WO2022121021A1 - 一种身份证号码检测方法、装置、可读存储介质和终端 - Google Patents

一种身份证号码检测方法、装置、可读存储介质和终端 Download PDF

Info

Publication number
WO2022121021A1
WO2022121021A1 PCT/CN2020/140583 CN2020140583W WO2022121021A1 WO 2022121021 A1 WO2022121021 A1 WO 2022121021A1 CN 2020140583 W CN2020140583 W CN 2020140583W WO 2022121021 A1 WO2022121021 A1 WO 2022121021A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
area
detection
line
card number
Prior art date
Application number
PCT/CN2020/140583
Other languages
English (en)
French (fr)
Inventor
王晓亮
魏东
杨旭
王丹丹
Original Assignee
广州广电运通金融电子股份有限公司
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 广州广电运通金融电子股份有限公司 filed Critical 广州广电运通金融电子股份有限公司
Publication of WO2022121021A1 publication Critical patent/WO2022121021A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • 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
    • G06V10/24—Aligning, centring, orientation detection or correction of the image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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
    • 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
    • G06V30/42—Document-oriented image-based pattern recognition based on the type of document

Definitions

  • the present invention relates to the technical field of information detection or intelligent vision, and in particular to an ID card number detection method, device, readable storage medium and terminal.
  • image recognition technology is gradually applied in security, military, medical, intelligent transportation and other fields, and technologies such as face recognition and fingerprint recognition are increasingly used in public security, finance, aerospace and other security fields.
  • image recognition is mainly used in the reconnaissance and identification of targets, through automatic image recognition technology to identify and strike enemy targets; in the medical field, various medical image analysis and diagnosis can be carried out through image recognition technology, On the one hand, it can greatly reduce the cost of medical treatment, and on the other hand, it can also help to improve the quality and efficiency of medical care; in the field of transportation, it can not only perform license plate recognition, but also be applied to the cutting-edge field of autonomous driving to achieve a clear view of roads, vehicles and pedestrians. Identify, improve the convenience of life and reduce people's travel costs.
  • ID cards are the most commonly used documents in daily life, and need to quickly and efficiently identify identity information in the fields of security, finance, and enterprise information management. Most of the early ID card information needs to be entered manually, which is very inefficient, and the long-term identification process will also make people's eyes tired, so manual entry is no longer suitable for the current situation of the rapid development of computers and other fields.
  • the technology of automatic identification or automatic extraction of ID card information has emerged, for complex scenes, such as the misalignment of the ID card in the vision, uneven illumination, external light field interference, and debris coverage, etc., the outline of the ID card and the image background are caused. The blurred boundary is not conducive to the accurate extraction of the ID card boundary, resulting in the reduction or failure of ID card number detection.
  • Deep learning method This method uses a large amount of labeled data to train the deep network in the model training stage, fits the network parameters, and realizes the modeling of the OCR (Optical Character Recognition, Optical Character Recognition) detection algorithm.
  • OCR Optical Character Recognition, Optical Character Recognition
  • the image is used as the input of the network, and the character region detection is realized through the network forward reasoning.
  • This method is currently a popular character detection method, but for the identification number detection task, this method has the following defects: (1) The non-document area image also participates in the network reasoning process, which wastes computing resources on the one hand; False detection of characters in the region existence requires additional processing logic to be eliminated; (2) This scheme consumes more computing resources, and the training and reasoning time is longer than this proposal; (3) Due to the inexplicability of the neural network, this method The frame of the positioned character area cannot accurately locate the smallest bounding rectangle of the character, and even cuts off part of the character area. That is, the traditional optical recognition (OCR) technology of document images is mainly used for high-definition scanned images. This method requires the recognized images to have clean Background, use standard print and have high resolution.
  • OCR optical recognition
  • the quality inspection video has extremely poor clarity, the environment is extremely complex, and the quality inspection requirements are diverse.
  • the current algorithms are all to solve a single problem, such as recognizing face and text information from high-quality images. , it is difficult to directly convert into complex business judgment. Therefore, a detection technology that can efficiently identify documents is required in a complex environment where the background of the non-document area is complex or the gradient change in the foreground area of the document is small.
  • the purpose of the present invention is to provide an ID card number detection method, device, readable storage medium and terminal, which can solve the above problems.
  • Design principle First, use the target detection algorithm to locate the position of the ID card, extract the image in the detection frame, normalize the size of the image, and select a mask of appropriate scale to perform morphological operations on the image, so that the ID number area can be connected and automatically Adapt the selection threshold to achieve binarization, apply hough to obtain a straight line, select the main direction of the image and rotate it, and extract the connected area and the ID number area with an aspect ratio that meets the requirements; among them, the Hough transform is a feature extraction (feature extraction). ), which are widely used in image analysis, computer vision and digital image processing. Hough transform is used to identify features in objects, such as lines. His algorithm flow is roughly as follows.
  • the algorithm will perform voting in the parameter space to determine the shape of the object, which is determined by the local accumulator space (accumulator space)
  • the maximum value is determined.
  • the principle is the representation of a straight line: for a straight line in the plane, in the hough transform, consider the representation method: use (r, theta) to represent a straight line. where r is the distance from the line to the origin, and theta is the angle between the vertical line of the line and the x-axis. That is, the parameter representing a straight line in the Hough transform becomes (r, theta).
  • Step 2 Plan the size of the ID card area , according to the position detected in the first step, cut out the ID card image, without changing the original aspect ratio, plan the long side as a fixed size, and plan the short side in the same proportion;
  • the length of the resulting image is L, the width is H, and the image is R;
  • the third step gelatinize the ID number area, design a mask corresponding to the planned size in the second step, and use the mask to perform morphological operations on the
  • the target detection algorithm used in the first step includes one of yolo and ssd.
  • the method for obtaining the inclination angle ⁇ is as follows: applying the hough change to perform line detection on the binarized image in the fourth step, and obtaining a line detection sequence
  • indexmax(P) j,j st
  • the rotation relation adopted in the sixth step is:
  • the dilation and binarization method of the rotated image is as follows:
  • the expansion method is to take a sliding window of size ⁇ +1 to slide on the image R', and the minimum pixel in the sliding window is the output of the position of the expansion method.
  • the implementation is as follows:
  • the inflated image is recorded as:
  • the expansion method is as follows:
  • R′ ij is the sub-image obtained by the sliding window at pixel position i, j
  • the dilated image S is obtained by applying an adaptive threshold algorithm to binarize the image TS.
  • the eighth step extracting the Unicom area whose aspect ratio meets the requirements is the ID card number area.
  • the steps are as follows: row projection and row cutting are performed on the image to obtain the Unicom row, and then perform column cutting for each Unicom row to obtain the Unicom row.
  • the area, the Unicom area and the aspect ratio that meets the prefabricated conditions are the ID number area.
  • the cutting method is as follows: traverse the projection value of the HOR i line line by line. Start line, if there is no end line, no new start line is recorded. If there is a start line, and the sum of the projection values of consecutive h lines is less than the fixed threshold TH, the end start line is recorded, and the start line and the end line are paired appears, note the m-th pair of starting line termination behaviors HS m , HE m :
  • the column projection of the connected row is cut by column to obtain the starting column and the end column of the connected area.
  • the cutting method is as follows: traverse VER mj column-by-column projection value of the connected row and column. Start column, if there is no end column, the new start column is not recorded, if there is a start column, and the sum of the projection values of consecutive w columns connected to the row and column is less than the fixed threshold TW, the end row is recorded, and the mth connection is recorded The n mth pair of the starting column and the ending column of the row
  • Connected area can be obtained by starting row, ending row, starting column and ending column
  • this area is the ID card number area, where T 1 and T 2 are the pre-designed aspect ratio thresholds of the ID card number.
  • the present invention also provides an ID card number detection device, the device includes: an acquisition unit for acquiring image information to be processed; a preprocessing unit for preprocessing the acquired image to obtain a normalized image R; the depth processing unit is used to gelatinize the numbered area in the preprocessed image R, wherein a rectangular mask of the same size as the character to be detected is used to perform an expansion operation on the image after the image planning, and the expansion operation is obtained.
  • the data extraction unit is used to detect the target information of the detected ID card area, and extract the connected area whose aspect ratio meets the requirements to be the ID card
  • the number area is obtained, and the characters in each line are detected in the area, and the area is selected and merged as the ID card number area, so as to extract the number in this area as the ID card number.
  • the present invention also provides a computer-readable storage medium on which computer instructions are stored, and when the computer instructions are executed, the steps of the aforementioned method for detecting an ID card number are executed.
  • the present invention also provides a terminal, comprising a memory and a processor, the memory stores computer instructions that can be run on the processor, and the processor executes the aforementioned method for detecting an ID card number when the processor runs the computer instructions A step of.
  • the beneficial effects of the present invention are: (1) compared with the traditional method, it solves the problem of the failure to detect the certificate number caused by the failure of the edge detection of the certificate, no matter the background of the non-certificate area is complex or the gradient change of the foreground area of the certificate is small, (2) Compared with the character detection algorithm based on deep learning, it improves the operation efficiency of the algorithm, saves computing resources, and does not need to manually design the processing logic of the non-certificate area.
  • the ID card number detection method of the present invention does not require tilt correction of ID cards, can improve ID card information extraction in complex shooting scenarios, greatly improves the recognition rate, and can be widely used in security, finance and other fields.
  • Fig. 1 is the flow chart of the ID card number detection method of the present invention
  • FIG. 2 is a schematic diagram of a structure module of an ID card number detection device.
  • a method for detecting an ID card number see Figure 1, the method includes:
  • the target detection algorithm adopted in the first step includes one of yolo and ssd.
  • the deep neural network model based on the YOLO algorithm is used to detect the collected images and obtain the corresponding ID card area. If the area occupied by the target information is too small, it can prevent the target information from being directly searched from the entire complex scene image. missed inspection.
  • the training process of the deep neural network model based on the YOLO algorithm may specifically include:
  • the ID card image under the complex background containing the label information of the ID card area is obtained as a training sample, and the obtained sample is divided into training set, verification set and test set.
  • preprocess the obtained ID card image under the complex background containing ID card area labeling information to eliminate samples that do not contain valid ID card area, and scale the size of the image to a preset size, such as 484 *484, etc., and perform data enhancement on the acquired samples through operations such as affine transformation, random cropping, and blurring.
  • the network structure of the deep neural network model includes 24 convolutional layers and 2 fully connected layers. Among them, for 24 convolutional layers, 1*1 convolution is mainly used for channel reduction, followed by 3*3 convolution. For the convolutional and fully connected layers, the Leaky ReLU activation function max(x, 0.1x) is used, and the last layer uses a linear activation function.
  • the features of the sample image are extracted through the Convolutional Neural Network (CNN), and a 539 vector is output in the last layer of fully connected (FC), and a 7* is obtained by reshape. 7*11 multidimensional matrix.
  • CNN Convolutional Neural Network
  • FC fully connected
  • 7*11 multidimensional matrix 7*11 multidimensional matrix.
  • each cell in the 7*7 cells can be used to predict the target area
  • "2" in the first 5*2 bits in the following 11 bits means Foreground and background
  • "5" represents the center point cx and cy of the prediction area
  • width w, height h and foreground confidence ⁇ c the last 1 bit of the 11 bits represents the target category confidence.
  • Step 2 Plan the size of the ID card area. According to the position detected in the first step, cut out the ID card image. Without changing the original aspect ratio, plan the long side as a fixed size and the short side as a fixed size. The sides are planned in the same proportion; the length of the planned image is L, the width is H, and the image is R.
  • the third step gelatinize the ID number area, design a mask corresponding to the planned size in the second step, and use the mask to perform morphological operations on the image to achieve the purpose of gelatinizing the image in the text area; set a rectangular mask , perform an expansion operation on the image after the image planning, wherein the mask size is the size of the character to be detected; obtain the corrected image after the expansion operation.
  • the fourth step edge detection, the edge detection image C is obtained by applying canny detection.
  • Step 5 Straight line detection to obtain the inclination angle ⁇ .
  • the acquisition method of the inclination angle ⁇ is as follows: apply the hough change to perform line detection on the binarized image in the fourth step, and obtain the line detection sequence
  • indexmax(P) j,j st
  • Step 6 Image rotation, applying the tilt angle calculated in Step 5 to rotate the identity sign image.
  • the rotation relation used in the sixth step is:
  • the dilation and binarization method of the rotated image is as follows
  • the expansion method is to take a sliding window of size ⁇ +1 to slide on the image R', and the minimum pixel in the sliding window is the output of the position of the expansion method.
  • the implementation is as follows:
  • the inflated image is recorded as:
  • the expansion method is as follows:
  • R′ ij is the sub-image obtained by the sliding window at pixel positions i and j:
  • the dilated image S is obtained by applying an adaptive threshold algorithm to binarize the image TS.
  • the eighth step extracting the Unicom area whose aspect ratio meets the requirements is the ID card number area.
  • the steps are as follows: perform row projection and row cutting on the image, obtain the Unicom row, and then perform column cutting on each row of Unicom row to obtain the Unicom area, Unicom
  • the area whose aspect ratio meets the prefabricated conditions is the ID number area.
  • the cutting method is as follows: traverse the projection value of the HOR i line line by line. Start line, if there is no end line, no new start line is recorded. If there is a start line, and the sum of the projection values of consecutive h lines is less than the fixed threshold TH, the end start line is recorded, and the start line and the end line are paired appears, note the m-th pair of starting line termination behaviors HS m , HE m :
  • the column projection of the connected row is cut by column to obtain the starting column and the end column of the connected area.
  • the cutting method is as follows: traverse VER mj column-by-column projection value of the connected row and column. Start column, if there is no end column, the new start column is not recorded, if there is a start column, and the sum of the projection values of consecutive w columns connected to the row and column is less than the fixed threshold TW, the end row is recorded, and the mth connection is recorded The n mth pair of the starting column and the ending column of the row
  • Connected regions can be obtained by starting row, ending row, starting column, and ending column:
  • this area is the ID card number area, where T 1 and T 2 are the pre-designed aspect ratio thresholds of the ID card number.
  • the image collected in the present invention is an image collected by a camera, which can be a static image (that is, an image collected separately), or an image in a video (that is, an image from a collected video according to a preset standard or random A selected image) can be used as the image source of the ID card of the present invention, and the embodiment of the present invention has no restrictions on all attributes such as the source, nature, size, and the like of the image.
  • embodiments of the present disclosure may also utilize, for example, but not limited to, image processing-based document detection algorithms (eg, edge detection, mathematical morphology, texture analysis-based localization, line detection, and edge detection). Statistical method, genetic algorithm, Hough transform and contour method, method based on wavelet transform, etc.), etc., are used for document detection on the captured image.
  • image processing-based document detection algorithms eg, edge detection, mathematical morphology, texture analysis-based localization, line detection, and edge detection.
  • the neural network when character detection is performed on the collected image by using the neural network, the neural network can be trained by using the sample image in advance, so that the trained neural network can effectively detect characters in the image.
  • the present invention also provides an ID card number detection device, see FIG. 2 , the device includes: an acquisition unit for acquiring image information to be processed; a preprocessing unit for preprocessing the acquired image to obtain a normalized The image R after the image R; the depth processing unit is used to gelatinize the number area in the image R after the preprocessing, in which a rectangular mask of the same size as the size of the character to be detected is used to perform the expansion operation on the image after the image planning , to obtain the corrected image after the expansion operation; then through edge detection, image rotation, expansion and binarization; the data extraction unit is used to detect the target information of the detected ID card area, and extract the connected area whose aspect ratio meets the requirements. That is, the ID card number area, and the characters in each line are detected in the area, and then the area is selected and merged as the ID card number area, so as to extract the number in this area as the ID card number.
  • the representation of the device in the hardware structure is as follows.
  • Acquisition unit - use hardware equipment, including but not limited to mobile phones, IPADs, ordinary cameras, CCD industrial cameras, scanners, etc., to collect image information on the front of the ID card. Note that the collected image should completely include the four borders of the ID card, And the inclination does not exceed plus or minus 20°, and the human eye can distinguish the ID card number.
  • Preprocessing unit depth processing unit, and data extraction unit—Using algorithms, programs, etc. stored in the memory, to perform corresponding processing and data extraction on the obtained images through the processor.
  • the output device including but not limited to the display screen of a tablet computer, computer, mobile phone, etc., outputs and displays the ID card number extracted by the processor.
  • the input/output interface and the network interface connect the acquisition device, the memory, the processor and the output device for signal connection, including electrical connection and telecommunication connection.
  • the present invention also provides a computer-readable storage medium on which computer instructions are stored, and when the computer instructions are executed, the steps of the aforementioned method for detecting an ID card number are executed.
  • the method for detecting the ID card number please refer to the detailed introduction in the foregoing part, and details are not repeated here.
  • Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), Flash Memory or other memory technology, Compact Disc Read Only Memory (CD-ROM), Digital Versatile Disc (DVD) or other optical storage, Magnetic tape cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information that can be accessed by a computing device.
  • computer-readable media does not include transitory computer-readable media, such as modulated data signals and carrier waves.
  • the present invention also provides a terminal, comprising a memory and a processor, the memory stores computer instructions that can be run on the processor, and the processor executes the aforementioned method for detecting an ID card number when the processor runs the computer instructions A step of.
  • a terminal comprising a memory and a processor
  • the memory stores computer instructions that can be run on the processor
  • the processor executes the aforementioned method for detecting an ID card number when the processor runs the computer instructions A step of.
  • the method for detecting the ID card number please refer to the detailed introduction in the foregoing part, and details are not repeated here.
  • the embodiments of the present application may be provided as methods, apparatuses, systems or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
  • computer-usable storage media including, but not limited to, disk storage, CD-ROM, optical storage, etc.

Abstract

提供了一种身份证号码检测方法、装置、可读存储介质和终端，方法首先应用目标检测算法定位身份证位置，提取检测框内图像，对图像进行尺寸归一化，选取适当尺度的掩模对图像进行形态学运算，使身份证号码区域实现联通，自适应选取阈值实现二值化，应用hough获取直线，选择图像主方向并进行旋转，提取长宽比符合要求的联通区域即为身份证号码区域。检测方案无需进行身份证的倾斜矫正，可以提高复杂拍摄场景下身份证信息提取，大大的提高识别率，在安保、金融等领域可得到广泛应用。

Description

一种身份证号码检测方法、装置、可读存储介质和终端 技术领域
本发明涉及信息检测或智能视觉技术领域，具体涉及一种身份证号码检测方法、装置、可读存储介质和终端。
背景技术
随着人工智能的兴起，图像识别技术逐步应用于安全、军事、医疗、智能交通等领域，人脸识别和指纹识别等技术越来越多的使用到公共安全、金融和航空航天等安全领域。在军事领域，图像识别主要应用于目标的侦查和识别，通过自动化的图像识别技术来进行敌方目标的识别并进行打击；在医疗领域，通过图像识别技术可以进行各类医学图像分析和诊断，一方面可以大大降低医疗的成本，另一方面也有助于提高医疗质量和效率；在交通领域不仅可以进行车牌识别，同时也可以应用到前沿的自动驾驶领域，实现对道路、车辆和行人的清晰识别，提高生活的便利并且降低人们出行成本。
对于证件图像识别，其中身份证作为日常生活中最常用的证件，在安防、金融、企事业信息管理领域需要快速高效识别身份信息。早期的身份证的信息大多需要人工录入，效率十分低下，而且长时间的识别过程也会使人眼疲劳，所以人工录入已经不适应于当今计算机等领域飞速发展的现状。虽然已出现了自动识别或自动提取身份证信息的技术，然而对于复杂场景，如证件在视觉内未对准、光照不均、外光场干扰、杂物覆盖等，导致身份证轮廓与图像背景边界模糊，不利于身份证边界的精确提取，从而导致身份 证号码检测效率降低或失败。为此也出现了一些解决方案如下。
传统方法：采用边缘检测算法，应用边缘检测算子定位证件边缘，应用边缘点直线拟合确定证件边缘直线与边缘直线交点信息从而确定证件偏转角度，对证件进行旋转，再应用图像处理方法检测证件号码位置，准确检测证件边缘点是该方法的核心步骤，而边缘检测算子对图像背景复杂程度要求较高，若图像背景前景区域梯度变化小，或背景区域存在大量边缘信息情况下，将导致证件边缘点检测失败，从而无法实现证件号码的检测。
深度学习方法：该方法在模型训练阶段应用大量标注数据对深度网络进行训练，拟合网络参数，实现OCR(Optical Character Recognition，光学字符识别)检测算法的建模，在模型预测阶段，将整张图像作为网络的输入，通过网络前向推理实现字符区域的检测。该方法为目前较为流行的字符检测方法，而对于证件号码检测任务，该方法存在如下缺陷(1)非证件区域图像也参加了网络推理过程，一方面浪费了计算资源，另一方面对于非证件区域存在做的字符存在误检测需要额外增加处理逻辑进行剔除；(2)该方案计算资源消耗较大，相比本提案训练和推理时间长；(3)因神经网络的不可解释行，该方法定位的字符区域边框存无法精确定位字符最小外接矩形框，甚至会切掉部分字符区域，即传统的证件图像光学识别(OCR)技术主要面向高清扫描的图像，该方法要求识别的图像拥有干净的背景、使用规范的印刷体并具有较高的分辨率。但是，自然场景中存在文本背景噪声大、文本分布不规范和自然光源影响等问题，OCR技术在实际自然场景中检测率并不理想，针对身份证等证件识别给后面步骤的字符识别带来压力。
综上，现有的身份证识别技术中，质检视频清晰度极差，环境极其复 杂，质检要求多样，目前算法均是解决单一问题，如从高质量的图像中识别人脸和文字信息，难以直接转化为复杂的业务判断，因此，需要一种在非证件区域背景复杂或证件前景区域梯度变化小的复杂环境下，均可高效识别证件的检测技术。
发明内容
为了克服现有技术的不足，本发明的目的在于提供一种身份证号码检测方法、装置、可读存储介质和终端，其能解决上述问题。
设计原理：首先应用目标检测检测算法定位身份证位置，提取检测框内图像，对图像进行尺寸归一化，选取适当尺度的掩模对图像进行形态学运算，使身份证号码区域实现联通，自适应选取阈值实现二值化，应用hough获取直线，选择图像主方向并进行旋转，提取长宽比符合要求的联通区域及为身份证号码区域；其中，霍夫变换是一种特征检测(feature extraction)，被广泛应用在图像分析(image analysis)、计算机视觉(computer vision)以及数位影像处理(digital image processing)。霍夫变换是用来辨别找出物件中的特征，例如：线条。他的算法流程大致如下，给定一个物件、要辨别的形状的种类，算法会在参数空间(parameter space)中执行投票来决定物体的形状，而这是由累加空间(accumulator space)里的局部最大值(local maximum)来决定。而对于霍夫变换进行直线检测，其原理为直线的表示：对于平面中的一条直线，在hough变换中，考虑表示方式为：使用(r,theta)来表示一条直线。其中r为该直线到原点的距离，theta为该直线的垂线与x轴的夹角。也就是霍夫变换中表示一条直线的参数变成了(r,theta)。.判断多个点是否在同一直线上：当对象变成点时，一个点可以发射出无数条直线，根据 霍夫变换的直线表达形式，假设这个点为i,则通过这个点的直线我们用(ri,thetai)表示。再假设一个点为j，则通过点j的一系列直线我们用(rj,thetaj)表示。两点决定一条直线，所以这两个点的直线必定有ri＝rj,thetai＝thetaj的时候。当是三个点，假设第三个点是k，则通过k点的一系列直线为(rk,thetak)，如果三点在一条直线上，那必定有某个ri＝rj＝rk＝r,thetai＝thetaj＝thetak＝theta。通过霍夫变换检测直线时需要找到这样一样直线，假设有N个点，检测其中的直线，也就是要找到具体的r和theata。对于上面所说的每个点可以通过无数条直线，这里设为n条(通常n＝180)，则一起可以找到Nn个(r,theata)，对这Nn个(r,theata)，可以利用统计学，统计到在theta＝某个值theta_i时，多个点的r近似相等于r_i。也就是说这多个点都在直线(r_i,theta_i)上。
本发明的目的采用以下技术方案实现：
一种身份证号码检测方法，方法包括：第一步：身份证检测，应用目标检测算法检测图像中身份证图像的位置，S＝(a,b,c,d),其中S为身份所在区域，a为该区域左上点横坐标，b为该区域左上点纵坐标，c为该区域右下点横坐标，d为该区域右下点纵坐标；第二步：身份证区域尺寸规一划，根据第一步的检测的位置，将身份证图像裁剪出来，在不改变原有长宽比情况下，将长边规一划为固定尺寸，短边同比例规一划；设规一划后的图像长为L，宽为H，图像为R；第三步：身份证号码区域糊化，设计与第二步规一划尺寸对应的掩模，使用该掩模对图像进行形态学运算，达到文字区域图像联通糊化的目的；设置矩形掩模，对图像规一划化后的图像进行膨胀操作，其中掩模尺寸为待检测字符大小；得到膨胀操作 后的矫正图像；第四步：边缘检测，应用canny检测得到边缘检测图像C；第五步：直线检测获取倾斜角度θ；第六步:图像旋转，应用第五步计算得到的倾斜角度对身份征图像进行旋转；第七步：对旋转图像膨胀和二值化；第八步：提取长宽比符合要求的联通区域即为身份证号码区域，二值化图像，先获取各行字符检测后选区域，再融合各行生产字符候选区域作为身份证号码区域。
优选的，在第一步中采用的目标检测算法包括yolo、ssd中的一种。
优选的，在第五步中，倾斜角度θ获取方法如下：应用hough变化对第四步二值化图像进行直线检测，获得直线检测序列
Figure PCTCN2020140583-appb-000001
Figure PCTCN2020140583-appb-000002
其中l i＝[k i,x 0i,y 0i,x 1i,y 1i,d i]，
Figure PCTCN2020140583-appb-000003
Figure PCTCN2020140583-appb-000004
其中
Figure PCTCN2020140583-appb-000005
Figure PCTCN2020140583-appb-000006
保留长度最长的直线，计算该直线的倾斜角度：
Figure PCTCN2020140583-appb-000007
其中，indexmax(P)＝j,j s.t.
Figure PCTCN2020140583-appb-000008
优选的，第六步中采用的旋转关系为：
Figure PCTCN2020140583-appb-000009
优选的，在第七步中，旋转图像的膨胀和二值化方法如下：
膨胀方法为取大小为δ+1的滑动窗在图像R′上滑动,滑动窗内的最小像素极为膨胀方法的在位置的输出,实现方式如下,
膨胀后的图像记为:
S＝[s i,j] H×W……………………………………………式6，
其中膨胀方法如下:
s i,j＝min(R′ ij)…………………………………………式7，
其中如下R′ ij为滑动窗在i,j像素位置取得的子图像
Figure PCTCN2020140583-appb-000010
对膨胀后的图像S应用自适应阈值算法获得而二化图像TS。
优选的，第八步,提取长宽比符合要求的联通区域即为身份证号码区域,步骤如下,对图像进行行投影并进行行切割,获取联通行再对各行联通行进行列切割,获取联通区域,联通区域宽高比满足预制条件的及为身份证号码区域.
对图像TS＝[ts ij]进行行投影,计算行投影值HOR i
Figure PCTCN2020140583-appb-000011
对行投影进行切割获取联通行,确定各联通行的起始行和终止行,切割方法如下,逐行遍历HOR i行投影值,若连续h行投影值的和大于固定阈值TH,则记录起始行,若没有出现终止行,则不记录新的起始行,若存在起始行,且连续h行投影值和小于固定阈值TH,则记录终止始行,起始行和终止行成对 出现,记第m对起始行终止行为HS m,HE m：
起始行HS m＝i,i st.
Figure PCTCN2020140583-appb-000012
终止行HE m＝i,i st.
Figure PCTCN2020140583-appb-000013
对每一个联通行进行列投影,对第m个联通行进行列投影，
VER mj＝∑ its i,j,i＝HS m…HE m。
对联通行的列投影进列切割获取联通区域的起始列和终止列,切割方法如下,逐列遍历VER mj联通行列投影值,若连续w列联通行列投影值的和大于固定阈值TW,则记录起始列,若没有出现终止列,则不记录新的起始列,若存在起始列,且连续w列联通行列投影值的和小于固定阈值TW,则记录终止行,记第m个联通行的第n m对起始列终止列为
Figure PCTCN2020140583-appb-000014
起始列
Figure PCTCN2020140583-appb-000015
终止列
Figure PCTCN2020140583-appb-000016
由起始行终止行起始列终止列可以获得联通区域
Figure PCTCN2020140583-appb-000017
若满足
Figure PCTCN2020140583-appb-000018
则该区域为身份证号码区域,其中T 1,T 2为预先设计的身份证号码的宽高比阈值。
本发明还提供了一种身份证号码检测装置，装置包括：获取单元，用于获取待处理的图像信息；预处理单元，用于对所获取的图像进行预处理，获 得归一化后的图像R；深度处理单元，用于对预处理后的图像R中号码区域糊化，其中采用与待检测字符大小相同尺寸的矩形掩模对图像规一划化后的图像进行膨胀操作，得到膨胀操作后的矫正图像；再通过边缘检测、图像旋转、膨胀和二值化；数据提取单元，用于对检测出的身份证区域进行目标信息检测，提取长宽比符合要求的联通区域即为身份证号码区域，并对区域内获取各行字符检测后选区域并融合作为身份证号码区域，以此提取此区域内号码作为身份证号码。
本发明还提供了一种计算机可读存储介质，其上存储有计算机指令，所述计算机指令运行时执行前述身份证号码检测方法的步骤。
本发明还提供了一种终端，包括存储器和处理器，所述存储器上储存有能够在所述处理器上运行的计算机指令，所述处理器运行所述计算机指令时执行前述身份证号码检测方法的步骤。
相比现有技术，本发明的有益效果在于：(1)相比传统方法，解决因证件边缘检测失败而导致的证件号码检测失败问题，无论非证件区域背景复杂或证件前景区域梯度变化小，均可有效实现深证好的检测，(2)相较基于深度学习的字符检测算法，提升了算法运算效率，节约运算资源，且无需人工设计非证件区域的处理逻辑。综上，本发明的身份证号码检测方法，无需进行身份证的倾斜矫正，可以提高复杂拍摄场景下身份证信息提取，大大的提高识别率，在安保、金融等领域可得到广泛应用。
附图说明
图1为本发明身份证号码检测方法流程图；
图2为身份证号码检测装置的结构模块示意图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明的一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本发明保护的范围。
第一实施例
一种身份证号码检测方法，参见图1，方法包括：
第一步：身份证检测，应用目标检测算法检测图像中身份证图像的位置，S＝(a,b,c,d),其中S为身份所在区域，a为该区域左上点横坐标，b为该区域左上点纵坐标，c为该区域右下点横坐标，d为该区域右下点纵坐标。
其中，在第一步中采用的目标检测算法包括yolo、ssd中的一种。
对于采用基于YOLO算法的深度神经网络模型对采集的的图像进行检测，得到对应的身份证区域，以在目标信息所占的区域过小时，可以防止直接从整个复杂场景图像中查找目标信息而造成的漏检。其中，所述基于YOLO算法的深度神经网络模型的训练过程，具体可以包括：
首先，获取含有身份证区域标注信息的复杂背景下的身份证图像，作为训练样本，并将所获取的样本划分为训练集、验证集和测试集。接着，对所获取的上述的含有身份证区域标注信息的复杂背景下的身份证图像进行预处理，以剔除不含有有效身份证区域样本，并将图像的尺寸缩放到 预设的大小，如484*484等，并通过仿射变换、随机裁剪、模糊等操作对所获取的样本进行数据增强。
随后，构建深度神经网络模型的网络结构。在本发明一实施例中，所构建的深度神经网络模型的网络结构包含24个卷积层和2个全连接层。其中，对于24个卷积层，主要使用1*1卷积进行信道缩减(channel reduction)，然后紧跟3*3卷积。对于卷积层和全连接层，采用Leaky ReLU激活函数max(x，0.1x)，最后一层采用线性激活函数。
在网络前期通过卷积神经网络(Convolutional Neural Network，CNN)提取样本图像的特征，并在最后一层全连接(Fully connected，FC)输出一个539的向量，通过重塑(reshape)获得一个7*7*11的多维矩阵。其中，所述7*7*11的多维矩阵中，7*7个单元格中的每个单元格均可以用于预测目标区域，之后的11位中的前5*2位中“2”表示前景和背景，“5”分别表示预测区域的中心点cx和cy，宽w、高h和前景的置信度×c，11位中的最后1位表示目标类别置信度。
第二步：身份证区域尺寸规一划，根据第一步的检测的位置，将身份证图像裁剪出来，在不改变原有长宽比情况下，将长边规一划为固定尺寸，短边同比例规一划；设规一划后的图像长为L，宽为H，图像为R。
第三步：身份证号码区域糊化，设计与第二步规一划尺寸对应的掩模，使用该掩模对图像进行形态学运算，达到文字区域图像联通糊化的目的；设置矩形掩模，对图像规一划化后的图像进行膨胀操作，其中掩模尺寸为待检测字符大小；得到膨胀操作后的矫正图像。
第四步：边缘检测，应用canny检测得到边缘检测图像C。
第五步：直线检测获取倾斜角度θ。
在第五步中，倾斜角度θ获取方法如下：应用hough变化对第四步二值化图像进行直线检测，获得直线检测序列
Figure PCTCN2020140583-appb-000019
Figure PCTCN2020140583-appb-000020
其中l i＝[k i,x 0i,y 0i,x 1i,y 1i,d i]，
Figure PCTCN2020140583-appb-000021
Figure PCTCN2020140583-appb-000022
其中
Figure PCTCN2020140583-appb-000023
Figure PCTCN2020140583-appb-000024
保留长度最长的直线，计算该直线的倾斜角度：
Figure PCTCN2020140583-appb-000025
其中，indexmax(P)＝j,j s.t.
Figure PCTCN2020140583-appb-000026
第六步:图像旋转，应用第五步计算得到的倾斜角度对身份征图像进行旋转。第六步中采用的旋转关系为：
Figure PCTCN2020140583-appb-000027
在第七步中，旋转图像的膨胀和二值化方法如下
膨胀方法为取大小为δ+1的滑动窗在图像R′上滑动,滑动窗内的最小像素极为膨胀方法的在位置的输出,实现方式如下,
膨胀后的图像记为:
S＝[s i,j] H×W……………………………………………式6，
其中膨胀方法如下:
s i,j＝min(R′ ij)…………………………………………式7，
其中如下R′ ij为滑动窗在i,j像素位置取得的子图像为：
Figure PCTCN2020140583-appb-000028
对膨胀后的图像S应用自适应阈值算法获得而二化图像TS。
第八步,提取长宽比符合要求的联通区域即为身份证号码区域,步骤如下,对图像进行行投影并进行行切割,获取联通行再对各行联通行进行列切割,获取联通区域,联通区域宽高比满足预制条件的及为身份证号码区域.
对图像TS＝[ts ij]进行行投影,计算行投影值HOR i，
Figure PCTCN2020140583-appb-000029
对行投影进行切割获取联通行,确定各联通行的起始行和终止行,切割方法如下,逐行遍历HOR i行投影值,若连续h行投影值的和大于固定阈值TH,则记录起始行,若没有出现终止行,则不记录新的起始行,若存在起始行,且连续h行投影值和小于固定阈值TH,则记录终止始行,起始行和终止行成对出现,记第m对起始行终止行为HS m,HE m：
起始行HS m＝i,i st.
Figure PCTCN2020140583-appb-000030
终止行HE m＝i,i st.
Figure PCTCN2020140583-appb-000031
对每一个联通行进行列投影,对第m个联通行进行列投影：
VER mj＝∑ its i,j,i＝HS m…HE m。
对联通行的列投影进列切割获取联通区域的起始列和终止列,切割方法如下,逐列遍历VER mj联通行列投影值,若连续w列联通行列投影值的和大于固定阈值TW,则记录起始列,若没有出现终止列,则不记录新的起始列,若存在起始列,且连续w列联通行列投影值的和小于固定阈值TW,则记录终止行,记第m个联通行的第n m对起始列终止列为
Figure PCTCN2020140583-appb-000032
起始列
Figure PCTCN2020140583-appb-000033
终止列
Figure PCTCN2020140583-appb-000034
由起始行终止行起始列终止列可以获得联通区域：
Figure PCTCN2020140583-appb-000035
若满足：
Figure PCTCN2020140583-appb-000036
则该区域为身份证号码区域,其中T 1,T 2为预先设计的身份证号码的宽高比阈值。
本发明中采集的图像，是通过摄像头采集的图像，可以是一张静态图像(即：单独采集的图像)，也可以是一张视频中图像(即从采集的视频中按照预设标准或随机选取的一张图像)，均可用于本发明身份证的图像源，本发明实施例对于图像的来源、性质、大小等等所有属性均无限制。
本领域技术人员基于本公开实施例的记载可以知悉，除了神经网络外，在本公开实施例还可以利用例如但不限于：基于图像处理的字符检测算法(例如，基于直方图粗分割和奇异值特征的字符/号码检测算法，基于二进小波变换的字符/号码检测算法，等等)，对采集图像进行字符检测。另外，除了神经网络外，在本公开实施例也可以利用例如但不限于：基于图像处理的证件检测算法(例如，边缘检测法，数学形态学法，基于纹理 分析的定位方法，行检测和边缘统计法，遗传算法，霍夫(Hough)变换和轮廓线法，基于小波变换的方法，等等)，等等，对采集图像进行证件检测。
本公开实施例中，通过神经网络对采集图像进行字符检测时，可以预先利用样本图像对神经网络进行训练，使得训练好的神经网络能够实现对图像中字符的有效检测。
第二实施例
本发明还提供了一种身份证号码检测装置，参见图2，装置包括：获取单元，用于获取待处理的图像信息；预处理单元，用于对所获取的图像进行预处理，获得归一化后的图像R；深度处理单元，用于对预处理后的图像R中号码区域糊化，其中采用与待检测字符大小相同尺寸的矩形掩模对图像规一划化后的图像进行膨胀操作，得到膨胀操作后的矫正图像；再通过边缘检测、图像旋转、膨胀和二值化；数据提取单元，用于对检测出的身份证区域进行目标信息检测，提取长宽比符合要求的联通区域即为身份证号码区域，并对区域内获取各行字符检测后选区域并融合作为身份证号码区域，以此提取此区域内号码作为身份证号码。
装置在硬件结构化中的表示如下。
获取单元-利用硬件设备，包括但不限于手机，IPAD，普通摄像头，CCD工业相机、扫描仪等，对身份证正面进行图像信息采集，注意采集到的图像应完全的包含身份证的四条边界，并且倾斜不超过正负20°，且人眼能分辨身份证号码。
预处理单元、深度处理单元和数据提取单元-利用存储在存储器中的算 法、程序等，通过处理器对获得的图像进行相应的处理和数据提取。
输出装置，包括但不限于平板电脑、计算机、手机等的显示屏，将处理器提取的身份证号码输出并显示。
输入/输出接口、网接口，将采集装置、存储器、处理器、输出装置进行信号连接，包括电性连接和电讯连接。
第三实施例
本发明还提供了一种计算机可读存储介质，其上存储有计算机指令，所述计算机指令运行时执行前述身份证号码检测方法的步骤。其中，所述的身份证号码检测方法请参见前述部分的详细介绍，此处不再赘述。
本领域普通技术人员可以理解上述实施例的各种方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成，该程序可以存储于计算机可读存储介质中，计算机可读介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括，但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带，磁带磁磁盘存储或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。按照本文中的界定，计算机可读介质不包括暂存电脑可读媒体(transitory media)，如调制的数据信号和载波。
第四实施例
本发明还提供了一种终端，包括存储器和处理器，所述存储器上储存有能够在所述处理器上运行的计算机指令，所述处理器运行所述计算机指令时执行前述身份证号码检测方法的步骤。其中，所述的身份证号码检测方法请参见前述部分的详细介绍，此处不再赘述。
上述方案解决了在复杂背景情况下，身份证轮廓与图像背景边界模糊，不利于身份证号码精确提取的问题。
还需要说明的是，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、商品或者设备不仅包括那些要素，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、商品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个......”限定的要素，并不排除在包括所述要素的过程、方法、商品或者设备中还存在另外的相同要素。
本领域技术人员应明白，本申请的实施例可提供为方法、装置、系统或计算机程序产品。因此，本申请可采用完全硬件实施例、完全软件实施例或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
最后应说明的是：以上实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围。

Claims (10)

  1. 一种身份证号码检测方法，其特征在于，方法包括：
    第一步：身份证检测，应用目标检测算法检测图像中身份证图像的位置，S＝(a,b,c,d),其中S为身份所在区域，a为该区域左上点横坐标，b为该区域左上点纵坐标，c为该区域右下点横坐标，d为该区域右下点纵坐标；
    第二步：身份证区域尺寸规一划，根据第一步的检测的位置，将身份证图像裁剪出来，在不改变原有长宽比情况下，将长边规一划为固定尺寸，短边同比例规一划；设规一划后的图像长为L，宽为H，图像为R；
    第三步：身份证号码区域糊化，设计与第二步规一划尺寸对应的掩模，使用该掩模对图像进行形态学运算，达到文字区域图像联通糊化的目的；设置矩形掩模，对图像规一划化后的图像进行膨胀操作，其中掩模尺寸为待检测字符大小；得到膨胀操作后的矫正图像；
    第四步：边缘检测，应用canny检测得到边缘检测图像C；
    第五步：直线检测获取倾斜角度θ；
    第六步:图像旋转，应用第五步计算得到的倾斜角度对身份征图像进行旋转；
    第七步：对旋转图像膨胀和二值化；
    第八步：提取长宽比符合要求的联通区域即为身份证号码区域，二值化图像，先获取各行字符检测后选区域，再融合各行生产字符候选区域作 为身份证号码区域。
  2. 根据权利要求1所述的身份证号码检测方法，其特征在于：
    在第一步中采用的目标检测算法包括yolo、ssd中的一种。
  3. 根据权利要求1所述的身份证号码检测方法，其特征在于：
    在第五步中，倾斜角度θ获取方法如下：应用hough变化对第四步二值化图像进行直线检测，获得直线检测序列
    Figure PCTCN2020140583-appb-100001
    Figure PCTCN2020140583-appb-100002
    其中
    Figure PCTCN2020140583-appb-100003
    Figure PCTCN2020140583-appb-100004
    其中
    Figure PCTCN2020140583-appb-100005
    Figure PCTCN2020140583-appb-100006
    保留长度最长的直线，计算该直线的倾斜角度：
    Figure PCTCN2020140583-appb-100007
    其中，
    Figure PCTCN2020140583-appb-100008
  4. 根据权利要求1所述的身份证号码检测方法，其特征在于：
    第六步中采用的旋转关系为：
    Figure PCTCN2020140583-appb-100009
  5. 根据权利要求1所述的身份证号码检测方法，其特征在于：
    在第七步中，旋转图像的膨胀和二值化方法如下：
    S＝[s i,j] H×W……………………………………………式6，
    其中膨胀方法如下:
    s i,j＝min(R′ ij)…………………………………………式7，
    其中如下R′ ij为滑动窗在i,j像素位置取得的子图像为：
    Figure PCTCN2020140583-appb-100010
    对膨胀后的图像S应用自适应阈值算法获得而二化图像TS。
  6. 根据权利要求1所述的身份证号码检测方法，其特征在于：
    在第八步中，对二值化后的图像先逐行扫描文字区域起始点和结束点，记各起始点结束点中间的区域为字符连续区域，记录各行连续区域满足先验条件的区域作为各行字符检测候选区域。
  7. 根据权利要求1或6所述的身份证号码检测方法，其特征在于：在第八步中，对二值化图像进行轮廓检测；获得轮廓外接矩形序列，
    Figure PCTCN2020140583-appb-100011
    其中r i＝[x i,y i,w i,h i]，选取宽高比满足如下先验条件的区域即为待检测字符区域，
    S＝[x i,y i,w i,h i],w i/h i＜T………………………………式10，
    行联通起始点s为
    Figure PCTCN2020140583-appb-100012
    行联通终止点e为
    Figure PCTCN2020140583-appb-100013
    若e-s＞T，逐行扫描记录满足条件的边界起始终止点，直至完成整个证件的扫描和识别。
  8. 一种身份证号码检测装置，其特征在于，包括：
    获取单元，用于获取待处理的图像信息；
    预处理单元，用于对所获取的图像进行预处理，获得归一化后的图像R；
    深度处理单元，用于对预处理后的图像R中号码区域糊化，其中采用与待检测字符大小相同尺寸的矩形掩模对图像规一划化后的图像进行膨胀操作，得到膨胀操作后的矫正图像；再通过边缘检测、图像旋转、膨胀和二值化；
    数据提取单元，用于对检测出的身份证区域进行目标信息检测，提取长宽比符合要求的联通区域即为身份证号码区域，并对区域内获取各行字符检测后选区域并融合作为身份证号码区域，以此提取此区域内号码作为身份证号码。
  9. 一种计算机可读存储介质，其上存储有计算机指令，其特征在于，所述计算机指令运行时执行权利要求1至7任一项所述的身份证号码检测方法的步骤。
  10. 一种终端，其特征在于，包括存储器和处理器，所述存储器上储存有能够在所述处理器上运行的计算机指令，所述处理器运行所述计算机指令时执行权利要求1至7任一项所述的身份证号码检测方法的步骤。
PCT/CN2020/140583 2020-12-10 2020-12-29 一种身份证号码检测方法、装置、可读存储介质和终端 WO2022121021A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011455531.1A CN112686247A (zh) 2020-12-10 2020-12-10 一种身份证号码检测方法、装置、可读存储介质和终端
CN202011455531.1 2020-12-10

Publications (1)

Publication Number Publication Date
WO2022121021A1 true WO2022121021A1 (zh) 2022-06-16

Family

ID=75448938

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/140583 WO2022121021A1 (zh) 2020-12-10 2020-12-29 一种身份证号码检测方法、装置、可读存储介质和终端

Country Status (2)

Country Link
CN (1) CN112686247A (zh)
WO (1) WO2022121021A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115804579A (zh) * 2022-11-21 2023-03-17 江西长冈医疗科技有限公司 一种规范化血压测量的过程控制系统及实现方法
CN117315664A (zh) * 2023-09-18 2023-12-29 山东博昂信息科技有限公司 一种基于图像序列的废钢斗号码识别方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7065261B1 (en) * 1999-03-23 2006-06-20 Minolta Co., Ltd. Image processing device and image processing method for correction of image distortion
CN103729645A (zh) * 2013-12-20 2014-04-16 湖北微模式科技发展有限公司 基于单目摄像头的二代证区域定位与提取的方法及装置
CN104050450A (zh) * 2014-06-16 2014-09-17 西安通瑞新材料开发有限公司 一种基于视频的车牌识别方法
CN104834929A (zh) * 2015-04-28 2015-08-12 安徽清新互联信息科技有限公司 一种基于方差特征的车牌校正方法
CN106156712A (zh) * 2015-04-23 2016-11-23 信帧电子技术（北京）有限公司 一种基于自然场景下的身份证号码识别方法与装置

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106446881B (zh) * 2016-07-29 2019-05-21 北京交通大学 从医疗化验单图像中提取化验结果信息的方法
CN108446699A (zh) * 2018-02-08 2018-08-24 东华大学 一种复杂场景下身份证图片信息识别系统
CN108694393A (zh) * 2018-05-30 2018-10-23 深圳市思迪信息技术股份有限公司 一种基于深度卷积的证件图像文本区域提取方法
CN109344727B (zh) * 2018-09-07 2020-11-27 苏州创旅天下信息技术有限公司 身份证文本信息检测方法及装置、可读存储介质和终端
CN110781890A (zh) * 2019-10-25 2020-02-11 上海德启信息科技有限公司 身份证识别方法、装置、电子设备及可读取存储介质
CN111898601A (zh) * 2020-07-14 2020-11-06 浙江大华技术股份有限公司 一种身份证要素提取方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7065261B1 (en) * 1999-03-23 2006-06-20 Minolta Co., Ltd. Image processing device and image processing method for correction of image distortion
CN103729645A (zh) * 2013-12-20 2014-04-16 湖北微模式科技发展有限公司 基于单目摄像头的二代证区域定位与提取的方法及装置
CN104050450A (zh) * 2014-06-16 2014-09-17 西安通瑞新材料开发有限公司 一种基于视频的车牌识别方法
CN106156712A (zh) * 2015-04-23 2016-11-23 信帧电子技术（北京）有限公司 一种基于自然场景下的身份证号码识别方法与装置
CN104834929A (zh) * 2015-04-28 2015-08-12 安徽清新互联信息科技有限公司 一种基于方差特征的车牌校正方法

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115804579A (zh) * 2022-11-21 2023-03-17 江西长冈医疗科技有限公司 一种规范化血压测量的过程控制系统及实现方法
CN115804579B (zh) * 2022-11-21 2024-01-26 江西长冈医疗科技有限公司 一种规范化血压测量的过程控制系统及实现方法
CN117315664A (zh) * 2023-09-18 2023-12-29 山东博昂信息科技有限公司 一种基于图像序列的废钢斗号码识别方法
CN117315664B (zh) * 2023-09-18 2024-04-02 山东博昂信息科技有限公司 一种基于图像序列的废钢斗号码识别方法

Also Published As

Publication number Publication date
CN112686247A (zh) 2021-04-20

Similar Documents

Publication Publication Date Title
CN107609549B (zh) 一种自然场景下证件图像的文本检测方法
CN110414507B (zh) 车牌识别方法、装置、计算机设备和存储介质
US11164027B2 (en) Deep learning based license plate identification method, device, equipment, and storage medium
CN112686812B (zh) 银行卡倾斜矫正检测方法、装置、可读存储介质和终端
EP2783328B1 (en) Text detection using multi-layer connected components with histograms
CN108596166A (zh) 一种基于卷积神经网络分类的集装箱箱号识别方法
CN108108734B (zh) 一种车牌识别方法及装置
CN110969164A (zh) 基于深度学习端对端的低照度成像车牌识别方法及装置
WO2022121021A1 (zh) 一种身份证号码检测方法、装置、可读存储介质和终端
CN110689003A (zh) 低照度成像车牌识别方法、系统、计算机设备及存储介质
CN112686258A (zh) 体检报告信息结构化方法、装置、可读存储介质和终端
KR20220049864A (ko) 인식한 번호판의 각도에 기초하여 차량의 번호를 인식하는 방법
CN111626145A (zh) 一种简捷有效的残缺表格识别及跨页拼接方法
CN111027564A (zh) 基于深度学习一体化的低照度成像车牌识别方法及装置
WO2022121025A1 (zh) 证件增减类别检测方法、装置、可读存储介质和终端
Chang Intelligent text detection and extraction from natural scene images
Huang et al. A back propagation based real-time license plate recognition system
Fang et al. 1-D barcode localization in complex background
CN110633705A (zh) 一种低照度成像车牌识别方法及装置
El Bahi et al. Document text detection in video frames acquired by a smartphone based on line segment detector and dbscan clustering
CN110276260B (zh) 一种基于深度摄像头的商品检测方法
Singh Texture-based real-time character extraction and recognition in natural images
CN116994261B (zh) 一种大数据精准教学智能题卡图像智能识别系统
CN112418210A (zh) 一种杆塔巡检信息智能分类方法
Roy Automatics number plate recognition using convolution neural network

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20964950

Country of ref document: EP

Kind code of ref document: A1