraction of card information.
  • the main purpose of this application is to provide a method, device, device, and computer storage medium for recognizing a card frame, aiming to solve the current technical problems of low card edge recognition efficiency and inaccurate recognition.
  • this application provides a method for recognizing a card frame.
  • the method for recognizing a card frame includes the following steps:
  • a straight line segment perpendicular to each other between every two vertex coordinates is used as the card frame of the card in the target image, and a recognition result containing the card frame is output.
  • the present application also provides a card frame recognition device, the card frame recognition device includes:
  • the request receiving module is configured to obtain the target image corresponding to the image recognition request when the image recognition request is received;
  • the image recognition module is used to input the preprocessed target image into a preset card recognition model for image recognition, and obtain image feature information of the target image;
  • a coordinate construction module configured to establish a rectangular coordinate system based on the image characteristic information, and obtain the characteristic point coordinates of each characteristic point in the image characteristic information according to the rectangular coordinate system;
  • a vertex determination module configured to establish an equation set based on the feature point coordinates of each of the feature points, and solve the equation set to obtain the vertex coordinates of the card in the target image;
  • the result output module is configured to use the straight line segment between each two vertex coordinates as the card frame of the card in the target image, and output the recognition result containing the card frame.
  • the present application also provides a card frame recognition device
  • the card frame recognition device includes: a memory, a processor, and a computer program stored on the memory and running on the processor, wherein:
  • a straight line segment perpendicular to each other between every two vertex coordinates is used as the card frame of the card in the target image, and a recognition result containing the card frame is output.
  • this application also provides a computer storage medium
  • the computer storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:
  • a straight line segment perpendicular to each other between every two vertex coordinates is used as the card frame of the card in the target image, and a recognition result containing the card frame is output.
  • the embodiment of the present application improves the efficiency of recognition of the target image.
  • the terminal establishes an equation set according to the image feature information of the target image to determine the card frame in the target image, so that the recognized card frame is more accurate.
  • FIG. 1 is a schematic diagram of a device structure of a hardware operating environment involved in a solution of an embodiment of the present application
  • FIG. 2 is a schematic flowchart of a first embodiment of a method for identifying a card frame according to this application;
  • FIG. 3 is a schematic diagram of functional modules of an embodiment of a card frame recognition device according to the present application.
  • the technical solution of the present application can also be applied to the field of blockchain and/or big data technology.
  • the data involved in this application such as identification results, can be stored in a database, or can be stored in a blockchain, which is not limited in this application.
  • Figure 1 is a terminal (also called a card frame recognition device) of the hardware operating environment involved in the solution of the embodiment of the present application.
  • the card frame recognition device can be composed of a separate card frame recognition device or Other devices are combined with the card frame recognition device to form a schematic diagram of the structure.
  • the terminal in the embodiments of this application can be a fixed terminal or a mobile terminal, such as smart air conditioners with networking functions, smart lights, smart power supplies, smart speakers, autonomous vehicles, PC (personal computer) personal computers, smart phones, and tablet computers. , E-book readers, portable computers, etc.
  • the terminal may include a processor 1001, for example, a central processing unit (CPU), a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002.
  • the communication bus 1002 is used to implement connection and communication between these components.
  • the user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface.
  • the network interface 1004 may optionally include a standard wired interface and a wireless interface (such as WIreless-FIdelity, WIFI interface).
  • the memory 1005 may be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), for example, a magnetic disk memory.
  • the memory 1005 may also be a storage device independent of the aforementioned processor 1001.
  • the terminal may also include a camera, an RF (Radio Frequency) circuit, a sensor, an audio circuit, and a WiFi module; an input unit, a display screen, and a touch screen; the network interface can be optional.
  • RF Radio Frequency
  • sensors such as light sensors, motion sensors and other sensors.
  • the light sensor may include an ambient light sensor and a proximity sensor; of course, the mobile terminal may also be equipped with other sensors such as gyroscope, barometer, hygrometer, thermometer, infrared sensor, etc., which will not be repeated here.
  • terminal structure shown in FIG. 1 does not constitute a limitation on the terminal, and may include more or less components than shown in the figure, or combine some components, or arrange different components.
  • the computer software product is stored in a storage medium (storage medium: also called computer storage medium, computer medium, readable medium, readable storage medium, computer readable storage medium, or directly called medium, etc., storage medium
  • storage medium can be a non-volatile readable storage medium, such as RAM, magnetic disk, optical disk, and includes several instructions to make a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute this application
  • the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a computer program.
  • the network interface 1004 is mainly used to connect to a back-end server and communicate with the back-end server;
  • the user interface 1003 is mainly used to connect to a client (user side) to communicate with the client;
  • the processor 1001 can be used to call a computer program stored in the memory 1005 and execute the steps in the card frame recognition method provided in the following embodiments of the present application.
  • the card frame identification method includes:
  • Step S10 when an image recognition request is received, a target image corresponding to the image recognition request is acquired.
  • the card frame recognition method in this embodiment is applied to a card frame recognition device.
  • the type of card frame recognition device is not limited.
  • a terminal is taken as an example for description.
  • the terminal receives an image recognition request and obtains the target image in the image recognition request.
  • the target image in this embodiment contains card information, and may also contain other information besides the card, for example, background information when the card was taken, card source information or card format information and so on.
  • the color and size of the target image are not specifically limited.
  • the target image may be in color or black and white.
  • the terminal After the terminal obtains the target image, the terminal preprocesses the target image, which specifically includes:
  • Step a1 cutting the target image to obtain a preliminary cut image, and performing binarization processing on the preliminary cut image to obtain a binarization processed image;
  • Step a3 Perform denoising processing on the binarized image, and complete the preprocessing of the target image.
  • the terminal cuts the target image to divide the target image into image units, cuts out the card information from the target image, and obtains the preliminary cut image.
  • the terminal performs image segmentation through any one or more of the following algorithms, for example, , 1.
  • Threshold-based segmentation method is to use one or several thresholds to divide the grayscale histogram of an image into several categories, and consider that pixels in the image with grayscale values in the same category belong to the same object.
  • the maximum correlation principle selects the threshold method, the method based on the stable state of the image topology, the Yager measure minimization method, the gray level co-occurrence matrix method, the variance method, the entropy method, the peak and valley analysis method, etc., among which, the adaptive threshold method , Maximum Entropy, Fuzzy Threshold, and Inter-Class Threshold are several algorithms that have improved the traditional threshold method more successfully); 2. Edge-based segmentation method (the segmentation method based on edge detection attempts to detect edges that contain different regions Solve segmentation problems); 3. Region-based segmentation methods (the essence of region segmentation is to connect images with similar properties to form the final segmentation area.
  • the method uses the local spatial information of the image, which can effectively overcome other
  • the method has the disadvantage that the image segmentation space is small and continuous.
  • this type of method if you start from the whole image, determine the area attribution of each pixel according to the criteria of the same area attribute feature, and form a regional map, which is often called regional growth. Segmentation method. If starting from the pixel, according to the criterion of the same area attribute feature, the connected pixels with close attributes are gathered into the area, which is the segmentation method of area growth); 4.
  • Image segmentation method based on cluster analysis is to represent the pixels in the image space with corresponding feature space points, segment the feature space according to their aggregation in the feature space, and then map them back to the original image space to obtain the segmentation result.
  • K-means and fuzzy C-means clustering algorithms are the most commonly used clustering algorithms
  • Segmentation method based on wavelet transform threshold image segmentation method based on wavelet transform uses dyadic wavelet transform to decompose the histogram of the image into different levels Then select the threshold threshold according to the given segmentation criteria and wavelet coefficients, and finally use the threshold to mark the image segmentation area); 6.
  • Segmentation method based on mathematical morphology is a nonlinear filtering method, It can be used to suppress noise, feature extraction, edge detection, image segmentation, etc.
  • 7. Segmentation method based on artificial neural network the basic idea of segmentation method based on neural network is to obtain a linear decision function by training a multilayer perceptron, and then Use the decision function to classify the pixels to achieve the purpose of segmentation; the terminal performs binarization processing on the preliminary cut image to obtain a binarized processed image; the terminal performs denoising processing on the binarized image to obtain denoising processing Image, the terminal completes the preprocessing of the target image.
  • the terminal performs binarization processing on the preliminary cut image to obtain a binarized image. Then, the terminal uses morphological operations to filter out the noise in the binarized image, and the purpose is to remove the binarized image. Process other noise points in the image except the card information to obtain a denoising processed image; wherein, the implementation process of using morphological operations to filter out the noise in the binarized image can refer to related technologies. Not limited.
  • the terminal adjusts the size and pixel value of the target image. For example, the terminal scales the size of the target image to 128*256; the terminal divides the pixel value of the target image by 255. In this embodiment, the terminal first The target image is preprocessed, and then the size of the target image is adjusted, and the pixel value of the target image is divided by 255 to reduce the target image pixels to facilitate post-image processing.
  • Step S20 Input the preprocessed target image into a preset card recognition model for image recognition, and obtain image feature information of the target image.
  • the card recognition model is preset in the terminal.
  • the preset card recognition model is a preset algorithm for card recognition.
  • the preset card recognition model randomly selects samples from historical card images, and marks the middle of the card and the card on some samples. Outside the edge, use the labeled samples as input to build an initial model (the initial model refers to a function that classifies the middle of the card and the outside of the edge of the card), and further extracts a preset proportion of unlabeled samples to train the initial model At the same time, the recognition accuracy rate of the training model is obtained.
  • the training model is used as the preset card recognition model
  • the preset card recognition model includes the encoder (the encoder is formed by the superposition of the down-sampling module and the bottleneck module), the decoder (the decoder is formed by the superimposition of the bottleneck module and the up-sampling module) and the weighted runner.
  • the terminal passes through the down-sampling module, The bottleneck module and the up-sampling module convolve the target image and pool the image feature information.
  • the terminal processes a target image with a pixel of 128x256 through a preset card recognition model to decompose a 64x32x64 image, and then the terminal performs a convolution operation on each 32x64 small image area. That is to say, the terminal extracts a small area of 32x64 and marks it as (1, 1), (1, 2), ..., from the starting coordinates, until (n, n), and then the extracted area Run the training one by one to obtain n sets, each set contains 32x64 convolutional features, the terminal will contain a combination of 32x64 convolutional features to obtain 4*128*256 image feature information.
  • step S30 a rectangular coordinate system is established based on the image characteristic information, and the characteristic point coordinates of each characteristic point in the image characteristic information are acquired according to the rectangular coordinate system.
  • the terminal establishes a rectangular coordinate system based on the image feature information.
  • the terminal uses the center of the image feature information as the coordinate origin.
  • the center of the image feature information refers to the circumscribed circle of the feature information, and the center of the circumscribed circle is the coordinate origin.
  • the Cartesian coordinate system determines the feature point coordinates of each feature point in the image feature information.
  • the terminal After determining the coordinate origin, the terminal obtains the coordinates of the end point of a straight line in the feature information, or the coordinates of the intersection of two straight lines, and the terminal according to the feature of each feature point Point coordinates determine the set of equations, specifically: because the card has 4 borders, 4 borders correspond to 4 straight line equations, there are two sets of two parallel borders, the slopes of the two straight line equations are the same, there are two sets of two vertical In the frame, the product of the slopes corresponding to the two straight line equations is equal to 1 or -1, or does not exist, and the terminal establishes an equation set according to the determined relationship of at least 4 points and slopes.
  • step S40 an equation set is established based on the feature point coordinates of each of the feature points, and the equation set is solved to obtain the coordinates of the vertices of the card in the target image.
  • step S40 includes:
  • Step b3 Calculate the straight line parameter calculation formula to obtain the straight line parameters, determine the straight line equation according to the straight line parameters, determine the intersection point coordinates by the straight line equation, and use the intersection point coordinates as the apex coordinates of the card in the target image, where:
  • step b3 in this embodiment further includes: the terminal obtains the linear parameter according to the linear parameter calculation formula, the terminal determines the linear equation corresponding to the linear parameter, extracts the slope in the linear equation, and judges whether the product of the slopes is negative one, or One slope is zero and the other slope does not exist; if the product between the slopes is negative one, or one slope is zero and the other slope does not exist, the intersection point coordinates are determined by the straight line equation, and the intersection point coordinates As the vertex coordinates of the card in the target image.
  • the terminal obtains 4 slopes, and first determines whether there are two sets of the same slope. If the terminal determines that there are no two sets of the same slope, then The terminal judges that the linear equation is wrong and outputs prompt information; if the terminal determines that there are two groups of the same slope, the terminal extracts a slope from each group and calculates whether the product between the slopes is negative one, or one slope is zero and the other slope Does not exist; if the product between the slopes is not negative one, or one slope is zero and the other does not exist, the terminal determines that the linear equation is wrong, and the terminal outputs a prompt message.
  • the terminal determines that the straight line equation is correct, the terminal determines the coordinates of the intersection point according to the direct equation, and the terminal uses the coordinates of the intersection point as the apex coordinates of the card in the target image.
  • an equation set is established based on the feature point coordinates of the feature point to obtain the vertex coordinates of the card in the image.
  • step S50 a straight line segment perpendicular to each other between each two vertex coordinates is used as the card frame of the card in the target image, and a recognition result including the card frame is output.
  • the terminal obtains 4 points, and the terminal connects the 4 points in pairs.
  • the terminal uses the straight line segment between each two vertex coordinates as the card border of the card in the target image, and outputs the recognition result including the card border .
  • what is obtained according to the image feature information is necessarily a straight edge, and when the model can be simplified, a very high degree of accuracy of card border recognition can also be obtained.
  • a card recognition model is preset in the terminal, and the terminal recognizes the target image through the preset card recognition model to obtain image feature information of the target image.
  • the preset card recognition model only performs simple processing on the target image, which reduces The recognition time of the target image improves the recognition efficiency of the target image.
  • the terminal establishes an equation set according to the image feature information of the target image to determine the card frame in the target image, so that the recognized card frame is more accurate.
  • This embodiment is a step before step S20 in the first embodiment.
  • This embodiment specifically describes the steps of creating a preset card recognition model, including:
  • Annotate card feature points on the image samples in the card sample set divide the card inner area and the card outer area according to the card feature points, and obtain the area division function between the card inner area and the card outer area;
  • the region segmentation function is iteratively trained through the iterative feature points, and the separation correctness rate of the trained region segmentation function is obtained, and the region segmentation function whose separation correctness rate reaches a preset threshold is used as a preset card recognition model.
  • the terminal when the terminal receives the model training instruction, the terminal obtains the card sample set corresponding to the model training instruction; the terminal marks the card feature points on the image samples in the card sample set, and the terminal divides the inner area of the card and the outer area of the card according to the feature points of the card Area, after determining the inner area of the card and the outer area of the card, the terminal fitting function obtains the area segmentation function between the inner area of the card and the outer area of the card, and the terminal uses this area segmentation function as the initial training model.
  • the terminal extracts iterative training samples from the card sample set at a preset ratio (the preset ratio refers to the preset training sample extraction ratio, in this embodiment, the preset ratio is set to 10 times), and the terminal obtains the card features in the iterative training sample Point, and use the card feature points in the iterative training sample as iterative feature points; the terminal iteratively trains the region segmentation function through iterative feature points, and obtains the separation accuracy of the trained region segmentation function, and the trained region is detected on the terminal
  • the region segmentation function whose separation accuracy rate of the segmentation function reaches a preset threshold (the preset threshold can be set according to specific scenarios, for example, the preset threshold is set to 95%) is used as the preset card recognition model.
  • the preset card recognition model obtained in this embodiment is a simplified model. Performing image recognition through the preset card recognition model can not only ensure the efficiency of image recognition, but also ensure the accuracy of image recognition.
  • This embodiment is a step after step S40 in the first embodiment of the present application.
  • the difference between this embodiment and the foregoing embodiment lies in:
  • the card image in the card frame is recognized by an optical character recognition algorithm to obtain the character information in the card image.
  • the terminal receives the card information recognition request, that is, after the terminal determines the card frame, the terminal outputs prompt information for the user to confirm.
  • the terminal receives the user confirmation instruction, the terminal automatically triggers the card information recognition request, and the terminal obtains the card frame The card image inside.
  • the terminal recognizes each single character image in the card image through the OCR (Optical Character Recognition) module, where the OCR module can use the CNN (Convolutional Neural Networks, convolutional neural network) technology to identify the single character image. Character images are classified and recognized. In this way, the card information in the card is recognized.
  • the optical character recognition algorithm is used to recognize the card image in the card frame to obtain the character information in the card image, so that the user does not need to actively input the card information, which reduces the user's operations.
  • This embodiment is a step after the third embodiment of the present application.
  • the difference between this embodiment and the foregoing embodiment lies in:
  • the terminal determines the card type of the card in the target image according to the character information.
  • the character information includes name, hometown, birthday, etc., and the terminal determines that the card is an ID card.
  • the character information includes: vehicle information, valid time, vehicle
  • the terminal determines that the card is a driver’s license.
  • the terminal obtains the card information record table corresponding to the card type.
  • the terminal adds the character information to the card information record table and adds identification information (identification information). It can refer to information that uniquely identifies the card, for example, the identification information is card number information).
  • the terminal When the terminal receives the card information query request, the terminal obtains the identification information corresponding to the card information query request; the terminal queries the card information record table, and the terminal obtains the character information corresponding to the card identification and outputs it for the user to view. In this embodiment, the terminal adds the recognized character information to the card information record table to facilitate later viewing.
  • an embodiment of the present application also proposes a card frame recognition device, and the card frame recognition device includes:
  • the request receiving module 10 is configured to obtain a target image corresponding to the image recognition request when the image recognition request is received;
  • the image recognition module 20 is configured to input the preprocessed target image into a preset card recognition model for image recognition, and obtain image feature information of the target image;
  • the coordinate construction module 30 is configured to establish a rectangular coordinate system based on the image characteristic information, and obtain the characteristic point coordinates of each characteristic point in the image characteristic information according to the rectangular coordinate system;
  • the vertex determination module 40 is configured to establish an equation set based on the feature point coordinates of each of the feature points, and solve the equation set to obtain the vertex coordinates of the card in the target image;
  • the result output module 50 is configured to use a straight line segment perpendicular to each other between each two vertex coordinates as the card frame of the card in the target image, and output the recognition result including the card frame.
  • the card frame recognition device includes:
  • An image segmentation module configured to cut the target image to obtain a preliminary cut image, and perform binarization processing on the preliminary cut image to obtain a binarization processed image
  • the image denoising module is used to denoise the binarized image and complete the preprocessing of the target image.
  • the card frame recognition device includes:
  • the sample acquisition module is configured to acquire the card sample set corresponding to the model training instruction when the model training instruction is received;
  • the labeling and division module is used to label the card feature points on the image samples in the card sample set, divide the card inner area and the card outer area according to the card feature points, and obtain the gap between the card inner area and the card outer area Region segmentation function;
  • a sample extraction module configured to extract iterative training samples from the card sample set at a preset ratio, and use card feature points in the iterative training samples as iterative feature points;
  • the model training module is used to iteratively train the region segmentation function through the iterative feature points, and obtain the separation accuracy rate of the trained region segmentation function, and use the region segmentation function whose separation accuracy rate reaches a preset threshold as the preliminary Set up a card recognition model.
  • the vertex determining module 40 includes:
  • the X_map is the characteristic value of the x-axis coordinate
  • y_map is the characteristic value of the y-axis coordinate
  • the A represents a set of straight line parameters
  • the vertex determination unit is used to calculate the straight line parameter calculation formula to obtain the straight line parameters, determine the straight line equation according to the straight line parameters, determine the intersection point coordinates by the straight line equation, and use the intersection point coordinates as the apex coordinates of the card in the target image ,
  • the vertex determination unit is further configured to:
  • intersection point coordinates are determined by the straight line equation, and the intersection point coordinates are used as the apex coordinates of the card in the target image.
  • the card frame recognition device includes:
  • a request receiving module configured to receive a card information recognition request, and obtain the card image in the card frame
  • the character acquisition module is used for recognizing the card image in the card frame through an optical character recognition algorithm to obtain character information in the card image.
  • the card frame recognition device includes:
  • An information adding module configured to determine the card type of the card in the target image according to the character information, add the character information to the card information record table corresponding to the card type, and add identification information;
  • the query receiving module is configured to obtain the identification information corresponding to the card information query request when the card information query request is received;
  • the information acquisition module is used for querying the card information record table, acquiring and outputting character information corresponding to the card identifier.
  • each functional module of the card frame recognition device can refer to the various embodiments of the card frame recognition method of the present application, which will not be repeated here.
  • the embodiment of the present application also proposes a computer storage medium.
  • a computer program is stored on the computer storage medium, and when the computer program is executed by a processor, the operations in the card frame recognition method provided in the above embodiments are implemented.
  • the storage medium (computer storage medium) involved in this application may be non-volatile or volatile.
  • the description is relatively simple, and for related parts, please refer to the part of the description of the method embodiment.
  • the device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separate. Some or all of the modules can be selected according to actual needs to achieve the objectives of the solution of the present application. Those of ordinary skill in the art can understand and implement without creative work.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Evolutionary Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Image Analysis (AREA)

Abstract

一种卡片边框识别方法，包括以下步骤：在接收到图像识别请求时，获取所述图像识别请求对应的目标图像(S10)；将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息(S20)；基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标(S30)；基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标(S40)；将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果(S50)。同时公开了一种卡片边框识别装置、设备和计算机存储介质，提高了卡片边框的识别效率和准确率。

Description

卡片边框识别方法、装置、设备和计算机存储介质
本申请要求于2020年3月16日提交中国专利局、申请号为202010182749.8，发明名称为“卡片边框识别方法、装置、设备和计算机存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及图像处理领域，尤其涉及卡片边框识别方法、装置、设备和计算机存储介质。
背景技术
随着身份证、社保卡和银行卡等各种卡片大量的使用，卡片识别的应用场景越来越多。
卡片识别首当其冲的就是卡片边框识别，发明人意识到，现有的卡片边框识别算法，主要是采用神经网络或者传统的边缘检测算法找到图片中的所有边缘信息，然后设置各种条件过滤掉一些边缘信息，得到卡片边框。这样的卡片识别算法较为复杂，卡片识别的效率低，且若卡片拍摄背景复杂或者拍摄得到的目标图像边缘模糊，容易出现识别错误，从而影响后续对卡片信息的提取。
发明内容
本申请的主要目的在于提供一种卡片边框识别方法、装置、设备和计算机存储介质，旨在解决当前卡板边缘识别效率低，识别不准确的技术问题。
为实现上述目的，本申请提供卡片边框识别方法，所述卡片边框识别方法包括以下步骤：
在接收到图像识别请求时，获取所述图像识别请求对应的目标图像；
将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息；
基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标；
基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标；
将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果。
此外，为实现上述目的，本申请还提供一种卡片边框识别装置，所述卡片边框识别装置包括：
请求接收模块，用于在接收到图像识别请求时，获取所述图像识别请求对应的目标图像；
图像识别模块，用于将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息；
坐标构建模块，用于基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标；
顶点确定模块，用于基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标；
结果输出模块，用于将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果。
此外，为实现上述目的，本申请还提供一种卡片边框识别设备；
所述卡片边框识别设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序，其中：
所述计算机程序被所述处理器执行时实现以下步骤：
在接收到图像识别请求时，获取所述图像识别请求对应的目标图像；
将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息；
基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标；
基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标；
将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果。
此外，为实现上述目的，本申请还提供计算机存储介质；
所述计算机存储介质上存储有计算机程序，所述计算机程序被处理器执行时实现以下步骤：
在接收到图像识别请求时，获取所述图像识别请求对应的目标图像；
将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息；
基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标；
基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标；
将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果。
本申请实施例提高了目标图像的识别效率，终端根据目标图像的图像特征信息建立方程组，确定目标图像中的卡片边框，使得识别到的卡片边框更加准确。
附图说明
图1是本申请实施例方案涉及的硬件运行环境的装置结构示意图；
图2为本申请卡片边框识别方法第一实施例的流程示意图；
图3为本申请卡片边框识别装置一实施例的功能模块示意图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
应当理解，此处所描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本申请的技术方案还可应用于区块链和/或大数据技术领域。可选的，本申请涉及的数据如识别结果等可存储于数据库中，或者可以存储于区块链中，本申请不做限定。
如图1所示，图1是本申请实施例方案涉及的硬件运行环境的终端(又叫卡片边框识别设备，其中，卡片边框识别设备可以是由单独的卡片边框识别装置构成，也可以是由其他装置与卡片边框识别装置组合形成)结构示意图。
本申请实施例终端可以固定终端，也可以是移动终端，如，带联网功能的智能空调、智能电灯、智能电源、智能音箱、自动驾驶汽车、PC(personal computer)个人计算机、智能手机、平板电脑、电子书阅读器、便携计算机等。
如图1所示，该终端可以包括：处理器1001，例如，中央处理器Central Processing Unit，CPU)，网络接口1004，用户接口1003，存储器1005，通信总线1002。其中，通信总线1002用于实现这些组件之间的连接通信。用户接口1003可以包括显示屏(Display)、输入单元比如键盘(Keyboard)，可选用户接口1003还可以包括标准的有线接口、无线接口。网络接口1004可选的可以包括标准的有线接口、无线接口(如无线保真WIreless-FIdelity，WIFI接口)。存储器1005可以是高速RAM存储器，也可以是稳定的存储器(non-volatile memory)，例如， 磁盘存储器。存储器1005可选的还可以是独立于前述处理器1001的存储装置。
可选地，终端还可以包括摄像头、RF(Radio Frequency，射频)电路，传感器、音频电路、WiFi模块；输入单元，比显示屏，触摸屏；网络接口可选除无线接口中除WiFi外，蓝牙、探针等等。其中，传感器比如光传感器、运动传感器以及其他传感器。具体地，光传感器可包括环境光传感器及接近传感器；当然，移动终端还可配置陀螺仪、气压计、湿度计、温度计、红外线传感器等其他传感器，在此不再赘述。
本领域技术人员可以理解，图1中示出的终端结构并不构成对终端的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。
如图1所示，该计算机软件产品存储在一个存储介质(存储介质：又叫计算机存储介质、计算机介质、可读介质、可读存储介质、计算机可读存储介质或者直接叫介质等，存储介质可以是非易失性可读存储介质，如RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法，作为一种计算机存储介质的存储器1005中可以包括操作系统、网络通信模块、用户接口模块以及计算机程序。
在图1所示的终端中，网络接口1004主要用于连接后台服务器，与后台服务器进行数据通信；用户接口1003主要用于连接客户端(用户端)，与客户端进行数据通信；而处理器1001可以用于调用存储器1005中存储的计算机程序，并执行本申请以下实施例提供的卡片边框识别方法中的步骤。
基于上述硬件结构，提出了本申请卡片边框识别方法的实施例。
参照图2，在本申请卡片边框识别方法的第一实施例中，所述卡片边框识别方法包括：
步骤S10，在接收到图像识别请求时，获取所述图像识别请求对应的目标图像。
本实施例中的卡片边框识别方法应用在卡片边框识别设备，卡片边框识别设备的种类不作限定，本实施例中以终端为例进行说明，终端接收图像识别请求，获取图像识别请求中的目标图像；可以理解的是，本实施例中目标图像中包含卡片信息，还可以包含除卡片之外的其他信息，例如，卡片拍摄时的背景信息、卡片的来源信息或者卡片的格式信息等等。此外，目标图像的颜色和尺寸不作具体限定，例如，目标图像可以是彩色或者黑白。
终端在获取到目标图像之后，终端对目标图像进行预处理，具体地，包括：
步骤a1，对所述目标图像进行切割，得到初步切割图像，对所述初步切割图像进行二值化处理，得到二值化处理图像；
步骤a3，对所述二值化处理图像进行去噪处理，并完成对所述目标图像的预处理。
即，终端对目标图像进行切割将目标图像分割为图像单元，从该目标图像中切割出卡片信息，得到初步切割图像，具体地，包括终端通过以下任意一种或者多种算法进行图像分割，例如，1.基于阈值的分割方法(阈值分割方法就是用一个或几个阈值将图像的灰度直方图分成几个类，认为图像中灰度值在同一类中的像素属于同一物体。例如，用最大相关性原则选择阈值的方法、基于图像拓扑稳定状态的方法、Yager测度极小化方法、灰度共生矩阵方法、方差法、熵法、峰值和谷值分析法等，其中，自适应阈值法、最大熵法、模糊阈值法、类间阈值法是对传统阈值法改进较成功的几种算法)；2.基于边缘的分割方法(基于边缘检测的分割方法试图通过检测包含不同区域的边缘来解决分割问题)；3.基于区域的分割方法(区域分割的实质就是把具有某种相似性质的像索连通，从而构成最终的分割区域。它利用了图像的局部空间信息，可有效地克服其他方法存在的图像分割空间小连续的缺点。在此类方法中，如果从全图出发，按区域属性特征一致的准则决定每个像元的区域归属，形成区域图，常称之为区域生长的分割方法。如果从像元出发，按区域属性特征一致的准则，将属性接近的连通像元聚集为区域，则是区域增长的分割方法)；4.基于聚类分析的图像分割方法(特征空间聚类法进行图像分割是将图像空间中的像素用对应的特征空 间点表示，根据它们在特征空间的聚集对特征空间进行分割，然后将它们映射回原图像空间，得到分割结果。其中，包括：K均值、模糊C均值聚类算法是最常用的聚类算法)；5.基于小波变换的分割方法(基于小波变换的阈值图像分割方法由二进小波变换将图像的直方图分解为不同层次的小波系数，然后依据给定的分割准则和小波系数选择阈值门限，最后利用阈值标出图像分割的区域)；6.基于数学形态学的分割方法(数学形态学是一种非线性滤波方法，可以用于抑制噪声、特性提取、边缘检测、图像分割等)；7.基于人工神经网络的分割方法(基于神经网络的分割方法的基本思想是通过训练多层感知机来得到线性决策函数，然后用决策函数对像素进行分类来达到分割的目的)；终端对该初步切割图像进行二值化处理，得到二值化处理图像；终端对该二值化处理图像进行去噪处理，得到去噪处理图像，以终端完成对目标图像的预处理。
本实施例中终端对初步切割图像进行二值化处理后，得到二值化处理图像，然后，终端运用形态学运算滤除该二值化处理图像中的噪声，其目的是去除该二值化处理图像中除卡片信息之外的其它噪点，得到去噪处理图像；其中，上述运用形态学运算滤除该二值化处理图像中的噪声的实现过程可以参见相关技术，本实施例在对此不做限定。
此外，在目标图像预处理之后，终端调整目标图像的大小和像素值，例如，终端将目标图像的大小缩放为128*256；终端将目标图像的像素值除以255，本实施例中终端先对目标图像进行预处理，然后，调整目标图像的大小，再将目标图像的像素值除以255降低目标图像像素，以方便后期图像处理。
步骤S20，将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息。
终端中预设卡片识别模型，预设卡片识别模型是预先设置的用于卡片识别的算法，预设卡片识别模型是从历史卡片图像中随机抽取样本，并在部分样本上进行标注卡片中部和卡片边沿外部，然后利用标注了的样本作为输入，构建一个初始模型(初始模型是指将卡片中部和卡片边沿外部进行分类的函数)，进一步地抽取预设比例的没有标注的样本对初始模型进行训练，同时获取训练模型的识别准确率，在识别准确率高于预设准确率(预设准确率可以根据具体场景设置，例如设置为98％)时，将该训练模型作为预设卡片识别模型，预设卡片识别模型中包含有编码器(编码器由下采样模块和瓶颈模块叠加形成)，解码器(解码器由瓶颈模块和上采样模块叠加形成)和加权运行器，终端通过下采样模块、瓶颈模块和上采样模块将目标图像进行卷积，池化得到图像特征信息。
例如，终端将像素为128x256的目标图像通过预设卡片识别模型处理，分解得到64x32x64的图像，然后终端对每个32x64的小块图像区域都进行卷积运算。也就是说，终端抽取32x64的小块区域，并且从起始坐标开始依次标记为(1，1)，(1，2)，...，一直到(n,n)，然后对抽取的区域逐个运行训练得到n个集合，每个集合含有32x64个卷积特征，终端将包含32x64个卷积特征的组合得到4*128*256的图像特征信息。
步骤S30，基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标。
终端基于图像特征信息建立直角坐标系，例如，终端以图像特征信息的中心作为坐标原点，其中，图像特征信息的中心是指将特征信息做外接圆，其中外接圆的圆心作为坐标原点，终端根据直角坐标系确定图像特征信息中各特征点的特征点坐标，例如，在确定坐标原点之后，终端获取特征信息中一条直线端点的坐标，或者两条直线交点的坐标，终端根据各特征点的特征点坐标确定方程组，具体地：由于卡片存在4条边框，4条边框的对应4个直线方程，存在两组两条平行的边框，两直线方程对应的斜率相同，存在两组两条垂直的边框，两直线方程对应的斜率的乘积等于1或-1，或者不存在，终端按照确定的至少4个点和斜率关系建立方程组。
步骤S40，基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标。
终端获取各特征点的特征点坐标，终端构建纵坐标为因变量，横坐标为自变量的方程组，终端通过建立方程组确定获取直线参数，终端在得到直线参数之后，终端确定直线参数对应的直线方程，从而获得目标图像中卡片的顶点坐标，具体地，步骤S40，包括：
步骤b1，基于各所述特征点的特征点坐标建立方程组，其中，所述方程组为：W*A*[y_map,1]＝W*x_map，所述W表示特征值，所述X_map为x轴坐标的特征值，y_map为y轴坐标的特征值，所述A表示直线参数集合；
步骤b2，求解所述方程组获得直线参数计算公式，其中，所述直线参数计算公式为A＝inv(T(Wy)*Wy)*(T(Wy)*Wx)，其中，所述T(x)为x的转置，所述inv(x)为x的逆；
步骤b3，计算所述直线参数计算公式获得直线参数，根据所述直线参数确定直线方程，通过所述直线方程确定交点坐标，将所述交点坐标作为所述目标图像中卡片的顶点坐标，其中，所述直线方程包括y＝a1x+b1和x＝a2y+b2，其中，所述a1、所述a2、所述b1和所述b2表示直线参数。
此外，本实施例中步骤b3，还包括：终端根据直线参数计算公式获得直线参数，终端确定直线参数对应的直线方程，提取直线方程中的斜率，判断斜率之间的乘积是否为负一，或者一个斜率为零且另一个斜率不存在；若所述斜率之间的乘积为负一，或者一个斜率为零且另一个斜率不存在，则通过所述直线方程确定交点坐标，将所述交点坐标作为所述目标图像中卡片的顶点坐标。
由于卡片的相邻两条边框相互垂直，在本实施例中终端获得直线方程之后，终端获取4个斜率，首先判断是否存在两组相同的斜率，若终端确定不存在两组相同的斜率，则终端判定直线方程错误输出提示信息；若终端确定存在两组相同的斜率，则终端从每一组中抽取一个斜率，计算斜率之间的乘积是否为负一，或者一个斜率为零且另一个斜率不存在；若斜率之间的乘积不是负一，或者一个斜率为零且另一个斜率不存在，则终端判定直线方程错误，终端输出提示信息，反之，若斜率之间的乘积是负一，或者一个斜率为零且另一个斜率不存在，则终端判定直线方程正确，终端根据直接方程确定交点坐标，终端将交点坐标作为目标图像中卡片的顶点坐标。
本实施例中根据特征点的特征点坐标建立方程组，获得图像中卡片的顶点坐标。
步骤S50，将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果。
本实施例中终端得到4个点，终端将4个点两两连接，终端将每两个顶点坐标之间相互垂直的直线段作为目标图像中卡片的卡片边框，并输出包含卡片边框的识别结果。本实施例中根据图像特征信息得到的必然是直线边缘，而且可以做到模型精简的情况下，也能获得很高的卡片边框识别更加准确度。
在本实施例中终端中预设卡片识别模型，终端通过预设卡片识别模型识别目标图像，获得目标图像的图像特征信息，本实施例中预设卡片识别模型只对目标图像进行简单处理，减少目标图像的识别时间，提高了目标图像的识别效率，终端根据目标图像的图像特征信息建立方程组，确定目标图像中的卡片边框，使得识别到的卡片边框更加准确。
进一步地，在本申请第一实施例的基础上，提出了本申请卡片边框识别方法的第二实施例。
本实施例是第一实施例中步骤S20之前的步骤，本实施例中具体说明了预设卡片识别模型创建的步骤，包括：
在接收到模型训练指令时，获取所述模型训练指令对应的卡片样本集；
在所述卡片样本集中的图像样本上标注卡片特征点，根据所述卡片特征点划分卡片内 部区域和卡片外部区域，获得所述卡片内部区域和所述卡片外部区域之间的区域分割函数；
以预设比例从所述卡片样本集中抽取迭代训练样本，将所述迭代训练样本中的卡片特征点作为迭代特征点；
通过所述迭代特征点对所述区域分割函数进行迭代训练，并获取训练的区域分割函数的分隔正确率，将所述分隔正确率达到预设阈值的区域分割函数作为预设卡片识别模型。
本实施例中终端在接收到模型训练指令时，终端获取模型训练指令对应的卡片样本集；终端在卡片样本集中的图像样本上标注卡片特征点，终端根据卡片特征点划分卡片内部区域和卡片外部区域，在确定卡片内部区域和卡片外部区域之后，终端拟合函数获得卡片内部区域和卡片外部区域之间的区域分割函数，终端将该区域分割函数作为初始训练模型。
终端以预设比例(预设比例是指预先设置的训练样本抽取比例，本实施例中将预设比例设置为10次)从卡片样本集中抽取迭代训练样本，终端获取迭代训练样本中的卡片特征点，并将迭代训练样本中的卡片特征点作为迭代特征点；终端通过迭代特征点对区域分割函数进行迭代训练，并获取训练的区域分割函数的分隔正确率，在终端检测到训练得到的区域分割函数分隔正确率达到预设阈值(预设阈值可以根据具体场景设置，例如预设阈值设置为95％)的区域分割函数作为预设卡片识别模型。
在本实施例中得到的预设卡片识别模型为一个简化的模型，通过该预设卡片识别模型进行图像识别既可以保证图像识别效率，又可以保证图像识别准确度。
进一步地，在本申请上述实施例的基础上，提出了本申请卡片边框识别方法的第三实施例。
本实施例是本申请第一实施例中步骤S40之后的步骤，本实施例与上述实施例的区别在于：
接收卡片信息识别请求，获取所述卡片边框内的卡片图像；
通过光学字符识别算法对所述卡片边框内的卡片图像进行识别，获得所述卡片图像中的字符信息。
在本实施例中终端接收卡片信息识别请求，即，终端确定卡片边框之后，终端输出提示信息以供用户确认，终端在接收到用户确认指令时，终端自动触发卡片信息识别请求，终端获取卡片边框内的卡片图像。
终端通过OCR(Optical Character Recognition，光学字符识别)模块对该卡片图像中的每个单字符图像进行识别，其中，在该OCR模块可以通过CNN(Convolutional Neural Networks，卷积神经网络)技术对该单字符图像进行分类识别。如此，即实现了对该卡片中的卡片信息进行识别。通过光学字符识别算法对所述卡片边框内的卡片图像进行识别，获得所述卡片图像中的字符信息，使得用户不需要主动地输入卡片信息，减少了用户的操作。
进一步地，在本申请上述实施例的基础上，提出了本申请卡片边框识别方法的第四实施例。
本实施例是本申请第三实施例之后的步骤，本实施例与上述实施例的区别在于：
根据所述字符信息确定所述目标图像中卡片的卡片类型，将所述字符信息添加到所述卡片类型对应的卡片信息记录表中，并添加标识信息；
在接收到卡片信息查询请求时，获取所述卡片信息查询请求对应的标识信息；
查询所述卡片信息记录表，获取所述卡片标识对应的字符信息并输出。
终端根据字符信息确定目标图像中卡片的卡片类型，例如，字符信息中包含姓名、籍贯、生日等等，终端确定该卡片为身份证，再比如，字符信息中包含：车辆信息、有效时间、车辆种类等等，终端确定该卡片为驾驶证，终端确定卡片类型之后，终端获取该卡片类型对应的卡片信息记录表，终端将字符信息添加到该卡片信息记录表中，并添加标识信息(标识信息可以是指唯一识别卡片的信息，例如标识信息为卡片编号信息)。
终端在接收到卡片信息查询请求时，终端获取卡片信息查询请求对应的标识信息；终端查询卡片信息记录表，终端获取卡片标识对应的字符信息并输出，以供用户查看。在本实施例中终端将识别得到的字符信息添加到卡片信息记录表中，方便后期的查看。
此外，参照图3，本申请实施例还提出一种卡片边框识别装置，所述卡片边框识别装置包括：
请求接收模块10，用于在接收到图像识别请求时，获取所述图像识别请求对应的目标图像；
图像识别模块20，用于将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息；
坐标构建模块30，用于基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标；
顶点确定模块40，用于基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标；
结果输出模块50，用于将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果。
在一实施例中，所述的卡片边框识别装置，包括：
图像分割模块，用于对所述目标图像进行切割，得到初步切割图像，对所述初步切割图像进行二值化处理，得到二值化处理图像；
图像去噪模块，用于对所述二值化处理图像进行去噪处理，并完成对所述目标图像的预处理。
在一实施例中，所述的卡片边框识别装置，包括：
样本获取模块，用于在接收到模型训练指令时，获取所述模型训练指令对应的卡片样本集；
标注划分模块，用于在所述卡片样本集中的图像样本上标注卡片特征点，根据所述卡片特征点划分卡片内部区域和卡片外部区域，获得所述卡片内部区域和所述卡片外部区域之间的区域分割函数；
样本抽取模块，用于以预设比例从所述卡片样本集中抽取迭代训练样本，将所述迭代训练样本中的卡片特征点作为迭代特征点；
模型训练模块，用于通过所述迭代特征点对所述区域分割函数进行迭代训练，并获取训练的区域分割函数的分隔正确率，将所述分隔正确率达到预设阈值的区域分割函数作为预设卡片识别模型。
在一实施例中，所述顶点确定模块40，包括：
方程构建单元，用于基于各所述特征点的特征点坐标建立方程组，其中，所述方程组为：W*A*[y_map,1]＝W*x_map，所述W表示特征值，所述X_map为x轴坐标的特征值，y_map为y轴坐标的特征值，所述A表示直线参数集合；
公式确定单元，用于求解所述方程组获得直线参数计算公式，其中，所述直线参数计算公式为A＝inv(T(Wy)*Wy)*(T(Wy)*Wx)，其中，所述T(x)为x的转置，所述inv(x)为x的逆；
顶点确定单元，用于计算所述直线参数计算公式获得直线参数，根据所述直线参数确定直线方程，通过所述直线方程确定交点坐标，将所述交点坐标作为所述目标图像中卡片的顶点坐标，其中，所述直线方程包括y＝a1x+b1和x＝a2y+b2，其中，所述a1、所述a2、所述b1和所述b2表示直线参数。
在一实施例中，所述顶点确定单元，还用于：
计算所述直线参数计算公式获得直线参数，根据所述直线参数确定直线方程，提取所 述直线方程中的斜率，判断所述斜率之间的乘积是否为负一，或者一个斜率为零且另一个斜率不存在；
若所述斜率之间的乘积为负一，或者一个斜率为零且另一个斜率不存在，则通过所述直线方程确定交点坐标，将所述交点坐标作为所述目标图像中卡片的顶点坐标。
在一实施例中，所述的卡片边框识别装置，包括：
请求接收模块，用于接收卡片信息识别请求，获取所述卡片边框内的卡片图像；
字符获取模块，用于通过光学字符识别算法对所述卡片边框内的卡片图像进行识别，获得所述卡片图像中的字符信息。
在一实施例中，所述的卡片边框识别装置，包括：
信息添加模块，用于根据所述字符信息确定所述目标图像中卡片的卡片类型，将所述字符信息添加到所述卡片类型对应的卡片信息记录表中，并添加标识信息；
查询接收模块，用于在接收到卡片信息查询请求时，获取所述卡片信息查询请求对应的标识信息；
信息获取模块，用于查询所述卡片信息记录表，获取所述卡片标识对应的字符信息并输出。
其中，卡片边框识别装置的各个功能模块实现的步骤可参照本申请卡片边框识别方法的各个实施例，此处不再赘述。
此外，本申请实施例还提出一种计算机存储介质。
所述计算机存储介质上存储有计算机程序，所述计算机程序被处理器执行时实现上述实施例提供的卡片边框识别方法中的操作。
可选的，本申请涉及的存储介质(计算机存储介质)可以是非易失性的，也可以是易失性的。
需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体/操作/对象与另一个实体/操作/对象区分开来，而不一定要求或者暗示这些实体/操作/对象之间存在任何这种实际的关系或者顺序；术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者系统不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者系统所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者系统中还存在另外的相同要素。
对于装置实施例而言，由于其基本相似于方法实施例，所以描述得比较简单，相关之处参见方法实施例的部分说明即可。以上所描述的装置实施例仅仅是示意性的，其中作为分离部件说明的单元可以是或者也可以不是物理上分开的。可以根据实际的需要选择中的部分或者全部模块来实现本申请方案的目的。本领域普通技术人员在不付出创造性劳动的情况下，即可以理解并实施。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在如上所述的一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法。
以上仅为本申请的优选实施例，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种卡片边框识别方法，其中，所述卡片边框识别方法包括以下步骤：
    在接收到图像识别请求时，获取所述图像识别请求对应的目标图像；
    将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息；
    基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标；
    基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标；
    将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果。
  2. 如权利要求1所述的卡片边框识别方法，其中，所述在接收到图像识别请求时，获取所述图像识别请求对应的目标图像的步骤之后，包括：
    对所述目标图像进行切割，得到初步切割图像，对所述初步切割图像进行二值化处理，得到二值化处理图像；
    对所述二值化处理图像进行去噪处理，并完成对所述目标图像的预处理。
  3. 如权利要求1所述的卡片边框识别方法，其中，所述将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息的步骤之前，包括：
    在接收到模型训练指令时，获取所述模型训练指令对应的卡片样本集；
    在所述卡片样本集中的图像样本上标注卡片特征点，根据所述卡片特征点划分卡片内部区域和卡片外部区域，获得所述卡片内部区域和所述卡片外部区域之间的区域分割函数；
    以预设比例从所述卡片样本集中抽取迭代训练样本，将所述迭代训练样本中的卡片特征点作为迭代特征点；
    通过所述迭代特征点对所述区域分割函数进行迭代训练，并获取训练的区域分割函数的分隔正确率，将所述分隔正确率达到预设阈值的区域分割函数作为预设卡片识别模型。
  4. 如权利要求1至3任意一项所述的卡片边框识别方法，其中，所述基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标的步骤，包括：
    基于各所述特征点的特征点坐标建立方程组，其中，所述方程组为：W*A*[y_map,1]＝W*x_map，所述W表示特征值，所述X_map为x轴坐标的特征值，y_map为y轴坐标的特征值，所述A表示直线参数集合；
    求解所述方程组获得直线参数计算公式，其中，所述直线参数计算公式为A＝inv(T(Wy)*Wy)*(T(Wy)*Wx)，其中，所述T(x)为x的转置，所述inv(x)为x的逆；
    计算所述直线参数计算公式获得直线参数，根据所述直线参数确定直线方程，通过所述直线方程确定交点坐标，将所述交点坐标作为所述目标图像中卡片的顶点坐标，其中，所述直线方程包括y＝a1x+b1和x＝a2y+b2，其中，所述a1、所述a2、所述b1和所述b2表示直线参数。
  5. 如权利要求4所述的卡片边框识别方法，其中，所述计算所述直线参数计算公式获得直线参数，根据所述直线参数确定直线方程，通过所述直线方程确定交点坐标，将所述交点坐标作为所述目标图像中卡片的顶点坐标的步骤，包括：
    计算所述直线参数计算公式获得直线参数，根据所述直线参数确定直线方程，提取所述直线方程中的斜率，判断所述斜率之间的乘积是否为负一，或者一个斜率为零且另一个斜率不存在；
    若所述斜率之间的乘积为负一，或者一个斜率为零且另一个斜率不存在，则通过所述直线方程确定交点坐标，将所述交点坐标作为所述目标图像中卡片的顶点坐标。
  6. 如权利要求1所述的卡片边框识别方法，其中，所述将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果的步骤之后，包括：
    接收卡片信息识别请求，获取所述卡片边框内的卡片图像；
    通过光学字符识别算法对所述卡片边框内的卡片图像进行识别，获得所述卡片图像中的字符信息。
  7. 如权利要求6所述的卡片边框识别方法，其中，所述通过光学字符识别算法对所述卡片边框内的卡片图像进行识别，获得所述卡片图像中的字符信息的步骤之后，包括：
    根据所述字符信息确定所述目标图像中卡片的卡片类型，将所述字符信息添加到所述卡片类型对应的卡片信息记录表中，并添加标识信息；
    在接收到卡片信息查询请求时，获取所述卡片信息查询请求对应的标识信息；
    查询所述卡片信息记录表，获取所述卡片标识对应的字符信息并输出。
  8. 一种卡片边框识别装置，其中，所述卡片边框识别装置包括：
    请求接收模块，用于在接收到图像识别请求时，获取所述图像识别请求对应的目标图像；
    图像识别模块，用于将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息；
    坐标构建模块，用于基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标；
    顶点确定模块，用于基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标；
    结果输出模块，用于将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果。
  9. 一种卡片边框识别设备，其中，所述卡片边框识别设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的计算机程序，其中：
    所述计算机程序被所述处理器执行时实现以下步骤：
    在接收到图像识别请求时，获取所述图像识别请求对应的目标图像；
    将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息；
    基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标；
    基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标；
    将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果。
  10. 如权利要求9所述的卡片边框识别设备，其中，所述在接收到图像识别请求时，获取所述图像识别请求对应的目标图像的步骤之后，所述计算机程序被所述处理器执行时还用于实现以下步骤：
    对所述目标图像进行切割，得到初步切割图像，对所述初步切割图像进行二值化处理，得到二值化处理图像；
    对所述二值化处理图像进行去噪处理，并完成对所述目标图像的预处理。
  11. 如权利要求9所述的卡片边框识别设备，其中，所述将经过预处理的目标图像输 入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息的步骤之前，所述计算机程序被所述处理器执行时还用于实现以下步骤：
    在接收到模型训练指令时，获取所述模型训练指令对应的卡片样本集；
    在所述卡片样本集中的图像样本上标注卡片特征点，根据所述卡片特征点划分卡片内部区域和卡片外部区域，获得所述卡片内部区域和所述卡片外部区域之间的区域分割函数；
    以预设比例从所述卡片样本集中抽取迭代训练样本，将所述迭代训练样本中的卡片特征点作为迭代特征点；
    通过所述迭代特征点对所述区域分割函数进行迭代训练，并获取训练的区域分割函数的分隔正确率，将所述分隔正确率达到预设阈值的区域分割函数作为预设卡片识别模型。
  12. 如权利要求9至11任意一项所述的卡片边框识别设备，其中，所述基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标时，具体实现：
    基于各所述特征点的特征点坐标建立方程组，其中，所述方程组为：W*A*[y_map,1]＝W*x_map，所述W表示特征值，所述X_map为x轴坐标的特征值，y_map为y轴坐标的特征值，所述A表示直线参数集合；
    求解所述方程组获得直线参数计算公式，其中，所述直线参数计算公式为A＝inv(T(Wy)*Wy)*(T(Wy)*Wx)，其中，所述T(x)为x的转置，所述inv(x)为x的逆；
    计算所述直线参数计算公式获得直线参数，根据所述直线参数确定直线方程，通过所述直线方程确定交点坐标，将所述交点坐标作为所述目标图像中卡片的顶点坐标，其中，所述直线方程包括y＝a1x+b1和x＝a2y+b2，其中，所述a1、所述a2、所述b1和所述b2表示直线参数。
  13. 如权利要求9所述的卡片边框识别设备，其中，所述将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果的步骤之后，所述计算机程序被所述处理器执行时还用于实现以下步骤：
    接收卡片信息识别请求，获取所述卡片边框内的卡片图像；
    通过光学字符识别算法对所述卡片边框内的卡片图像进行识别，获得所述卡片图像中的字符信息。
  14. 如权利要求13所述的卡片边框识别设备，其中，所述通过光学字符识别算法对所述卡片边框内的卡片图像进行识别，获得所述卡片图像中的字符信息的步骤之后，所述计算机程序被所述处理器执行时还用于实现以下步骤：
    根据所述字符信息确定所述目标图像中卡片的卡片类型，将所述字符信息添加到所述卡片类型对应的卡片信息记录表中，并添加标识信息；
    在接收到卡片信息查询请求时，获取所述卡片信息查询请求对应的标识信息；
    查询所述卡片信息记录表，获取所述卡片标识对应的字符信息并输出。
  15. 一种计算机存储介质，其中，所述计算机存储介质上存储有计算机程序，所述计算机程序被处理器执行时实现以下步骤：
    在接收到图像识别请求时，获取所述图像识别请求对应的目标图像；
    将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息；
    基于所述图像特征信息建立直角坐标系，根据所述直角坐标系获取所述图像特征信息中各特征点的特征点坐标；
    基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标；
    将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输 出包含所述卡片边框的识别结果。
  16. 如权利要求15所述的计算机存储介质，其中，所述在接收到图像识别请求时，获取所述图像识别请求对应的目标图像的步骤之后，所述计算机程序被处理器执行时还用于实现以下步骤：
    对所述目标图像进行切割，得到初步切割图像，对所述初步切割图像进行二值化处理，得到二值化处理图像；
    对所述二值化处理图像进行去噪处理，并完成对所述目标图像的预处理。
  17. 如权利要求15所述的计算机存储介质，其中，所述将经过预处理的目标图像输入至预设卡片识别模型中进行图像识别，获得所述目标图像的图像特征信息的步骤之前，所述计算机程序被处理器执行时还用于实现以下步骤：
    在接收到模型训练指令时，获取所述模型训练指令对应的卡片样本集；
    在所述卡片样本集中的图像样本上标注卡片特征点，根据所述卡片特征点划分卡片内部区域和卡片外部区域，获得所述卡片内部区域和所述卡片外部区域之间的区域分割函数；
    以预设比例从所述卡片样本集中抽取迭代训练样本，将所述迭代训练样本中的卡片特征点作为迭代特征点；
    通过所述迭代特征点对所述区域分割函数进行迭代训练，并获取训练的区域分割函数的分隔正确率，将所述分隔正确率达到预设阈值的区域分割函数作为预设卡片识别模型。
  18. 如权利要求15至17任意一项所述的计算机存储介质，其中，所述基于各所述特征点的特征点坐标建立方程组，求解所述方程组获得所述目标图像中卡片的顶点坐标时，具体实现：
    基于各所述特征点的特征点坐标建立方程组，其中，所述方程组为：W*A*[y_map,1]＝W*x_map，所述W表示特征值，所述X_map为x轴坐标的特征值，y_map为y轴坐标的特征值，所述A表示直线参数集合；
    求解所述方程组获得直线参数计算公式，其中，所述直线参数计算公式为A＝inv(T(Wy)*Wy)*(T(Wy)*Wx)，其中，所述T(x)为x的转置，所述inv(x)为x的逆；
    计算所述直线参数计算公式获得直线参数，根据所述直线参数确定直线方程，通过所述直线方程确定交点坐标，将所述交点坐标作为所述目标图像中卡片的顶点坐标，其中，所述直线方程包括y＝a1x+b1和x＝a2y+b2，其中，所述a1、所述a2、所述b1和所述b2表示直线参数。
  19. 如权利要求15所述的计算机存储介质，其中，所述将每两个顶点坐标之间相互垂直的直线段作为所述目标图像中卡片的卡片边框，并输出包含所述卡片边框的识别结果的步骤之后，所述计算机程序被处理器执行时还用于实现以下步骤：
    接收卡片信息识别请求，获取所述卡片边框内的卡片图像；
    通过光学字符识别算法对所述卡片边框内的卡片图像进行识别，获得所述卡片图像中的字符信息。
  20. 如权利要求19所述的计算机存储介质，其中，所述通过光学字符识别算法对所述卡片边框内的卡片图像进行识别，获得所述卡片图像中的字符信息的步骤之后，所述计算机程序被处理器执行时还用于实现以下步骤：
    根据所述字符信息确定所述目标图像中卡片的卡片类型，将所述字符信息添加到所述卡片类型对应的卡片信息记录表中，并添加标识信息；
    在接收到卡片信息查询请求时，获取所述卡片信息查询请求对应的标识信息；
    查询所述卡片信息记录表，获取所述卡片标识对应的字符信息并输出。
PCT/CN2020/117773 2020-03-16 2020-09-25 卡片边框识别方法、装置、设备和计算机存储介质 Ceased WO2021184718A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010182749.8A CN111507324B (zh) 2020-03-16 2020-03-16 卡片边框识别方法、装置、设备和计算机存储介质
CN202010182749.8 2020-03-16

Publications (1)

Publication Number Publication Date
WO2021184718A1 true WO2021184718A1 (zh) 2021-09-23

Family

ID=71871579

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/117773 Ceased WO2021184718A1 (zh) 2020-03-16 2020-09-25 卡片边框识别方法、装置、设备和计算机存储介质

Country Status (2)

Country Link
CN (1) CN111507324B (zh)
WO (1) WO2021184718A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114279367A (zh) * 2021-12-23 2022-04-05 元气森林(北京)食品科技集团有限公司 陈列柜开门角度的检测方法、陈列柜及存储介质
CN114509013A (zh) * 2022-02-14 2022-05-17 知微行易(上海)智能科技有限公司 一种基于计算机视觉的线缆质检方法
CN117968523A (zh) * 2023-12-19 2024-05-03 深圳市纬迪科技有限公司 基于图像的切割位置监测方法、装置、设备及介质

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111507324B (zh) * 2020-03-16 2024-05-31 平安科技（深圳）有限公司 卡片边框识别方法、装置、设备和计算机存储介质
CN114322849A (zh) * 2020-09-29 2022-04-12 昆山午力电子有限公司 基于图像识别的物品特征定位系统
CN112419336B (zh) * 2020-11-24 2021-09-14 哈尔滨市科佳通用机电股份有限公司 基于小波变换的动车齿轮箱漏油故障识别方法及系统
CN118212188A (zh) * 2024-03-12 2024-06-18 广东兴艺数字印刷股份有限公司 一种动漫卡自动检测控制方法、系统、设备及介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105139378A (zh) * 2015-07-31 2015-12-09 小米科技有限责任公司 卡片边界检测方法及装置
US20160110627A1 (en) * 2013-04-01 2016-04-21 Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirket System and method for describing image outlines
CN108647634A (zh) * 2018-05-09 2018-10-12 深圳壹账通智能科技有限公司 图像边框查找方法、装置、计算机设备及存储介质
CN110059702A (zh) * 2019-03-29 2019-07-26 北京奇艺世纪科技有限公司 一种物体轮廓识别方法及装置
CN111507324A (zh) * 2020-03-16 2020-08-07 平安科技（深圳）有限公司 卡片边框识别方法、装置、设备和计算机存储介质

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106845508B (zh) * 2015-12-07 2019-05-17 腾讯科技（深圳）有限公司 一种检测图像中信息卡的方法、装置和系统
CN106250894B (zh) * 2016-07-26 2021-10-26 北京小米移动软件有限公司 卡片信息识别方法及装置
CN108446621A (zh) * 2018-03-14 2018-08-24 平安科技（深圳）有限公司 票据识别方法、服务器及计算机可读存储介质
CN110008944B (zh) * 2019-02-20 2024-02-13 平安科技（深圳）有限公司 基于模板匹配的ocr识别方法及装置、存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160110627A1 (en) * 2013-04-01 2016-04-21 Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirket System and method for describing image outlines
CN105139378A (zh) * 2015-07-31 2015-12-09 小米科技有限责任公司 卡片边界检测方法及装置
CN108647634A (zh) * 2018-05-09 2018-10-12 深圳壹账通智能科技有限公司 图像边框查找方法、装置、计算机设备及存储介质
CN110059702A (zh) * 2019-03-29 2019-07-26 北京奇艺世纪科技有限公司 一种物体轮廓识别方法及装置
CN111507324A (zh) * 2020-03-16 2020-08-07 平安科技（深圳）有限公司 卡片边框识别方法、装置、设备和计算机存储介质

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114279367A (zh) * 2021-12-23 2022-04-05 元气森林(北京)食品科技集团有限公司 陈列柜开门角度的检测方法、陈列柜及存储介质
CN114509013A (zh) * 2022-02-14 2022-05-17 知微行易(上海)智能科技有限公司 一种基于计算机视觉的线缆质检方法
CN114509013B (zh) * 2022-02-14 2024-06-11 知微行易(上海)智能科技有限公司 一种基于计算机视觉的线缆质检方法
CN117968523A (zh) * 2023-12-19 2024-05-03 深圳市纬迪科技有限公司 基于图像的切割位置监测方法、装置、设备及介质

Also Published As

Publication number Publication date
CN111507324B (zh) 2024-05-31
CN111507324A (zh) 2020-08-07

Similar Documents

Publication Publication Date Title
WO2021184718A1 (zh) 卡片边框识别方法、装置、设备和计算机存储介质
CN112686812B (zh) 银行卡倾斜矫正检测方法、装置、可读存储介质和终端
US12494070B2 (en) Feature density object classification, systems and methods
WO2019218824A1 (zh) 一种移动轨迹获取方法及其设备、存储介质、终端
WO2019169772A1 (zh) 图片处理方法、电子装置及存储介质
CN113627423A (zh) 圆形印章字符识别方法、装置、计算机设备和存储介质
CN111695373B (zh) 斑马线的定位方法、系统、介质及设备
WO2020224221A1 (zh) 跟踪方法、装置、电子设备及存储介质
CN110414649B (zh) Dm码的定位方法、装置、终端及存储介质
CN107918767B (zh) 目标检测方法、装置、电子设备及计算机可读介质
CN111582134B (zh) 证件边沿检测方法、装置、设备和介质
CN111080665B (zh) 图像边框识别方法、装置、设备和计算机存储介质
CN112396060B (zh) 基于身份证分割模型的身份证识别方法及其相关设备
CN113033256B (zh) 一种指尖检测模型的训练方法和设备
CN117234357A (zh) 一种触摸屏多点触摸识别方法、装置及电子设备
WO2025055548A1 (zh) 图像处理方法、装置及相关设备
CN110348326A (zh) 基于身份证识别和多设备访问的家庭健康信息处理方法
CN113469070A (zh) 车载人员数量检测方法、装置和服务器
CN110795995B (zh) 数据处理方法、装置及计算机可读存储介质
CN114283087B (zh) 一种图像去噪方法及相关设备
CN113688839B (zh) 视频处理方法及装置、电子设备、计算机可读存储介质
WO2019071663A1 (zh) 电子装置、虚拟样本生成方法及存储介质
CN117746450A (zh) 一种面向设备运维的移动端定值表单识别方法
WO2020244076A1 (zh) 人脸识别方法、装置、电子设备及存储介质
Thilagavathy et al. Fuzzy based edge enhanced text detection algorithm using MSER

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: 20925990

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: 20925990

Country of ref document: EP

Kind code of ref document: A1