WO2020177470A1 - 验证码识别方法、装置、终端及存储介质 - Google Patents

验证码识别方法、装置、终端及存储介质 Download PDF

Info

Publication number
WO2020177470A1
WO2020177470A1 PCT/CN2019/130580 CN2019130580W WO2020177470A1 WO 2020177470 A1 WO2020177470 A1 WO 2020177470A1 CN 2019130580 W CN2019130580 W CN 2019130580W WO 2020177470 A1 WO2020177470 A1 WO 2020177470A1
Authority
WO
WIPO (PCT)
Prior art keywords
recognition
verification code
preset
image
character image
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/CN2019/130580
Other languages
English (en)
French (fr)
Inventor
王卡风
须成忠
曹廷荣
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.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
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 Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Publication of WO2020177470A1 publication Critical patent/WO2020177470A1/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
    • G06V10/26—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

Definitions

  • the present invention relates to the field of Internet security technology, in particular to a verification code identification method, device, terminal and storage medium.
  • the main steps of the method for identifying numbers, characters and Chinese characters are as follows: First, the interference information in the verification code is removed through image processing techniques such as binarization, spatial filter, and transformation. Secondly, methods such as projection, clustering or target detection are used to determine the position of the character in the image and perform cutting. Third, the feature values of the extracted characters are classified using algorithms such as SVM or convolutional neural network in machine learning. Finally, the value of the final verification code is obtained according to the classification result.
  • image processing techniques such as binarization, spatial filter, and transformation.
  • methods such as projection, clustering or target detection are used to determine the position of the character in the image and perform cutting.
  • the feature values of the extracted characters are classified using algorithms such as SVM or convolutional neural network in machine learning.
  • the value of the final verification code is obtained according to the classification result.
  • none of the existing verification code recognition methods can well solve the problem of the segmentation of pasted characters, resulting in low recognition accuracy of the verification code.
  • the present invention provides a verification code recognition method, device, terminal and storage medium, so as to solve the problem that the pasted characters cannot be well segmented in the existing verification code recognition process, resulting in low verification code recognition accuracy.
  • the present invention provides a verification code identification method, which includes:
  • the steps of preprocessing the verification code image include:
  • Binarize the gray-scale verification code image and then confirm the interference area and/or interference line through the spatial filter, and delete the interference area and/or interference line.
  • the step of calculating the recognition accuracy rate of each character image to be recognized includes:
  • the step of adjusting the preset cutting range and re-cutting, and then training the preset recognition model according to the cutting result until the recognition accuracy exceeds the preset threshold includes:
  • the step of recognizing the character image to be trained through the preset recognition model it further includes:
  • the present invention also provides a verification code identification device, which includes:
  • the acquisition module is used to acquire the verification code image and preprocess the verification code image
  • the cutting module is used to cut the verification code image according to the preset cutting range to obtain multiple character images to be recognized;
  • the recognition module is used to recognize multiple character images to be recognized through a preset recognition model
  • the calculation module is used to calculate the recognition accuracy of each character image to be recognized
  • the training module is used to adjust the preset cutting range and cut again when the recognition accuracy rate does not exceed the preset threshold, and then train the preset recognition model according to the cutting result until the recognition accuracy exceeds the preset threshold.
  • the acquisition module includes:
  • the image acquisition unit is used to acquire the verification code image
  • Graying unit used for graying the verification code image
  • the denoising unit is used to binarize the gray-scale verification code image, confirm the interference area and/or interference line through the spatial filter, and delete the interference area and/or interference line.
  • the calculation module includes:
  • the statistical unit is used to count the number of correct recognition and the number of incorrect recognition of each character image to be recognized;
  • the training module includes:
  • the adjustment unit is used to adjust the preset cutting range
  • the re-cutting unit is used to use the U-Net network to cut the verification code image according to the adjusted preset cutting range to obtain the character image to be trained;
  • the recognition and calculation unit is used to recognize the character image to be trained through a preset recognition model, and calculate the recognition accuracy of the character image to be trained according to the recognition result;
  • the loop recognition unit is used to readjust the preset cutting range when the recognition accuracy of the character image to be trained does not exceed the preset threshold, and perform cutting and recognition again until the recognition accuracy of the character image to be trained exceeds the preset threshold.
  • the training module also includes:
  • the occlusion unit is used to extract the depth features of the character image to be trained and perform random occlusion of the depth features.
  • the present invention also provides a terminal, which includes a memory and a processor, the processor is coupled to the memory, and a computer program that can run on the processor is stored in the memory;
  • the present invention also provides a storage medium on which a computer program is stored.
  • the computer program is executed by a processor, the steps in any one of the above verification code identification methods are implemented.
  • the present invention recognizes the cut characters through a preset recognition model, and calculates the recognition accuracy rate.
  • the recognition accuracy rate is lower than the preset threshold
  • the verification code is confirmed as a verification code that is difficult to recognize.
  • the characters cannot be divided well because they are pasted together, confirm that the verification code is a verification code that is difficult to recognize.
  • the preset recognition model is trained until the recognition accuracy exceeds the preset threshold, so as to achieve the purpose of enhancing the recognition ability of the preset recognition model and improve the recognition accuracy of the preset recognition model.
  • FIG. 1 is a flowchart of the first embodiment of the verification code identification method of the present invention
  • FIG. 2 is a flowchart of a second embodiment of the verification code identification method of the present invention.
  • FIG. 3 is a flowchart of a third embodiment of the verification code identification method of the present invention.
  • FIG. 4 is a flowchart of a fourth embodiment of the verification code identification method of the present invention.
  • FIG. 5 is a flowchart of a fifth embodiment of the verification code identification method of the present invention.
  • FIG. 6 is a schematic diagram of functional modules of the first embodiment of the verification code identification device of the present invention.
  • FIG. 7 is a schematic diagram of functional modules of the second embodiment of the verification code identification device of the present invention.
  • FIG. 8 is a schematic diagram of functional modules of a third embodiment of the verification code identification device of the present invention.
  • FIG. 9 is a schematic diagram of functional modules of a fourth embodiment of the verification code identification device of the present invention.
  • FIG. 10 is a schematic diagram of functional modules of a fifth embodiment of the verification code identification device of the present invention.
  • FIG. 11 is a schematic diagram of a framework of an embodiment of a terminal according to the present invention.
  • FIG. 1 shows an embodiment of the verification code identification method of the present invention.
  • the verification code identification method includes the following steps:
  • Step S1 obtaining a verification code image, and preprocessing the verification code image.
  • the verification code image is preprocessed to remove noise in the verification code image and reduce the influence of noise on the verification code recognition.
  • step S1 the step of preprocessing the verification code image specifically includes the following steps:
  • Step S10 Perform a grayscale operation on the verification code image.
  • the color image of the verification code image is gray-scaled by the component method, specifically the brightness of the three components in the color image is used as the gray value of the three gray-scale images, and a gray-scale can be selected according to the application needs.
  • Image see the following formula:
  • Step S11 Binarize the gray-scale verification code image, confirm the interference area and/or interference line through the spatial filter, and delete the interference area and/or interference line.
  • binarization specifically refers to an image containing target objects, background and noise.
  • the most frequently used method is to set a global threshold. T, use T to divide the image data into two parts: the pixel group larger than T and the pixel group smaller than T.
  • the pixel value of the pixel group larger than T is set to white (or black), and the pixel value of the pixel group smaller than T is set to black (or white).
  • the spatial filter consists of a neighborhood (usually a small rectangle) and a predefined operation performed on the image pixels surrounded by the neighborhood. Filtering produces a new pixel. The coordinates of the new pixel are equal to the coordinates of the center of the neighborhood. The value of the pixel is the result of the filtering operation. After the center of the spatial filter accesses each pixel in the input image, a processed (filtered) image is generated.
  • the noise in the verification code image is removed by a spatial filter to facilitate subsequent image cutting processing, and it can also achieve better results when pasting characters. Cutting effect.
  • Step S2 cutting the verification code image according to the preset cutting range to obtain multiple character images to be recognized.
  • the preset cutting range is the width of one character in the verification code image. After the preset cutting range is obtained, the character area of the verification code image is cut according to the preset cutting range, so as to obtain a plurality of character images to be recognized.
  • Step S3 Recognizing a plurality of character images to be recognized through a preset recognition model.
  • the preset recognition model is obtained by training according to the convolutional neural network, a plurality of character images to be recognized are recognized through the preset recognition model, and the recognition result is obtained and recorded.
  • the cutting effect is better.
  • the image of the characters to be recognized has less noise and is easy to recognize.
  • the pasted part of the characters will have more Too much noise may cause the recognition error of the preset recognition model. Therefore, each character image to be recognized has a recognition accuracy rate and a recognition error rate.
  • Step S4 Calculate the recognition accuracy of each character image to be recognized.
  • the recognition accuracy rate of each character image to be recognized is calculated according to the recognition result obtained by the preset recognition model.
  • step S4 includes the following steps:
  • Step S40 Count the number of correct recognition times and the number of incorrect recognition times of each character image to be recognized.
  • the preset recognition model recognizes the character image to be recognized, it is determined whether the recognition is correct. If the recognition is correct, the number of correct recognitions is recorded once, and the number of incorrect recognitions is recorded once if the recognition is wrong.
  • Step S5 When the recognition accuracy rate does not exceed the preset threshold, adjust the preset cutting range and cut again, and then train the preset recognition model according to the cutting result until the recognition accuracy exceeds the preset threshold.
  • the recognition accuracy rate of the character image to be recognized does not exceed the preset threshold, it means that the character image to be recognized is difficult to recognize.
  • the preset recognition model is trained until the recognition accuracy rate of the to-be-recognized character image exceeds the preset threshold.
  • the preset threshold is preferably 70%.
  • step S5 includes the following steps:
  • Step S50 adjusting the preset cutting range.
  • the cutting range can be adjusted according to (a,b), (a,b+ ⁇ ), (a- ⁇ ,b), (a- ⁇ ,b+ ⁇ ), where (a , b) is the width of a character, and ⁇ is one tenth of the width of a character. Moreover, the adjustment of the cutting range only changes the width in one dimension, thus avoiding many invalid calculations.
  • Step S51 using the U-Net network to cut the verification code image according to the adjusted preset cutting range to obtain the character image to be trained.
  • step S52 the character image to be trained is recognized through a preset recognition model, and the recognition accuracy of the character image to be trained is calculated according to the recognition result.
  • the character image to be trained is recognized again through the preset recognition model, and the recognition accuracy of the character image to be trained is calculated again.
  • Step S53 When the recognition accuracy of the character image to be trained does not exceed the preset threshold, readjust the preset cutting range, and perform cutting and recognition again until the recognition accuracy of the character image to be trained exceeds the preset threshold.
  • the recognition accuracy rate of the character image to be trained does not exceed the preset threshold, it means that the character image to be trained cut according to the adjusted preset cutting range still cannot be well recognized.
  • adjust the preset again Set the cutting range for cutting, and perform the recognition operation and the operation of calculating the recognition accuracy again, and loop in turn until the recognition accuracy of the character image to be trained exceeds the preset threshold, indicating that the preset recognition model has been able to recognize the current cutting well
  • the range is the optimal cutting range and the optimal cutting range is recorded. When this type of character image is encountered in the subsequent verification code recognition process, the cutting and recognition can be performed according to the optimal cutting range.
  • step S52 the following steps are further included before step S52:
  • Step S60 Extract the depth features of the character image to be trained, and perform random occlusion on the depth features.
  • the cut characters are recognized by the preset recognition model, and the recognition accuracy rate is calculated.
  • the recognition accuracy rate is lower than the preset threshold
  • the verification code is confirmed as a verification code that is difficult to recognize.
  • it is pasted together and cannot be divided well confirm that the verification code is a verification code that is difficult to recognize.
  • the cutting range After adjusting the cutting range, perform cutting again, and use the cut verification code to train the preset recognition model , Until the recognition accuracy exceeds the preset threshold, so as to achieve the purpose of enhancing the recognition ability of the preset recognition model and improve the recognition accuracy of the preset recognition model.
  • FIG. 6 shows an embodiment of the verification code identification device of the present invention.
  • the verification code identification device includes an acquisition module 10, a cutting module 11, an identification module 12, a calculation module 13 and a training module 14.
  • the obtaining module 10 is used to obtain the verification code image and preprocess the verification code image; the cutting module 11 is used to cut the verification code image according to the preset cutting range to obtain multiple character images to be recognized; recognition module 12. It is used to recognize multiple character images to be recognized through a preset recognition model; the calculation module 13 is used to calculate the recognition accuracy of each character image to be recognized; the training module 14 is used when the recognition accuracy does not exceed the preset When setting the threshold, adjust the preset cutting range and cut again, and then train the preset recognition model according to the cutting result until the recognition accuracy exceeds the preset threshold.
  • the acquisition module 10 includes an image acquisition unit 100, a grayscale unit 101 and a denoising unit 102.
  • the image acquisition unit 100 is used to obtain a verification code image; the gray-scale unit 101 is used to perform a gray-scale operation on the verification code image; and the denoising unit 102 is used to perform binary operations on the gray-scale verification code image
  • the spatial filter is used to confirm the interference area and/or interference line, and delete the interference area and/or interference line.
  • the calculation module 13 includes a statistics unit 130 and a calculation unit 131.
  • the training module 14 includes an adjustment unit 140, a re-cutting unit 141, a recognition and calculation unit 142, and a cycle recognition unit 143.
  • the adjustment unit 140 is used to adjust the preset cutting range; the re-cutting unit 141 is used to use the U-Net network to cut the verification code image according to the adjusted preset cutting range to obtain the character image to be trained; recognition and calculation The unit 142 is used to recognize the character image to be trained using a preset recognition model, and to calculate the recognition accuracy of the character image to be trained according to the recognition result; the loop recognition unit 143 is used to recognize when the recognition accuracy of the character image to be trained does not exceed the preset When setting the threshold, readjust the preset cutting range, and perform cutting and recognition again until the recognition accuracy of the character image to be trained exceeds the preset threshold.
  • the training module 14 further includes an occlusion unit 144 for extracting the depth features of the character image to be trained and randomly occluding the depth features.
  • Fig. 11 shows a schematic block diagram of a terminal provided by another embodiment of the present invention.
  • the terminal in this embodiment includes: one or at least two processors 80, a memory 81, and the A computer program 810 running on the processor 80.
  • the processor 80 executes the computer program 810, it implements the steps in the verification code identification method described in the above embodiment, for example: step S1-step S5 shown in FIG. 1.
  • the processor 80 executes the computer program 810, the functions of the modules/units in the above-mentioned embodiment of the precipitation forecasting device based on multi-mode integration are realized, for example: the functions of the module 10-module 14 shown in FIG.
  • the computer program 810 may be divided into one or more modules/units, and the one or more modules/units are stored in the memory 81 and executed by the processor 80 to complete the application.
  • One or more modules/units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 810 in the terminal.
  • the terminal includes but is not limited to a processor 80 and a memory 81.
  • FIG. 8 is only an example of the terminal, and does not constitute a limitation on the terminal. It may include more or less components than shown in the figure, or combine some components, or different components, such as a terminal. It can also include input devices, output devices, network access devices, buses, and so on.
  • the processor 80 can be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), ready-made Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • the memory 81 can be a read-only memory, a static storage device that can store static information and instructions, a random access memory, or a dynamic storage device that can store information and instructions, or it can be an electrically erasable programmable read-only memory or a read-only optical disk. , Or other optical disk storage, optical disk storage, magnetic disk storage media or other magnetic storage devices.
  • the memory 81 and the processor 80 may be connected through a communication bus, or may be integrated with the processor 80.
  • the disclosed device and method may be implemented in other ways.
  • the device embodiments described above are only illustrative, for example, the division of modules or units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or It can be integrated into another device, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • each unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.
  • the embodiment of the present application also provides a storage medium for storing a computer program, which contains program data designed for executing the above-mentioned verification code identification method embodiment of the present application.
  • a storage medium for storing a computer program, which contains program data designed for executing the above-mentioned verification code identification method embodiment of the present application.
  • the integrated module/unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, this application implements all or part of the processes in the above-mentioned embodiment methods, and can also be completed by instructing relevant hardware through a computer program 810.
  • the computer program 810 can be stored in a computer-readable storage medium. When executed by the processor 80, 810 can implement the steps of the foregoing method embodiments.
  • the computer program 810 includes computer program code, and the computer program code may be in the form of source code, object code, executable file, or some intermediate forms.
  • the computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access Memory (RAM, Random Access Memory), electric carrier signal, telecommunications signal, software distribution medium, etc.
  • ROM Read-Only Memory
  • RAM random access Memory
  • electric carrier signal telecommunications signal
  • software distribution medium etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Character Discrimination (AREA)

Abstract

一种验证码识别方法、装置、终端及存储介质。该方法包括：获取验证码图像，并对验证码图像进行预处理（S1）；根据预设切割范围对验证码图像进行切割，得到多个待识别字符图像（S2）；通过预设识别模型对多个待识别字符图像进行识别（S3）；计算每个待识别字符图像的识别准确率（S4）；当识别准确率未超过预设阈值时，调整预设切割范围并重新切割，再根据切割结果训练预设识别模型，直至识别准确率超过预设阈值（S5）。该方法在验证码识别过程中，将难以识别的验证码重新切割并对预设识别模型进行训练，从而使得预设识别模型的识别能力更强，提升预设识别模型的识别准确率。

Description

验证码识别方法、装置、终端及存储介质 技术领域
本发明涉及互联网安全技术领域，尤其涉及一种验证码识别方法、装置、终端及存储介质。
背景技术
在人机智能交互中，尤其是让机器模拟人的行为，来智能识别验证码是一项重要的技术。验证码有多种类型，基于图像的验证码是最为广泛使用的技术之一。那么让机器自动识别图像验证码就是完成机器模拟人的一项基础技术之一，基于图像的验证码识别一般要对字符进行切割，然后使用机器学习或者深度学习的方法对字符进行分类。字符切割的准确程度直接影响到最终预测结果的好坏。在实际中验证码都是随机生成的，不同的生成算法得到的验证码有不同的切割方法。
现阶验证码种类繁多，对于数字、字符和汉字类型验证码识别的方法主要步骤如下：首先，通过二值化、空间滤波器、变换等图像处理技术去除验证码中的干扰信息。其次，使用投影、类聚或目标检测等方法确定字符在图像中的位置并进行切割。再次，提取出字符的特征值利用机器学习中的SVM或卷积神经网络等算法进行分类。最后，根据分类的结果得到最终验证码的值。但是，现有的验证码识别方法都没有很好的解决黏贴字符分割的问题，导致验证码的识别准确率不高。
发明内容
本发明提供了一种验证码识别方法、装置、终端及存储介质，以解决现有验证码识别过程中黏贴字符不能很好的分割，导致验证码识别准确率不高的问题。
为了解决上述问题，本发明提供了一种验证码识别方法，其包括：
获取验证码图像，并对验证码图像进行预处理；
根据预设切割范围对验证码图像进行切割，得到多个待识别字符图像；
通过预设识别模型对多个待识别字符图像进行识别；
计算每个待识别字符图像的识别准确率；
当识别准确率未超过预设阈值时，调整预设切割范围并重新切割，再根据切割结果训练预设识别模型，直至识别准确率超过预设阈值。
作为本发明的进一步改进，对验证码图像进行预处理的步骤包括：
将验证码图像进行灰度化操作；
对灰度化的验证码图像进行二值化处理，再通过空间滤波器确认干扰区域和/或干扰线条，并删除干扰区域和/或干扰线条。
作为本发明的进一步改进，计算每个待识别字符图像的识别准确率的步骤，包括：
统计每个待识别字符图像的识别正确次数和识别错误次数；
计算每个待识别字符图像的识别准确率，识别准确率＝识别正确次数/(识别正确次数+识别错误次数)。
作为本发明的进一步改进，调整预设切割范围并重新切割，再根据切割结果训练预设识别模型，直至识别准确率超过预设阈值的步骤，包括：
调整预设切割范围；
利用U-Net网络根据调整后的预设切割范围对验证码图像进行切割，得到待训练字符图像；
通过预设识别模型对待训练字符图像进行识别，并根据识别结果计算待训练字符图像的识别准确率；
当待训练字符图像的识别准确率未超过预设阈值时，重新调整预设切割范围，并再次进行切割和识别，直至待训练字符图像的识别准确率超过预设阈值。
作为本发明的进一步改进，通过预设识别模型对待训练字符图像进行识别的步骤之前，还包括：
提取待训练字符图像的深度特征，并对深度特征进行随机遮挡。
为了解决上述问题，本发明还提供了一种验证码识别装置，其包括：
获取模块，用于获取验证码图像，并对验证码图像进行预处理；
切割模块，用于根据预设切割范围对验证码图像进行切割，得到多个待识 别字符图像；
识别模块，用于通过预设识别模型对多个待识别字符图像进行识别；
计算模块，用于计算每个待识别字符图像的识别准确率；
训练模块，用于当识别准确率未超过预设阈值时，调整预设切割范围并重新切割，再根据切割结果训练预设识别模型，直至识别准确率超过预设阈值。
作为本发明的进一步改进，获取模块包括：
图像获取单元，用于获取验证码图像；
灰度化单元，用于将验证码图像进行灰度化操作；
去噪单元，用于对灰度化的验证码图像进行二值化处理，再通过空间滤波器确认干扰区域和/或干扰线条，并删除干扰区域和/或干扰线条。
作为本发明的进一步改进，计算模块包括：
统计单元，用于统计每个待识别字符图像的识别正确次数和识别错误次数；
计算单元，用于计算每个待识别字符图像的识别准确率，识别准确率＝识别正确次数/(识别正确次数+识别错误次数)。
作为本发明的进一步改进，训练模块包括：
调整单元，用于调整预设切割范围；
重新切割单元，用于利用U-Net网络根据调整后的预设切割范围对验证码图像进行切割，得到待训练字符图像；
识别与计算单元，用于通过预设识别模型对待训练字符图像进行识别，并根据识别结果计算待训练字符图像的识别准确率；
循环识别单元，用于当待训练字符图像的识别准确率未超过预设阈值时，重新调整预设切割范围，并再次进行切割和识别，直至待训练字符图像的识别准确率超过预设阈值。
作为本发明的进一步改进，训练模块还包括：
遮挡单元，用于提取待训练字符图像的深度特征，并对深度特征进行随机遮挡。
为了解决上述问题，本发明还提供了一种终端，其包括存储器和处理器，处理器耦接存储器，存储器上存储有可在处理器上运行的计算机程序；
处理器执行计算机程序时，实现上述任一项验证码识别方法中的步骤。
为了解决上述问题，本发明还提供了一种存储介质，其上存储有计算机程 序，计算机程序被处理器执行时，实现上述任一项验证码识别方法中的步骤。
相比于现有技术，本发明通过预设识别模型识别切割后的字符，并计算识别准确率，当识别准确率低于预设阈值时，确认该验证码为难以识别的验证码，通过将该组，当字符因为黏贴在一起而不能很好的进行分割时，确认该验证码为难以识别的验证码，此时，调整切割范围后，重新进行切割，并采用切割后的验证码对预设识别模型进行训练，直至识别准确率超过预设阈值时为止，从而达到增强预设识别模型识别能力的目的，提高了预设识别模型的识别准确率。
附图说明
图1为本发明验证码识别方法第一个实施例的流程图；
图2为本发明验证码识别方法第二个实施例的流程图；
图3为本发明验证码识别方法第三个实施例的流程图；
图4为本发明验证码识别方法第四个实施例的流程图；
图5为本发明验证码识别方法第五个实施例的流程图；
图6为本发明验证码识别装置第一个实施例的功能模块示意图；
图7为本发明验证码识别装置第二个实施例的功能模块示意图；
图8为本发明验证码识别装置第三个实施例的功能模块示意图；
图9为本发明验证码识别装置第四个实施例的功能模块示意图；
图10为本发明验证码识别装置第五个实施例的功能模块示意图；
图11为本发明终端一个实施例的框架示意图。
具体实施方式
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用来限定本发明。
图1展示了本发明验证码识别方法的一个实施例。如图1所示，本实施例中， 该验证码识别方法包括以下步骤：
步骤S1，获取验证码图像，并对验证码图像进行预处理。
具体地，获取验证码图像之后，对验证码图像进行预处理，以去除验证码图像中的噪声，降低噪声对验证码识别的影响。
在一些实施例中，如图2所示，步骤S1中，对验证码图像进行预处理的步骤，具体包括以下步骤：
步骤S10，将验证码图像进行灰度化操作。
具体地，将验证码图像的彩色图像通过分量法进行灰度化处理，具体为将彩色图像中的三分量的亮度作为三个灰度图像的灰度值，可根据应用需要选取一种灰度图像，参见下述公式：
Gray 1(i，j)＝R(i，j)；
Gray 2(i，j)＝G(i，j)；
Gray 3(i，j)＝B(i，j)；
其中，Gray n(i，j)(n＝1,2,3)为转换后的灰度图像在(i，j)处的灰度值，R、G、B分别表示彩色图像中像素点红、绿、蓝的像素值。
步骤S11，对灰度化的验证码图像进行二值化处理，再通过空间滤波器确认干扰区域和/或干扰线条，并删除干扰区域和/或干扰线条。
需要说明的是，二值化具体是指一幅图像包含目标物体、背景还有噪声，要想从多值的数字图像中直接提取出目标物体，最经常使用的方法就是设定一个全局的阈值T，用T将图像的数据分成两部分：大于T的像素群和小于T的像素群。将大于T的像素群的像素值设定为白色(或者黑色)，小于T的像素群的像素值设定为黑色(或者白色)。例如：计算每个像素的(R+G+B)/3，假设>127，则设置该像素为白色，即R＝G＝B＝255；否则设置为黑色，即R＝G＝B＝0。从而完成对图像的二值化处理。空间滤波器由一个邻域(通常是一个较小的矩形)、对该邻域所包围图像像素执行的预定义操作组成。滤波产生一个新像素，新像素的坐标等于邻域中心的坐标，像素的值是滤波操作的结果。空间滤波器的中心访问输入图像中的每个像素后，就生成了处理(滤波)之后的图像。
具体地，本实施例在对验证码图像进行二值化处理之后，通过空间滤波器去除验证码图像中的噪声，方便后续进行图像切割处理，在针对黏贴字符时，也能够达到较好的切割效果。
步骤S2，根据预设切割范围对验证码图像进行切割，得到多个待识别字符图像。
具体地，通常地该预设切割范围为验证码图像中一个字符的宽度。在获取预设切割范围之后，根据该预设切割范围对验证码图像的字符区域进行切割，从而得到多个待识别字符图像。
步骤S3，通过预设识别模型对多个待识别字符图像进行识别。
具体地，该预设识别模型根据卷积神经网络训练得到，通过该预设识别模型对多个待识别字符图像进行识别，得到识别结果并进行记录。针对于未黏贴在一起的字符，其切割效果较好，待识别字符图像中的噪声较少，容易识别，而针对于黏贴在一起的字符，切割之后，字符的黏贴部分会存在较多的噪声，可能会导致预设识别模型识别错误。因此，每个待识别字符图像均存在识别准确率和识别错误率。
步骤S4，计算每个待识别字符图像的识别准确率。
具体地，根据预设识别模型得到的识别结果来计算每个待识别字符图像的识别准确率。
在一些实施例中，如图3所示，步骤S4包括以下步骤：
步骤S40，统计每个待识别字符图像的识别正确次数和识别错误次数。
具体地，在预设识别模型对待识别字符图像进行识别之后，判断是否识别正确，若识别正确，则记录一次识别正确次数，若识别错误，则记录一次识别错误次数。
步骤S41，计算每个待识别字符图像的识别准确率，识别准确率＝识别正确次数/(识别正确次数+识别错误次数)。
步骤S5，当识别准确率未超过预设阈值时，调整预设切割范围并重新切割，再根据切割结果训练预设识别模型，直至识别准确率超过预设阈值。
具体地，当待识别字符图像的识别准确率未超过预设阈值时，说明该待识别字符图像较难识别，此时，为了增强预设识别模型的识别能力，则将该较难识别的待识别字符图像作为样本，对预设识别模型进行训练，直至该待识别字符图像的识别准确率超过预设阈值时为止。而当待识别字符图像的识别准确率超过预设阈值时，说明该待识别字符图像的识别准确率高，不需要再进行训练。
需要说明的是，该预设阈值优选为70％。
在一些实施例中，如图4所示，步骤S5包括以下步骤：
步骤S50，调整预设切割范围。
具体地，本实施例中，切割范围可按照(a,b),(a,b+δ),(a-δ,b)，(a-δ,b+δ)来进行调整，其中(a,b)为一个字符的宽度，δ为一个字符宽度的十分之一。并且，切割范围的调整仅在一个维度上进行宽度变化，因此避免了很多无效计算。
步骤S51，利用U-Net网络根据调整后的预设切割范围对验证码图像进行切割，得到待训练字符图像。
具体地，使用labelme软件对验证码图像进行标记后，按照调整后的预设切割范围进行切割得到待训练字符图像。
步骤S52，通过预设识别模型对待训练字符图像进行识别，并根据识别结果计算待训练字符图像的识别准确率。
具体地，通过预设识别模型再次去识别该待训练字符图像，并再次计算待训练字符图像的识别准确率。
步骤S53，当待训练字符图像的识别准确率未超过预设阈值时，重新调整预设切割范围，并再次进行切割和识别，直至待训练字符图像的识别准确率超过预设阈值。
具体地，当待训练字符图像的识别准确率未超过预设阈值时，说明根据调整后的预设切割范围切割得到的待训练字符图像仍然不能够很好的进行识别，此时，再次调整预设切割范围进行切割，并再次执行识别操作和计算识别准确率操作，依次循环，直至待训练字符图像的识别准确率超过预设阈值时，说明预设识别模 型已经能够很好的识别当前的切割范围为最佳切割范围并记录该最佳切割范围，在后续验证码识别过程中遇到该种字符图像时，按照该最佳切割范围进行切割和识别即可。
进一步的，为了提高预设识别模型的识别准确率，如图5所示，在步骤S52之前还包括以下步骤：
步骤S60，提取待训练字符图像的深度特征，并对深度特征进行随机遮挡。
具体地，提取待训练字符图像的深度特征，并针对深度特征划分不同的类别，并且建立每个因子特征与相应类别的权重关系，再减弱背景(或噪声)因子的权重，从而达到降低噪声影响，产生更好的学习效果，增强预设识别模型的识别能力。
本实施例通过预设识别模型识别切割后的字符，并计算识别准确率，当识别准确率低于预设阈值时，确认该验证码为难以识别的验证码，通过将该组，当字符因为黏贴在一起而不能很好的进行分割时，确认该验证码为难以识别的验证码，此时，调整切割范围后，重新进行切割，并采用切割后的验证码对预设识别模型进行训练，直至识别准确率超过预设阈值时为止，从而达到增强预设识别模型识别能力的目的，提高了预设识别模型的识别准确率。
图6展示了本发明验证码识别装置的一个实施例。如图6所示，在本实施例中，该验证码识别装置包括获取模块10、切割模块11、识别模块12、计算模块13和训练模块14。
其中，获取模块10，用于获取验证码图像，并对验证码图像进行预处理；切割模块11，用于根据预设切割范围对验证码图像进行切割，得到多个待识别字符图像；识别模块12，用于通过预设识别模型对多个待识别字符图像进行识别；计算模块13，用于计算每个待识别字符图像的识别准确率；训练模块14，用于当识别准确率未超过预设阈值时，调整预设切割范围并重新切割，再根据切割结果训练预设识别模型，直至识别准确率超过预设阈值。
上述实施例的基础上，其他实施例中，如图7所示，获取模块10包括图像获取单元100、灰度化单元101和去噪单元102。
其中，图像获取单元100，用于获取验证码图像；灰度化单元101，用于将验证码图像进行灰度化操作；去噪单元102，用于对灰度化的验证码图像进行二值化处理，再通过空间滤波器确认干扰区域和/或干扰线条，并删除干扰区域和/或干扰线条。
上述实施例的基础上，其他实施例中，如图8所示，计算模块13包括统计单元130和计算单元131。
其中，统计单元130，用于统计每个待识别字符图像的识别正确次数和识别错误次数；计算单元131，用于计算每个待识别字符图像的识别准确率，识别准确率＝识别正确次数/(识别正确次数+识别错误次数)。
上述实施例的基础上，其他实施例中，如图9所示，训练模块14包括调整单元140、重新切割单元141、识别与计算单元142和循环识别单元143。
其中，调整单元140，用于调整预设切割范围；重新切割单元141，用于利用U-Net网络根据调整后的预设切割范围对验证码图像进行切割，得到待训练字符图像；识别与计算单元142，用于通过预设识别模型对待训练字符图像进行识别，并根据识别结果计算待训练字符图像的识别准确率；循环识别单元143，用于当待训练字符图像的识别准确率未超过预设阈值时，重新调整预设切割范围，并再次进行切割和识别，直至待训练字符图像的识别准确率超过预设阈值。
上述实施例的基础上，其他实施例中，如图10所示，训练模块14还包括遮挡单元144，用于提取待训练字符图像的深度特征，并对深度特征进行随机遮挡。
关于上述实施例中验证码识别装置各模块实现技术方案的其他细节，可参见上述实施例中的验证码识别方法中的描述，此处不再赘述。
需要说明的是，本说明书中的各个实施例均采用递进的方式描述，每个实施例重点说明的都是与其他实施例的不同之处，各个实施例之间相同相似的部分互相参见即可。对于装置类实施例而言，由于其与方法实施例基本相似，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
图11展示了本发明又一个实施例提供的终端的示意框图，参见图11，该实施例中的终端包括：一个或至少两个处理器80、存储器81以及存储在该存储器 81中并可在处理器80上运行的计算机程序810。处理器80执行计算机程序810时，实现上述实施例描述的验证码识别方法中的步骤，例如：图1所示的步骤S1-步骤S5。或者，处理器80执行计算机程序810时，实现上述基于多模式集成的降水预报装置实施例中各模块/单元的功能，例如：图6所示模块10-模块14的功能。
计算机程序810可以被分割成一个或多个模块/单元，一个或者多个模块/单元被存储在存储器81中，并由处理器80执行，以完成本申请。一个或多个模块/单元可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述计算机程序810在终端中的执行过程。
终端包括但不仅限于处理器80、存储器81。本领域技术人员可以理解，图8仅仅是终端的一个示例，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，例如终端还可以包括输入设备、输出设备、网络接入设备、总线等。
处理器80可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
存储器81可以是只读存储器、可存储静态信息和指令的静态存储设备、随机存取存储器、或者可存储信息和指令的动态存储设备，也可以是电可擦可编程只读存储器、只读光盘、或其他光盘存储、光碟存储、磁盘存储介质或者其他磁存储设备。存储器81与处理器80可以通过通信总线相连接，也可以和处理器80集成在一起。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述或记载的部分，可以参见其它实施例的相关描述。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例 的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
在本申请所提供的实施例中，应该理解到，所揭露的装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，模块或单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个装置，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通讯连接可以是通过一些接口，装置或单元的间接耦合或通讯连接，可以是电性，机械或其它的形式。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
本申请实施例还提供了一种存储介质，用于存储计算机程序，其包含用于执行本申请上述验证码识别方法实施例所设计的程序数据。通过执行该存储介质中存储的计算机程序，可以实现本申请提供的验证码识别方法。
集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请实现上述实施例方法中的全部或部分流程，也可以通过计算机程序810来指令相关的硬件来完成，计算机程序810可存储于一计算机可读存储介质中，该计算机程序810在被处理器80执行时，可实现上述各个方法实施例的步骤。其中，计算机程序810包括计算机程序代码，计算机程序代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。计算机可读介质可以包括：能够携带计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random  Access Memory)、电载波信号、电信信号以及软件分发介质等。需要说明的是，计算机可读介质包含的内容可以根据司法管辖区内立法和专利实践的要求进行适当的增减，例如在某些司法管辖区，根据立法和专利实践，计算机可读介质不包括是电载波信号和电信信号。
需要说明的是，本说明书中的各个实施例均采用递进的方式描述，每个实施例重点说明的都是与其他实施例的不同之处，各个实施例之间相同相似的部分互相参见即可。对于装置类实施例而言，由于其与方法实施例基本相似，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
以上对发明的具体实施方式进行了详细说明，但其只作为范例，本发明并不限制于以上描述的具体实施方式。对于本领域的技术人员而言，任何对该发明进行的等同修改或替代也都在本发明的范畴之中，因此，在不脱离本发明的精神和原则范围下所作的均等变换和修改、改进等都应涵盖在本发明的范围内。

Claims (12)

  1. 一种验证码识别方法，其特征在于，其包括：
    获取验证码图像，并对所述验证码图像进行预处理；
    根据预设切割范围对所述验证码图像进行切割，得到多个待识别字符图像；
    通过预设识别模型对所述多个待识别字符图像进行识别；
    计算每个待识别字符图像的识别准确率；
    当所述识别准确率未超过预设阈值时，调整所述预设切割范围并重新切割，再根据切割结果训练所述预设识别模型，直至所述识别准确率超过所述预设阈值。
  2. 根据权利要求1所述的验证码识别方法，其特征在于，所述对所述验证码图像进行预处理的步骤包括：
    将所述验证码图像进行灰度化操作；
    对灰度化的验证码图像进行二值化处理，再通过空间滤波器确认干扰区域和/或干扰线条，并删除所述干扰区域和/或干扰线条。
  3. 根据权利要求1所述的验证码识别方法，其特征在于，所述计算每个待识别字符图像的识别准确率的步骤，包括：
    统计每个待识别字符图像的识别正确次数和识别错误次数；
    计算每个待识别字符图像的识别准确率，所述识别准确率＝识别正确次数/(识别正确次数+识别错误次数)。
  4. 根据权利要求1所述的验证码识别方法，其特征在于，所述调整所述预设切割范围并重新切割，再根据切割结果训练所述预设识别模型，直至所述识别准确率超过所述预设阈值的步骤，包括：
    调整所述预设切割范围；
    利用U-Net网络根据所述调整后的预设切割范围对所述验证码图像进行切割，得到所述待训练字符图像；
    通过所述预设识别模型对所述待训练字符图像进行识别，并根据识别结果计算所述待训练字符图像的识别准确率；
    当所述待训练字符图像的识别准确率未超过所述预设阈值时，重新调整所述预设切割范围，并再次进行切割和识别，直至所述待训练字符图像的识别准确率超过所述预设阈值。
  5. 根据权利要求4所述的验证码识别方法，其特征在于，所述通过所述预设识别模型对所述待训练字符图像进行识别的步骤之前，还包括：
    提取所述待训练字符图像的深度特征，并对所述深度特征进行随机遮挡。
  6. 一种验证码识别装置，其特征在于，其包括：
    获取模块，用于获取验证码图像，并对所述验证码图像进行预处理；
    切割模块，用于根据预设切割范围对所述验证码图像进行切割，得到多个待识别字符图像；
    识别模块，用于通过预设识别模型对所述多个待识别字符图像进行识别；
    计算模块，用于计算每个待识别字符图像的识别准确率；
    训练模块，用于当所述识别准确率未超过预设阈值时，调整所述预设切割范围并重新切割，再根据切割结果训练所述预设识别模型，直至所述识别准确率超过所述预设阈值。
  7. 根据权利要求6所述的验证码识别装置，其特征在于，所述获取模块包括：
    图像获取单元，用于获取所述验证码图像；
    灰度化单元，用于将所述验证码图像进行灰度化操作；
    去噪单元，用于对灰度化的验证码图像进行二值化处理，再通过空间滤波器确认干扰区域和/或干扰线条，并删除所述干扰区域和/或干扰线条。
  8. 根据权利要求6所述的验证码识别装置，其特征在于，所述计算模块包括：
    统计单元，用于统计每个待识别字符图像的识别正确次数和识别错误次数；
    计算单元，用于计算每个待识别字符图像的识别准确率，所述识别准确率＝识别正确次数/(识别正确次数+识别错误次数)。
  9. 根据权利要求6所述的验证码识别装置，其特征在于，所述训练模块包括：
    调整单元，用于调整所述预设切割范围；
    重新切割单元，用于利用U-Net网络根据所述调整后的预设切割范围对所述验证码图像进行切割，得到所述待训练字符图像；
    识别与计算单元，用于通过所述预设识别模型对所述待训练字符图像进行识别，并根据识别结果计算所述待训练字符图像的识别准确率；
    循环识别单元，用于当所述待训练字符图像的识别准确率未超过所述预设阈值时，重新调整所述预设切割范围，并再次进行切割和识别，直至所述待训练字符图像的识别准确率超过所述预设阈值。
  10. 根据权利要求9所述的验证码识别装置，其特征在于，所述训练模块还包括：
    遮挡单元，用于提取所述待训练字符图像的深度特征，并对所述深度特征进行随机遮挡。
  11. 一种终端，其特征在于，其包括存储器和处理器，所述处理器耦接所 述存储器，所述存储器上存储有可在所述处理器上运行的计算机程序；
    所述处理器执行所述计算机程序时，实现权利要求1-5任一项所述验证码识别方法中的步骤。
  12. 一种存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时，实现权利要求1-5任一项所述验证码识别方法中的步骤。
PCT/CN2019/130580 2019-03-04 2019-12-31 验证码识别方法、装置、终端及存储介质 Ceased WO2020177470A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910158617.9 2019-03-04
CN201910158617.9A CN109919160B (zh) 2019-03-04 2019-03-04 验证码识别方法、装置、终端及存储介质

Publications (1)

Publication Number Publication Date
WO2020177470A1 true WO2020177470A1 (zh) 2020-09-10

Family

ID=66963048

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/130580 Ceased WO2020177470A1 (zh) 2019-03-04 2019-12-31 验证码识别方法、装置、终端及存储介质

Country Status (2)

Country Link
CN (1) CN109919160B (zh)
WO (1) WO2020177470A1 (zh)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112446377A (zh) * 2020-11-18 2021-03-05 深圳市元征科技股份有限公司 一种数据处理方法、数据处理装置及电子设备
CN113761962A (zh) * 2021-09-13 2021-12-07 武汉先同科技有限公司 一种赋码产品视觉检测方法、系统以及存储介质
CN113868620A (zh) * 2021-10-14 2021-12-31 咪咕文化科技有限公司 一种验证码的验证方法、装置及设备
CN114329414A (zh) * 2021-12-27 2022-04-12 杭州卓健信息科技股份有限公司 一种不定长通用数字字母验证码识别系统及方法
CN115001670A (zh) * 2022-05-25 2022-09-02 武汉极意网络科技有限公司 基于自学习的验证码生成方法、系统、设备及存储介质
CN115546226A (zh) * 2022-10-08 2022-12-30 智慧芽信息科技(苏州)有限公司 验证码图像识别方法、模板生成方法、装置和存储介质
CN115712887A (zh) * 2023-01-09 2023-02-24 成方金融科技有限公司 图片验证码识别方法、装置、电子设备及存储介质

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109919160B (zh) * 2019-03-04 2021-03-23 深圳先进技术研究院 验证码识别方法、装置、终端及存储介质
CN110348450A (zh) * 2019-07-15 2019-10-18 中国工商银行股份有限公司 用于图像验证码的安全评估方法、装置和计算机系统
US11204987B2 (en) * 2019-11-07 2021-12-21 Nxp B.V. Method for generating a test for distinguishing humans from computers
CN110909807A (zh) * 2019-11-26 2020-03-24 深圳市信联征信有限公司 基于深度学习的网络验证码识别方法、装置及计算机设备
CN111274957A (zh) * 2020-01-20 2020-06-12 阳光人寿保险股份有限公司 网页页面验证码识别方法、装置、终端和计算机存储介质
CN111340137A (zh) * 2020-03-26 2020-06-26 上海眼控科技股份有限公司 图像识别方法、装置及存储介质
CN112989312B (zh) * 2020-11-30 2024-04-30 北京金堤科技有限公司 验证码的识别方法、装置、电子设备和存储介质
CN113627395B (zh) * 2021-09-17 2023-11-17 平安银行股份有限公司 文本识别方法、装置、介质及电子设备
CN114067324B (zh) * 2021-10-27 2025-10-17 厦门友微科技有限公司 一种验证码图片识别方法、装置、设备及存储介质
CN115174122B (zh) * 2022-07-18 2024-12-03 湖北天融信网络安全技术有限公司 验证码生成方法、验证码验证方法、装置、设备及介质
CN115482539A (zh) * 2022-08-30 2022-12-16 中电信数智科技有限公司 一种基于图像识别的收货单生成方法和装置
CN116347419A (zh) * 2023-03-23 2023-06-27 国网湖南省电力有限公司 移动终端识别方法及系统
CN116311297A (zh) * 2023-04-12 2023-06-23 国网河北省电力有限公司 基于计算机视觉的电子凭据图像识别解析方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130298195A1 (en) * 2010-02-19 2013-11-07 Microsoft Corporation Image-Based CAPTCHA Exploiting Context in Object Recognition
CN108596168A (zh) * 2018-04-20 2018-09-28 北京京东金融科技控股有限公司 用于识别图像中字符的方法、装置及介质
CN109919160A (zh) * 2019-03-04 2019-06-21 深圳先进技术研究院 验证码识别方法、装置、终端及存储介质

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1999064982A1 (en) * 1998-06-09 1999-12-16 Kent Ridge Digital Labs A recognition process
CN101408933A (zh) * 2008-05-21 2009-04-15 浙江师范大学 基于粗网格特征提取和bp神经网络的车牌字符识别方法
CN101630360B (zh) * 2008-07-14 2012-12-19 上海分维智能科技有限公司 一种在高清晰图像中识别车牌的方法
CN101859382B (zh) * 2010-06-03 2013-07-31 复旦大学 一种基于最大稳定极值区域的车牌检测与识别的方法
CN102930277B (zh) * 2012-09-19 2016-04-27 上海珍岛信息技术有限公司 一种基于识别反馈的字符图像验证码识别方法
CN104408452B (zh) * 2014-11-17 2019-01-15 深圳市比一比网络科技有限公司 一种基于旋转投影宽度的拉丁字符倾斜纠正方法及系统
CN104408455B (zh) * 2014-11-27 2017-09-12 上海理工大学 粘连字符分割方法
CN105069456B (zh) * 2015-07-30 2018-10-09 北京邮电大学 一种车牌字符分割方法及装置
CN105654042B (zh) * 2015-12-23 2019-04-09 华南理工大学 玻璃温度计的检定温度字符识别方法
CN106874901B (zh) * 2017-01-17 2020-07-03 北京智元未来科技有限公司 一种行驶证识别方法及装置
CN107688809B (zh) * 2017-08-28 2019-08-27 深圳壹账通智能科技有限公司 验证码识别方法、装置、计算机设备及计算机存储介质
CN108875745B (zh) * 2018-08-09 2021-11-19 同程网络科技股份有限公司 组合验证码的分割识别方法及其系统
CN109086772A (zh) * 2018-08-16 2018-12-25 成都市映潮科技股份有限公司 一种扭曲粘连字符图片验证码的识别方法及系统

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130298195A1 (en) * 2010-02-19 2013-11-07 Microsoft Corporation Image-Based CAPTCHA Exploiting Context in Object Recognition
CN108596168A (zh) * 2018-04-20 2018-09-28 北京京东金融科技控股有限公司 用于识别图像中字符的方法、装置及介质
CN109919160A (zh) * 2019-03-04 2019-06-21 深圳先进技术研究院 验证码识别方法、装置、终端及存储介质

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
BAI PEIRUI ,WANG JINBO , DING GUOMEI: "A General CAPTCHA Recognition Algorithm Based on Image Segmentation", JOURNAL OF SHANDONG UNIVERSITY OF SCIENCE AND TECHNOLOGY (NATURAL SCIENCE), vol. 37, no. 3, 30 June 2018 (2018-06-30), pages 106 - 113, XP055731837, ISSN: 1672-3767, DOI: 10.16452/j.cnki.sdkjzk.2018.03.014 *
LIU HUAN , SHAO WEIYUAN ,GUO YUEFEI: "Research on CAPTCHA Recognition with Convolutional Neural Networks", COMPUTER ENGINEERING AND APPLICATIONS, vol. 52, no. 18, 15 September 2016 (2016-09-15), pages 1 - 7, XP055731840, ISSN: 1002-8331, DOI: 10.3778/j.issn.1002-8331.1603-0181 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112446377A (zh) * 2020-11-18 2021-03-05 深圳市元征科技股份有限公司 一种数据处理方法、数据处理装置及电子设备
CN113761962A (zh) * 2021-09-13 2021-12-07 武汉先同科技有限公司 一种赋码产品视觉检测方法、系统以及存储介质
CN113761962B (zh) * 2021-09-13 2022-09-16 武汉先同科技有限公司 一种赋码产品视觉检测方法、系统以及存储介质
CN113868620A (zh) * 2021-10-14 2021-12-31 咪咕文化科技有限公司 一种验证码的验证方法、装置及设备
CN114329414A (zh) * 2021-12-27 2022-04-12 杭州卓健信息科技股份有限公司 一种不定长通用数字字母验证码识别系统及方法
CN115001670A (zh) * 2022-05-25 2022-09-02 武汉极意网络科技有限公司 基于自学习的验证码生成方法、系统、设备及存储介质
CN115546226A (zh) * 2022-10-08 2022-12-30 智慧芽信息科技(苏州)有限公司 验证码图像识别方法、模板生成方法、装置和存储介质
CN115712887A (zh) * 2023-01-09 2023-02-24 成方金融科技有限公司 图片验证码识别方法、装置、电子设备及存储介质

Also Published As

Publication number Publication date
CN109919160B (zh) 2021-03-23
CN109919160A (zh) 2019-06-21

Similar Documents

Publication Publication Date Title
WO2020177470A1 (zh) 验证码识别方法、装置、终端及存储介质
US11423701B2 (en) Gesture recognition method and terminal device and computer readable storage medium using the same
CN102496023B (zh) 像素层面的感兴趣区域提取方法
CN110705583A (zh) 细胞检测模型训练方法、装置、计算机设备及存储介质
CN103824091B (zh) 一种用于智能交通系统的车牌识别方法
CN114581709B (zh) 识别医学图像中的目标的模型训练、方法、设备及介质
CN105844277B (zh) 标签识别方法和装置
WO2017121018A1 (zh) 二维码图像处理的方法和装置、终端、存储介质
CN110956081B (zh) 车辆与交通标线位置关系的识别方法、装置及存储介质
CN110969046B (zh) 人脸识别方法、设备及计算机可读存储介质
CN112801232A (zh) 一种应用于处方录入的扫描识别方法及系统
CN110570442A (zh) 一种复杂背景下轮廓检测方法、终端设备及存储介质
CN108764328A (zh) 太赫兹图像危险品识别方法、装置、设备及可读存储介质
CN107563386A (zh) 元器件验证方法、装置、设备与计算机可读存储介质
CN115082776A (zh) 一种基于图像识别的电能表自动检测系统及方法
CN113537211A (zh) 一种基于非对称iou的深度学习车牌框定位方法
CN105809089A (zh) 复杂背景下的多人脸检测方法及装置
CN112084874B (zh) 一种物体检测方法、装置及终端设备
CN111986176A (zh) 一种裂纹图像识别方法、系统、终端及可读存储介质
CN111488811A (zh) 人脸识别方法、装置、终端设备及计算机可读介质
CN112561893B (zh) 图片匹配方法、装置、电子设备及存储介质
Chen et al. Image segmentation based on mathematical morphological operator
CN111199240A (zh) 银行卡识别模型的训练方法、银行卡识别方法以及装置
CN111062309A (zh) 一种雨天交通标志检测方法、存储介质及系统
CN112183540B (zh) 标签参数的提取方法和装置、存储介质、电子装置

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: 19917732

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: 19917732

Country of ref document: EP

Kind code of ref document: A1

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

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 02.02.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 19917732

Country of ref document: EP

Kind code of ref document: A1