WO2022105179A1 - 生物特征图像识别方法、装置、电子设备及可读存储介质 - Google Patents

生物特征图像识别方法、装置、电子设备及可读存储介质 Download PDF

Info

Publication number
WO2022105179A1
WO2022105179A1 PCT/CN2021/097072 CN2021097072W WO2022105179A1 WO 2022105179 A1 WO2022105179 A1 WO 2022105179A1 CN 2021097072 W CN2021097072 W CN 2021097072W WO 2022105179 A1 WO2022105179 A1 WO 2022105179A1
Authority
WO
WIPO (PCT)
Prior art keywords
image set
sample image
image
target sample
training
Prior art date
Application number
PCT/CN2021/097072
Other languages
English (en)
French (fr)
Inventor
李佳琳
李昌昊
王健宗
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2022105179A1 publication Critical patent/WO2022105179A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/18Eye characteristics, e.g. of the iris
    • G06V40/197Matching; Classification
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/18Eye characteristics, e.g. of the iris
    • G06V40/193Preprocessing; Feature extraction

Definitions

  • the present application relates to the technical field of artificial intelligence, and in particular, to a biometric image recognition method, apparatus, electronic device, and readable storage medium.
  • biometric identification as an effective way to identify passengers, is widely used in various fields such as stations, airports, finance, and network security.
  • the inventor realized that the current biometric identification is to collect the to-be-recognized image of the identified person, and to extract the image features of the to-be-recognized image through a trained deep learning model and match the image features in the database to achieve the effect of recognition.
  • a biometric image recognition method comprising:
  • the feature vector is used for comparison and identification in a preset image feature vector library to obtain the identification result.
  • a biometric image recognition device includes:
  • a training sample generation module is used to obtain a first training image set, and use the first training image set to train a pre-built generative adversarial network model to obtain a data enhancement model; obtain a first sample image set, use the data
  • the enhancement model performs data enhancement on the first sample image set to obtain a second sample image set; summarizes the first sample image set and the second sample image set to obtain a first target sample image set; performing data amplification on the first target sample image set to obtain a second target sample image set;
  • a model training module used to train a pre-built deep learning network model by using the second target sample image set to obtain an image recognition model
  • the image recognition module is used for, when receiving the biometric image to be recognized, use the image recognition model to perform feature extraction on the biometric image to be recognized to obtain a feature vector; use the feature vector in a preset image feature vector The comparison and identification are performed in the library to obtain the identification result.
  • An electronic device comprising:
  • the processor executes the computer program stored in the memory to realize the following steps:
  • the feature vector is used for comparison and identification in a preset image feature vector library to obtain the identification result.
  • a computer-readable storage medium having at least one computer program stored in the computer-readable storage medium, the at least one computer program being executed by a processor to implement the following steps:
  • the feature vector is used for comparison and identification in a preset image feature vector library to obtain the identification result.
  • the present application can improve the accuracy of biometric image recognition.
  • FIG. 1 is a schematic flowchart of a biometric image recognition method provided by an embodiment of the present application.
  • FIG. 2 is a schematic flowchart of obtaining a recognition result in a biometric image recognition method provided by an embodiment of the present application
  • FIG. 3 is a schematic block diagram of a biometric image recognition device provided by an embodiment of the present application.
  • FIG. 4 is a schematic diagram of the internal structure of an electronic device for implementing a method for recognizing a biometric image provided by an embodiment of the present application;
  • An embodiment of the present application provides a method for identifying a biometric image.
  • the execution body of the biometric image recognition method includes, but is not limited to, at least one of electronic devices that can be configured to execute the method provided by the embodiments of the present application, such as a server and a terminal.
  • the biometric image recognition method can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform.
  • the server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like.
  • the biometric image recognition method includes:
  • the first training image set is an iris image set
  • the iris image is an eye image including an iris region.
  • the generative adversarial network model is divided into two parts: a generator and a discriminator.
  • the generator is used to generate an image according to the first training image set
  • the discriminator is used to determine whether the generated image is a fake image generated by the generator or a corresponding real image in the first training image set, so that the generated image is a fake image generated by the generator or a corresponding real image in the first training image set.
  • the generated image generated by the generator can approximate the real image in the first training image set, and a generative model is obtained according to the generation network at this time, which is used to generate a new image and solve the problem that the image sample is too small.
  • using the first training image set to train the pre-built generative adversarial network model includes: constructing a first loss function; based on the first loss function, using the first loss function
  • the training image set performs alternate iterative training of the generator and the discriminator on the first generative adversarial network model; when the value of the first loss function reaches a first preset threshold, the training is stopped to obtain the data enhancement model;
  • the first loss function is:
  • the L GAN is a pre-built adversarial loss function
  • z represents a preset random parameter variable
  • D is the generator in the first generative adversarial network model
  • G is the first generative adversarial network model.
  • Discriminator x is the real image in the image generated by the generator
  • P (x) is the probability distribution of the real image in the generated image
  • P (z) is the probability distribution of the false image in the generated image
  • E is Expected value calculation function.
  • the first sample image set is a different iris image set from the first training image set.
  • the data enhancement model is used to process all the images in the first sample image set to obtain the second sample image set.
  • the first sample image set includes images a and b, input image a into the data enhancement model to obtain image A, input image b into the data enhancement model to obtain image B, and summarize image A and image B to obtain the first image Two-sample image set.
  • the performing data augmentation on the first target sample image set includes: performing translation, flipping and color adjustment operations on all images in the first target sample image set to obtain augmentation an image set; summarizing the augmented image set and the first target sample image set; labeling the images in the summarized image set to obtain the second target sample image set.
  • the label area is the iris area.
  • the Label Me image labeling tool can be used to manually mark the label area in the embodiment of the present application.
  • the second target sample image set may be stored in a blockchain node.
  • the deep learning network model may include a convolutional neural network model or the like.
  • the embodiment of the present application uses the second target sample image set to train the pre-built deep learning network model, including:
  • Step A according to the preset convolution pooling times, perform a convolution pooling operation on the second target sample image set to obtain a feature set;
  • Step B use a preset activation function to calculate the feature set to obtain a predicted value, obtain the label value of the label area corresponding to each image in the second target sample image set, and obtain the predicted value according to the predicted value and the The label value is calculated using the pre-built second loss function to obtain the loss value;
  • the label value and the label area are in one-to-one correspondence, for example, the label value corresponding to the label area is 1, and the label value corresponding to the non-label area is 0.
  • Step C Compare the size of the loss value with the second preset threshold value, when the loss value is greater than or equal to the second preset threshold value, return to the step A; or when the loss value is less than the first When the preset threshold is reached, the training is stopped to obtain the image recognition model.
  • performing a convolution pooling operation on the second target sample image set to obtain a first feature set includes: performing a convolution operation on the second target sample image set to obtain a first volume product data set; perform a maximum pooling operation on the first convolution data set to obtain the feature set.
  • ⁇ ' represents the channel number of the first convolution data set
  • represents the channel number of the second target sample image set
  • k is the size of the preset convolution kernel
  • f is the step of the preset convolution operation Amplitude
  • p is the preset data zero padding matrix.
  • the first activation function described in the preferred embodiment of the present application includes:
  • ⁇ t represents the predicted value
  • s represents the data in the feature set
  • the first loss function described in the preferred embodiment of the present application includes:
  • L ce represents the loss value
  • N is the data number of the second target sample image set
  • i is a positive integer
  • y i is the label value
  • p i is the predicted value
  • the biometric image to be recognized is an iris image that needs to be recognized.
  • performing feature extraction on the to-be-recognized biometric image by using the image recognition model to obtain a feature vector includes: using the image recognition model to perform image recognition on the to-be-recognized image , extract the output value of the fully connected layer in the image recognition model after the image recognition is completed, and obtain the feature vector.
  • the extracting the output value of the fully connected layer in the image recognition model after the image recognition is completed, to obtain the feature vector includes: according to the image recognition model in the fully connected layer of the node in the fully connected layer. sequence, extract the output values of all nodes in the fully connected layer in the image recognition model after the image recognition is completed and combine them vertically to obtain the feature vector, for example: the fully connected layer has 3 nodes, which are the first node, The second node and the third node, after completing the image recognition, the output value of the first node of the fully connected layer is 1, the output value of the second node is 3, and the output value of the third node is 5. 5 The three eigenvalues are combined vertically in the node order to obtain the eigenvector
  • the use of the feature vector to perform comparison and identification in a preset image feature vector library to obtain a recognition result includes:
  • the embodiment of the present application uses a cosine similarity algorithm to calculate the similarity between the feature vector and each image feature vector in the image feature vector library.
  • the identification result is successful; if there is no similarity value greater than or equal to the similarity value set in the similarity value set the third preset threshold, the identification result is identification failure.
  • the third preset threshold is 0.85, there is a similarity value of 0.9 that is greater than the third preset threshold of 0.85, and the recognition result is that the recognition is successful;
  • the third preset threshold is 0.95, there is no similarity value greater than or equal to 0.95, and the recognition result is recognition failure.
  • FIG. 3 it is a functional block diagram of the biometric image recognition device of the present application.
  • the biometric image recognition apparatus 100 described in this application may be installed in an electronic device.
  • the biometric image recognition device may include a training sample generation module 101, a model training module 102, and an image recognition module 103.
  • the modules described in the present invention may also be referred to as units, which refer to a device that can be used by an electronic device. A series of computer program segments executed by a processor and capable of performing fixed functions and stored in the memory of an electronic device.
  • each module/unit is as follows:
  • the training sample generation module 101 is used to obtain a first training image set, and use the first training image set to train a pre-built generative adversarial network model to obtain a data enhancement model; obtain the first sample image set, use the The data enhancement model performs data enhancement on the first sample image set to obtain a second sample image set; summarizes the first sample image set and the second sample image set to obtain a first target sample image set; Data augmentation is performed on the first target sample image set to obtain a second target sample image set.
  • the first training image set is an iris image set
  • the iris image is an eye image including an iris region.
  • the generative adversarial network model is divided into two parts: a generator and a discriminator.
  • the generator is used to generate an image according to the first training image set
  • the discriminator is used to determine whether the generated image is a fake image generated by the generator or a corresponding real image in the first training image set, so that the generated image is a fake image generated by the generator or a corresponding real image in the first training image set.
  • the generated image generated by the generator can approximate the real image in the first training image set, and a generation model is obtained according to the generation network at this time, which is used to generate a new image and solve the problem that the image sample is too small.
  • the training sample generation module 101 uses the first training image set to train the pre-built generative adversarial network model, including: constructing a first loss function; based on the first loss function, Use the first training image set to perform alternate iterative training of generator and discriminator on the first generative adversarial network model; when the value of the first loss function reaches a first preset threshold, stop training, and obtain the Describe the data augmentation model;
  • the first loss function is:
  • the L GAN is a pre-built adversarial loss function
  • z represents a preset random parameter variable
  • D is the generator in the first generative adversarial network model
  • G is the first generative adversarial network model.
  • Discriminator x is the real image in the image generated by the generator
  • P (x) is the probability distribution of the real image in the generated image
  • P (z) is the probability distribution of the false image in the generated image
  • E is Expected value calculation function.
  • the first sample image set is a different iris image set from the first training image set.
  • the training sample generation module 101 in this embodiment of the present application processes all the images in the first sample image set by using the data enhancement model,
  • the second set of sample images is obtained.
  • the first sample image set includes images a and b, input image a into the data enhancement model to obtain image A, input image b into the data enhancement model to obtain image B, and summarize image A and image B to obtain the first image Two-sample image set.
  • the training sample generation module 101 summarizes the first sample image set and the second sample image set to obtain a first target sample image set
  • the training sample generation module 101 performs data augmentation on the first target sample image set.
  • the training sample generation module 101 performs data augmentation on the first target sample image set, including: performing translation flipping and color adjustment on all images in the first target sample image set operations to obtain an augmented image set; summarizing the augmented image set and the first target sample image set; and labeling the images in the summarized image set to obtain the second target sample image set.
  • the label area is the iris area.
  • the Label Me image labeling tool can be used to manually mark the label area in the embodiment of the present application.
  • the second target sample image set may be stored in a blockchain node.
  • the model training module 102 is configured to use the second target sample image set to train a pre-built deep learning network model to obtain an image recognition model.
  • the deep learning network model may include a convolutional neural network model or the like.
  • model training module 102 uses the following means to train the pre-built deep learning network model, including:
  • Step A according to the preset convolution pooling times, perform a convolution pooling operation on the second target sample image set to obtain a feature set;
  • Step B use a preset activation function to calculate the feature set to obtain a predicted value, obtain the label value of the label area corresponding to each image in the second target sample image set, and obtain the predicted value according to the predicted value and the The label value is calculated using the pre-built second loss function to obtain the loss value;
  • the label value and the label area are in one-to-one correspondence, for example, the label value corresponding to the label area is 1, and the label value corresponding to the non-label area is 0.
  • Step C Compare the size of the loss value with the second preset threshold value, when the loss value is greater than or equal to the second preset threshold value, return to the step A; or when the loss value is less than the first When the preset threshold is reached, the training is stopped to obtain the image recognition model.
  • the model training module 102 performs a convolution pooling operation on the second target sample image set to obtain a first feature set, including: performing a convolution operation on the second target sample image set Obtain a first convolution data set; perform a maximum pooling operation on the first convolution data set to obtain the feature set.
  • ⁇ ' represents the number of channels of the first convolution data set
  • represents the number of channels of the second target sample image set
  • k is the size of the preset convolution kernel
  • f is the step of the preset convolution operation Amplitude
  • p is the preset data zero padding matrix.
  • the first activation function described in the preferred embodiment of the present application includes:
  • ⁇ t represents the predicted value
  • s represents the data in the feature set
  • the first loss function described in the preferred embodiment of the present application includes:
  • L ce represents the loss value
  • N is the data number of the second target sample image set
  • i is a positive integer
  • y i is the label value
  • p i is the predicted value
  • the image recognition module 103 is configured to perform feature extraction on the biometric image to be recognized by using the image recognition model when receiving the biometric image to be recognized to obtain a feature vector; use the feature vector in the preset image. Compare and identify in the feature vector library to get the identification result.
  • the biometric image to be recognized is an iris image that needs to be recognized.
  • the image recognition module 103 uses the image recognition model to perform feature extraction on the biometric image to be recognized to obtain a feature vector, including: using the image recognition model to perform a feature extraction on the to-be-recognized biometric image Perform image recognition on the image, extract the output value of the fully connected layer in the image recognition model after the image recognition is completed, and obtain the feature vector.
  • the image recognition module 103 uses the following means to obtain the feature vector, including: extracting the image after image recognition is completed according to the order of nodes in the fully connected layer in the image recognition model Identify the output values of all nodes in the fully connected layer in the model and combine them vertically to obtain the feature vector.
  • the fully connected layer has 3 nodes, which are the first node, the second node, and the third node in order to complete image recognition.
  • the output value of the first node of the post-full connection layer is 1, the output value of the second node is 3, and the output value of the third node is 5.
  • the three eigenvalues 1, 3, and 5 are vertically combined in the order of the nodes.
  • the image recognition module 103 in the embodiment of the present application obtains the recognition result by using the following means, including:
  • the embodiment of the present application uses a cosine similarity algorithm to calculate the similarity between the feature vector and each image feature vector in the image feature vector library.
  • Screening and comparing is performed according to the similarity value set to obtain the identification result.
  • the identification result is successful; if there is no similarity value greater than or equal to the similarity value set in the similarity value set the third preset threshold, the identification result is identification failure.
  • the third preset threshold is 0.85, there is a similarity value of 0.9 that is greater than the third preset threshold of 0.85, and the recognition result is that the recognition is successful;
  • the third preset threshold is 0.95, there is no similarity value greater than or equal to 0.95, and the recognition result is recognition failure.
  • FIG. 4 it is a schematic structural diagram of an electronic device implementing the biometric image recognition method of the present application.
  • the electronic device 1 may include a processor 10, a memory 11 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a biometric image recognition program.
  • the memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, magnetic disk, CD etc.
  • the memory 11 may be an internal storage unit of the electronic device 1 , such as a mobile hard disk of the electronic device 1 .
  • the memory 11 may also be an external storage device of the electronic device 1, such as a pluggable mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital) equipped on the electronic device 1. , SD) card, flash memory card (Flash Card), etc.
  • the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device.
  • the memory 11 can not only be used to store application software installed in the electronic device 1 and various types of data, such as codes of biometric image recognition programs, etc., but also can be used to temporarily store data that has been output or will be output.
  • the processor 10 may be composed of integrated circuits, for example, may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits packaged with the same function or different functions, including one or more integrated circuits.
  • Central Processing Unit CPU
  • microprocessor digital processing chip
  • graphics processor and combination of various control chips, etc.
  • the processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect the various components of the entire electronic device, by running or executing programs or modules (such as biological components) stored in the memory 11. feature image recognition program, etc.), and call data stored in the memory 11 to execute various functions of the electronic device 1 and process data.
  • the bus may be a peripheral component interconnect (PCI for short) bus or an extended industry standard architecture (Extended industry standard architecture, EISA for short) bus or the like.
  • PCI peripheral component interconnect
  • EISA Extended industry standard architecture
  • the bus can be divided into address bus, data bus, control bus and so on.
  • the bus is configured to implement connection communication between the memory 11 and at least one processor 10 and the like.
  • FIG. 4 only shows an electronic device with components. Those skilled in the art can understand that the structure shown in FIG. 4 does not constitute a limitation on the electronic device 1, and may include fewer or more components than those shown in the drawings. components, or a combination of certain components, or a different arrangement of components.
  • the electronic device 1 may also include a power supply (such as a battery) for powering the various components, preferably, the power supply may be logically connected to the at least one processor 10 through a power management device, so that the power management
  • the device implements functions such as charge management, discharge management, and power consumption management.
  • the power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and any other components.
  • the electronic device 1 may further include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.
  • the electronic device 1 may also include a network interface, optionally, the network interface may include a wired interface and/or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is usually used in the electronic device 1 Establish a communication connection with other electronic devices.
  • a network interface optionally, the network interface may include a wired interface and/or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is usually used in the electronic device 1 Establish a communication connection with other electronic devices.
  • the electronic device 1 may further include a user interface, and the user interface may be a display (Display), an input unit (eg, a keyboard (Keyboard)), optionally, the user interface may also be a standard wired interface or a wireless interface.
  • the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, and the like.
  • the display may also be appropriately called a display screen or a display unit, which is used for displaying information processed in the electronic device 1 and for displaying a visualized user interface.
  • the biometric image recognition program 12 stored in the memory 11 in the electronic device 1 is a combination of multiple computer programs, and when running in the processor 10, can realize:
  • the feature vector is used for comparison and identification in a preset image feature vector library to obtain the identification result.
  • modules/units integrated in the electronic device 1 may be stored in a computer-readable storage medium.
  • the computer-readable medium may be non-volatile or volatile.
  • the computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, removable hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) .
  • the computer usable storage medium may mainly include a stored program area and a stored data area, wherein the stored program area may store an operating system, an application program required for at least one function, and the like; Using the created data, etc., the application program executed by the processor can implement the following steps:
  • the feature vector is used for comparison and identification in a preset image feature vector library to obtain the identification result.
  • modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.
  • each functional module in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units can be implemented in the form of hardware, or can be implemented in the form of hardware plus software function modules.
  • the blockchain referred to in this application is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

Abstract

一种生物特征图像识别方法、装置、电子设备及可读存储介质,其中方法包括:利用训练完成的数据增强模型对第一样本图像集进行数据增强,得到第二样本图像集(S2);汇总第一样本图像集及第二样本图像集,得到第一目标样本图像集(S3);对第一目标样本图像集进行数据扩增,得到第二目标样本图像集(S4);利用第二目标样本图像集训练完成的图像识别模型对待识别生物特征图像进行特征提取,得到特征向量(S6);利用特征向量在预设的图像特征向量库中进行比对识别,得到识别结果(S7)。该方法可以生物特征图像识别的准确率。

Description

生物特征图像识别方法、装置、电子设备及可读存储介质
本申请要求于2020年11月23日提交中国专利局、申请号为CN202011322604.X,发明名称为“生物特征图像识别方法、装置、电子设备及可读存储介质”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
本申请涉及人工智能技术领域,尤其涉及一种生物特征图像识别方法、装置、电子设备及可读存储介质。
背景技术
随着通信和交通技术的迅速发展,生物特征识别作为辨别乘客身份的有效方式,被广泛应用于车站、机场、金融、网络安全等各种领域。
发明人意识到目前生物特征识别是采集到被识别人的待识别图像,通过训练的深度学习模型提取待识别图像的图像特征与数据库中的图像特征进行匹配达到识别的效果。
但是深度学习模型的训练需要大量的生物特征图像,由于生物特征图像具有隐私性且不易获取,所以深度学习模型往往因为缺少足够的训练样本导致特征提取能力差,从而影响生物特征图像识别的准确率。
发明内容
一种生物特征图像识别方法,包括:
获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;
获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;
汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;
利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
一种生物特征图像识别装置,所述装置包括:
训练样本生成模块,用于获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
模型训练模块,用于利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
图像识别模块,用于当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
一种电子设备,所述电子设备包括:
存储器,存储至少一个计算机程序;及
处理器,执行所述存储器中存储的计算机程序以实现如下步骤:
获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;
获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;
汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;
利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
一种计算机可读存储介质,所述计算机可读存储介质中存储有至少一个计算机程序,所述至少一个计算机程序被处理器执行以实现如下步骤:
获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;
获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;
汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;
利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
本申请可以提高生物特征图像识别的准确率。
附图说明
图1为本申请一实施例提供的生物特征图像识别方法的流程示意图;
图2为本申请一实施例提供的生物特征图像识别方法中得到识别结果的流程示意图;
图3为本申请一实施例提供的生物特征图像识别装置的模块示意图;
图4为本申请一实施例提供的实现生物特征图像识别方法的电子设备的内部结构示意图;
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。
本申请实施例提供一种生物特征图像识别方法。所述生物特征图像识别方法的执行主体包括但不限于服务端、终端等能够被配置为执行本申请实施例提供的该方法的电子设备中的至少一种。换言之,所述生物特征图像识别方法可以由安装在终端设备或服务端设备的软件或硬件来执行,所述软件可以是区块链平台。所述服务端包括但不限于:单台服务器、服务器集群、云端服务器或云端服务器集群等。
参照图1所示的本申请一实施例提供的生物特征图像识别方法的流程示意图,在本申请实施例中,所述生物特征图像识别方法包括:
S1、获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;
本申请实施例中,所述第一训练图像集为虹膜图像集合,所述虹膜图像为包含虹膜区域的眼睛图像。
本领域技术人员应该了解,所述生成对抗网络模型分为生成器和鉴别器两个部分。在训练过程中,生成器用来根据所述第一训练图像集生成图像,所述鉴别器用来判断生成图像是生成器生成的虚假图像还是所述第一训练图像集中的对应的真实图像,使生成器生成的生成图像,能逼近所述第一训练图像集中真实图像,根据此时的生成网络得到一个生成模型,用来生成新的图像,解决图像样本过小的问题。
详细地,本申请实施例中所述利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,包括:构建第一损失函数;基于所述第一损失函数,利用所述第一训练图像集对所述第一生成对抗网络模型进行生成器与鉴别器的交替迭代训练;当所述第一损失函数的值达到第一预设阈值时,停止训练,得到所述数据增强模型;
其中,所述第一损失函数为:
Figure PCTCN2021097072-appb-000001
其中,所述L GAN为预构建的对抗损失函数,z表示预设的随机参数变量,D为所述第一生成对抗网络模型中的生成器,G为所述第一生成对抗网络模型中的鉴别器,x为所述生成器生成图像中的真实图像,P (x)为所述生成图像中真实图像的概率分布,P (z)为所述生成图像中虚假图像的概率分布,E为期望值计算函数。
S2、获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;
本申请实施例中,所述第一样本图像集是与所述第一训练图像集不同的虹膜图像集合。
由于所述第一样本图像集样本较小,且不易获取,所以本申请实施例利用所述数据增强模型对所述第一样本图像集中的所有图像进行处理,得到所述第二样本图像集。例如:所述第一样本图像集中包含图像a、b,将图像a输入所述数据增强模型得到图像A,将图像b输入所述数据增强模型得到图像B,汇总图像A及图像B得到第二样本图像集。
S3、汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
S4、对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
本申请实施例中,为了提高后续模型的鲁棒性,对所述第一目标样本图像集进行数据扩增。
详细地,本申请实施例中,所述对所述第一目标样本图像集进行数据扩增,包括:对所述第一目标样本图像集中的所有图像进行平移翻转及颜色调整操作,得到扩增图像集;汇总所述扩增图像集及所述第一目标样本图像集;对汇总后的图像集中的图像进行标签区域标记,得到所述第二目标样本图像集。本申请实施例中,所述标签区域为虹膜区域,较佳地,本申请实施例可使用Label Me图像标注工具人工进行标签区域标记。
本申请的另一实施例中,为了保证数据的隐私性,所述第二目标样本图像集可以存储在区块链节点中。
S5、利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像特征提取模型;
较佳地,本申请实施例中,所述深度学习网络模型可以包括卷积神经网络模型等。
进一步地,本申请实施例利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,包括:
步骤A:根据预设的卷积池化次数,对所述第二目标样本图像集进行卷积池化操作,得到特征集;
步骤B:利用预设的激活函数对所述特征集进行计算得到预测值,获取所述第二目标样本图像集中每张图像对应的所述标签区域的标签值,根据所述预测值及所述标签值,利用预构建的第二损失函数进行计算,得到损失值;
本申请实施例中所述标签值与所述标签区域是一一对应的,例如:所述标签区域对应的标签值为1,非标签区域对应的标签值为0。
步骤C:对比所述损失值与第二预设阈值的大小,当所述损失值大于或等于所述第二预设阈值时,返回所述步骤A;或者当所述损失值小于所述第二预设阈值时,停止训练,得到所述图像识别模型。
详细地,本申请实施例中所述对所述第二目标样本图像集进行卷积池化操作得到第一特征集,包括:对所述第二目标样本图像集进行卷积操作得到第一卷积数据集;对所述第一卷积数据集进行最大池化操作得到所述特征集。
进一步地,所述卷积操作为:
Figure PCTCN2021097072-appb-000002
其中,ω’表示所述第一卷积数据集的通道数,ω表示所述第二目标样本图像集的通道数,k为预设卷积核的大小,f为预设卷积操作的步幅,p为预设数据补零矩阵。
进一步地,本申请较佳实施例所述第一激活函数包括:
Figure PCTCN2021097072-appb-000003
其中,μ t表示所述预测值,s表示所述特征集中的数据。
详细地,本申请较佳实施例所述第一损失函数包括:
Figure PCTCN2021097072-appb-000004
其中,L ce表示所述损失值,N为所述第二目标样本图像集的数据数目,i为正整数,y i为所述标签值,p i为所述预测值。
S6、当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;
本申请实施例中,所述待识别生物特征图像为需要识别的虹膜图像。
详细地,本申请实施例中,所述利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量,包括:利用所述图像识别模型对所述待识别图像进行图像识别,提取完成图像识别后所述图像识别模型中全连接层的输出值,得到所述特征向量。
进一步地,本申请实施例中,所述提取完成图像识别后所述图像识别模型中全连接层的输出值,得到所述特征向量,包括:根据所述图像识别模型中全连接层中节点的顺序,提取完成图像识别后的所述图像识别模型中全连接层所有节点的输出值并纵向组合,得到所述特征向量,例如:全连接层共有3个节点,按顺序分别为第一节点、第二节点、第三节点,完成图像识别后全连接层第一个节点的输出值为1、第二个节点的输出值为3、第三个节点的输出值为5,将1、3、5三个特征值按节点顺序纵向组合得到特征向量
Figure PCTCN2021097072-appb-000005
S7、利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
详细地,本申请实施例中,参阅图2所示,所述利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果,包括:
S71、计算所述特征向量与所述图像特征向量库中每个图像特征向量的相似度,得到对应的相似度值;
较佳地,本申请实施例利用余弦相似度算法计算所述特征向量与所述图像特征向量库中每个图像特征向量的相似度。
S72、汇总所有的所述相似度值,得到相似度值集;
S73、根据所述相似度值集进行筛选比对,得到所述识别结果。
本申请实施例中,若所述相似度值集中存在相似度值大于或等于第三预设阈值,则所述识别结果为识别成功;若所述相似度值集中不存在相似度值大于或等于所述第三预设阈值,则所述识别结果为识别失败。例如:相似度集中共有0.6、0.7、0.9三个相似度值,当所述第三预设阈值为0.85时,存在相似度值0.9大于第三预设阈值0.85,所述识别结果为识别成功;当所述第三预设阈值为0.95时,不存在相似度值大于或等于0.95,所述识别结果为识别失败。
如图3所示,是本申请生物特征图像识别装置的功能模块图。
本申请所述生物特征图像识别装置100可以安装于电子设备中。根据实现的功能,所述生物特征图像识别装置可以包括训练样本生成模块101、模型训练模块102、图像识别模块103,本发所述模块也可以称之为单元,是指一种能够被电子设备处理器所执行,并且能够完成固定功能的一系列计算机程序段,其存储在电子设备的存储器中。
在本实施例中,关于各模块/单元的功能如下:
所述训练样本生成模块101用于获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集。
本申请实施例中,所述第一训练图像集为虹膜图像集合,所述虹膜图像为包含虹膜区域的眼睛图像。
本领域技术人员应该了解,所述生成对抗网络模型分为生成器和鉴别器两个部分。在训练过程中,生成器用来根据所述第一训练图像集生成图像,所述鉴别器用来判断生成图像是生成器生成的虚假图像还是所述第一训练图像集中的对应的真实图像,使生成器生成的生成图像,能逼近所述第一训练图像集中真实图像,根据此时的生成网络得到一个生成模型,用来对生成新的图像,解决图像样本过小的问题。
详细地,本申请实施例中所述训练样本生成模块101利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,包括:构建第一损失函数;基于所述第一损失函数,利用所述第一训练图像集对所述第一生成对抗网络模型进行生成器与鉴别器的交替迭代训练;当所述第一损失函数的值达到第一预设阈值时,停止训练,得到所述数据增强模型;
其中,所述第一损失函数为:
Figure PCTCN2021097072-appb-000006
其中,所述L GAN为预构建的对抗损失函数,z表示预设的随机参数变量,D为所述第一生成对抗网络模型中的生成器,G为所述第一生成对抗网络模型中的鉴别器,x为所述生成器生成图像中的真实图像,P (x)为所述生成图像中真实图像的概率分布,P (z)为所述生成图像中虚假图像的概率分布,E为期望值计算函数。
本申请实施例中,所述第一样本图像集是与所述第一训练图像集不同的虹膜图像集合。
由于所述第一样本图像集样本较小,且不易获取,所以本申请实施例所述训练样本生成模块101利用所述数据增强模型对所述第一样本图像集中的所有图像进行处理,得到所述第二样本图像集。例如:所述第一样本图像集中包含图像a、b,将图像a输入所述数据增强模型得到图像A,将图像b输入所述数据增强模型得到图像B,汇总图像A及图像B得到第二样本图像集。
所述训练样本生成模块101汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
本申请实施例中,为了提高后续模型的鲁棒性,所述训练样本生成模块101对所述第一目标样本图像集进行数据扩增。
详细地,本申请实施例中,所述训练样本生成模块101对所述第一目标样本图像集进行数据扩增,包括:对所述第一目标样本图像集中的所有图像进行平移翻转及颜色调整操作,得到扩增图像集;汇总所述扩增图像集及所述第一目标样本图像集;对汇总后的图像集中的图像进行标签区域标记,得到所述第二目标样本图像集。本申请实施例中,所述标签区域为虹膜区域,较佳地,本申请实施例可使用Label Me图像标注工具人工进行标签区域标记。
本申请的另一实施例中,为了保证数据的隐私性,所述第二目标样本图像集可以存储在区块链节点中。
所述模型训练模块102用于利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型。
较佳地,本申请实施例中,所述深度学习网络模型可以包括卷积神经网络模型等。
进一步地,本申请实施例所述模型训练模块102利用如下手段对预构建的深度学习网络模型进行训练,包括:
步骤A:根据预设的卷积池化次数,对所述第二目标样本图像集进行卷积池化操作,得到特征集;
步骤B:利用预设的激活函数对所述特征集进行计算得到预测值,获取所述第二目标样本图像集中每张图像对应的所述标签区域的标签值,根据所述预测值及所述标签值,利用预构建的第二损失函数进行计算,得到损失值;
本申请实施例中所述标签值与所述标签区域是一一对应的,例如:所述标签区域对应的标签值为1,非标签区域对应的标签值为0。
步骤C:对比所述损失值与第二预设阈值的大小,当所述损失值大于或等于所述第二预设阈值时,返回所述步骤A;或者当所述损失值小于所述第二预设阈值时,停止训练,得到所述图像识别模型。
详细地,本申请实施例中所述模型训练模块102对所述第二目标样本图像集进行卷积池化操作得到第一特征集,包括:对所述第二目标样本图像集进行卷积操作得到第一卷积数据集;对所述第一卷积数据集进行最大池化操作得到所述特征集。
进一步地,所述卷积操作为:
Figure PCTCN2021097072-appb-000007
其中,ω’表示所述第一卷积数据集的通道数,ω表示所述第二目标样本图像集的通道数,k为预设卷积核的大小,f为预设卷积操作的步幅,p为预设数据补零矩阵。
进一步地,本申请较佳实施例所述第一激活函数包括:
Figure PCTCN2021097072-appb-000008
其中,μ t表示所述预测值,s表示所述特征集中的数据。
详细地,本申请较佳实施例所述第一损失函数包括:
Figure PCTCN2021097072-appb-000009
其中,L ce表示所述损失值,N为所述第二目标样本图像集的数据数目,i为正整数,y i为所述标签值,p i为所述预测值。
所述图像识别模块103用于当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
本申请实施例中,所述待识别生物特征图像为需要识别的虹膜图像。
详细地,本申请实施例中,所述图像识别模块103利用所述图像识别模型对所述待识 别生物特征图像进行特征提取,得到特征向量,包括:利用所述图像识别模型对所述待识别图像进行图像识别,提取完成图像识别后所述图像识别模型中全连接层的输出值,得到所述特征向量。
进一步地,本申请实施例中,所述图像识别模块103利用如下手段得到所述特征向量,包括:根据所述图像识别模型中全连接层中节点的顺序,提取完成图像识别后的所述图像识别模型中全连接层所有节点的输出值并纵向组合,得到所述特征向量,例如:全连接层共有3个节点,按顺序分别为第一节点、第二节点、第三节点,完成图像识别后全连接层第一个节点的输出值为1、第二个节点的输出值为3、第三个节点的输出值为5,将1、3、5三个特征值按节点顺序纵向组合得到特征向量
Figure PCTCN2021097072-appb-000010
详细地,本申请实施例中所述图像识别模块103利用如下手段得到识别结果,包括:
计算所述特征向量与所述图像特征向量库中每个图像特征向量的相似度,得到对应的相似度值;
较佳地,本申请实施例利用余弦相似度算法计算所述特征向量与所述图像特征向量库中每个图像特征向量的相似度。
汇总所有的所述相似度值,得到相似度值集;
根据所述相似度值集进行筛选比对,得到所述识别结果。
本申请实施例中,若所述相似度值集中存在相似度值大于或等于第三预设阈值,则所述识别结果为识别成功;若所述相似度值集中不存在相似度值大于或等于所述第三预设阈值,则所述识别结果为识别失败。例如:相似度集中共有0.6、0.7、0.9三个相似度值,当所述第三预设阈值为0.85时,存在相似度值0.9大于第三预设阈值0.85,所述识别结果为识别成功;当所述第三预设阈值为0.95时,不存在相似度值大于或等于0.95,所述识别结果为识别失败。
如图4所示,是本申请实现生物特征图像识别方法的电子设备的结构示意图。
所述电子设备1可以包括处理器10、存储器11和总线,还可以包括存储在所述存储器11中并可在所述处理器10上运行的计算机程序,如生物特征图像识别程序。
其中,所述存储器11至少包括一种类型的可读存储介质,所述可读存储介质包括闪存、移动硬盘、多媒体卡、卡型存储器(例如:SD或DX存储器等)、磁性存储器、磁盘、光盘等。所述存储器11在一些实施例中可以是电子设备1的内部存储单元,例如该电子设备1的移动硬盘。所述存储器11在另一些实施例中也可以是电子设备1的外部存储设备,例如电子设备1上配备的插接式移动硬盘、智能存储卡(Smart Media Card,SMC)、安全数字(Secure Digital,SD)卡、闪存卡(Flash Card)等。进一步地,所述存储器11还可以既包括电子设备1的内部存储单元也包括外部存储设备。所述存储器11不仅可以用于存储安装于电子设备1的应用软件及各类数据,例如生物特征图像识别程序的代码等,还可以用于暂时地存储已经输出或者将要输出的数据。
所述处理器10在一些实施例中可以由集成电路组成,例如可以由单个封装的集成电路所组成,也可以是由多个相同功能或不同功能封装的集成电路所组成,包括一个或者多个中央处理器(Central Processing unit,CPU)、微处理器、数字处理芯片、图形处理器及各种控制芯片的组合等。所述处理器10是所述电子设备的控制核心(Control Unit),利用各种接口和线路连接整个电子设备的各个部件,通过运行或执行存储在所述存储器11内的程序或者模块(例如生物特征图像识别程序等),以及调用存储在所述存储器11内的数据,以执行电子设备1的各种功能和处理数据。
所述总线可以是外设部件互连标准(peripheral component interconnect,简称PCI)总线或扩展工业标准结构(extended industry standard architecture,简称EISA)总线等。该总 线可以分为地址总线、数据总线、控制总线等。所述总线被设置为实现所述存储器11以及至少一个处理器10等之间的连接通信。
图4仅示出了具有部件的电子设备,本领域技术人员可以理解的是,图4示出的结构并不构成对所述电子设备1的限定,可以包括比图示更少或者更多的部件,或者组合某些部件,或者不同的部件布置。
例如,尽管未示出,所述电子设备1还可以包括给各个部件供电的电源(比如电池),优选地,电源可以通过电源管理装置与所述至少一个处理器10逻辑相连,从而通过电源管理装置实现充电管理、放电管理、以及功耗管理等功能。电源还可以包括一个或一个以上的直流或交流电源、再充电装置、电源故障检测电路、电源转换器或者逆变器、电源状态指示器等任意组件。所述电子设备1还可以包括多种传感器、蓝牙模块、Wi-Fi模块等,在此不再赘述。
进一步地,所述电子设备1还可以包括网络接口,可选地,所述网络接口可以包括有线接口和/或无线接口(如WI-FI接口、蓝牙接口等),通常用于在该电子设备1与其他电子设备之间建立通信连接。
可选地,该电子设备1还可以包括用户接口,用户接口可以是显示器(Display)、输入单元(比如键盘(Keyboard)),可选地,用户接口还可以是标准的有线接口、无线接口。可选地,在一些实施例中,显示器可以是LED显示器、液晶显示器、触控式液晶显示器以及OLED(Organic Light-Emitting Diode,有机发光二极管)触摸器等。其中,显示器也可以适当的称为显示屏或显示单元,用于显示在电子设备1中处理的信息以及用于显示可视化的用户界面。
应该了解,所述实施例仅为说明之用,在专利申请范围上并不受此结构的限制。
所述电子设备1中的所述存储器11存储的生物特征图像识别程序12是多个计算机程序的组合,在所述处理器10中运行时,可以实现:
获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;
获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;
汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;
利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
具体地,所述处理器10对上述计算机程序的具体实现方法可参考图1对应实施例中相关步骤的描述,在此不赘述。
进一步地,所述电子设备1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。所述计算机可读介质可以是非易失性的,也可以是易失性的。所述计算机可读介质可以包括:能够携带所述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)。
进一步地,所述计算机可用存储介质可主要包括存储程序区和存储数据区,其中,存储程序区可存储操作系统、至少一个功能所需的应用程序等;存储数据区可存储根据区块链节点的使用所创建的数据等,所述应用程序被处理器执行可以实现以下步骤:
获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训 练,得到数据增强模型;
获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;
汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;
利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
在本申请所提供的几个实施例中,应该理解到,所揭露的设备,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,所述模块的划分,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。
所述作为分离部件说明的模块可以是或者也可以不是物理上分开的,作为模块显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
另外,在本申请各个实施例中的各功能模块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,也可以采用硬件加软件功能模块的形式实现。
对于本领域技术人员而言,显然本申请不限于上述示范性实施例的细节,而且在不背离本申请的精神或基本特征的情况下,能够以其他的具体形式实现本申请。
因此,无论从哪一点来看,均应将实施例看作是示范性的,而且是非限制性的,本申请的范围由所附权利要求而不是上述说明限定,因此旨在将落在权利要求的等同要件的含义和范围内的所有变化涵括在本申请内。不应将权利要求中的任何附关联图标记视为限制所涉及的权利要求。
本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
此外,显然“包括”一词不排除其他单元或步骤,单数不排除复数。系统权利要求中陈述的多个单元或装置也可以由一个单元或装置通过软件或者硬件来实现。第二等词语用来表示名称,而并不表示任何特定的顺序。
最后应说明的是,以上实施例仅用以说明本申请的技术方案而非限制,尽管参照较佳实施例对本申请进行了详细说明,本领域的普通技术人员应当理解,可以对本申请的技术方案进行修改或等同替换,而不脱离本申请技术方案的精神和范围。

Claims (20)

  1. 一种生物特征图像识别方法,其中,所述方法包括:
    获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;
    获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;
    汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
    对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
    利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
    当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;
    利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
  2. 如权利要求1所述的生物特征图像识别方法,其中,所述利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型,包括:
    构建第一损失函数;
    基于所述第一损失函数,利用所述第一训练图像集对所述生成对抗网络模型进行生成器与鉴别器的交替迭代训练;
    当所述第一损失函数的值达到第一预设阈值时,停止训练,得到所述数据增强模型。
  3. 如权利要求1所述的生物特征图像识别方法,其中,所述对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集,包括:
    对所述第一目标样本图像集中的所有图像进行平移翻转及颜色调整操作,得到扩增图像集;
    汇总所述扩增图像集及所述第一目标样本图像集;
    对汇总后的图像集中的图像进行标签区域标记,得到所述第二目标样本图像集。
  4. 如权利要求3所述的生物特征图像识别方法,其中,所述利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型,包括:
    特征提取步骤:根据预设的卷积池化次数,对所述第二目标样本图像集进行卷积池化操作,得到特征集;
    损失计算步骤:利用预设的激活函数对所述特征集进行计算得到预测值,获取所述第二目标样本图像集中每张图像对应的所述标签区域的标签值,根据所述预测值及所述标签值,利用预构建的第二损失函数进行计算,得到损失值;
    训练判断步骤:对比所述损失值与第二预设阈值的大小,当所述损失值大于或等于所述第二预设阈值时,返回所述特征提取步骤;或者当所述损失值小于所述第二预设阈值时,停止训练,得到所述图像识别模型。
  5. 如权利要求1所述的生物特征图像识别方法,其中,所述利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量,包括:
    利用所述图像识别模型对所述待识别图像进行图像识别;
    提取完成图像识别后所述图像识别模型中全连接层的输出值,得到所述特征向量。
  6. 如权利要求1所述的生物特征图像识别方法,其中,所述利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果,包括:
    计算所述特征向量与所述图像特征向量库中每个图像特征向量的相似度,得到对应的相似度值;
    汇总所有的所述相似度值,得到相似度值集;
    根据所述相似度值集进行筛选比对,得到所述识别结果。
  7. 如权利要求5所述的生物特征图像识别方法,其中,所述根据所述相似度值集进行筛选比对,得到所述识别结果,包括:
    若所述相似度值集中存在相似度值大于或等于第三预设阈值,则所述识别结果为识别通过;或者
    若所述相似度值集中不存在相似度值大于或等于所述第三预设阈值,则所述识别结果为识别不通过。
  8. 一种生物特征图像识别装置,其中,包括:
    训练样本生成模块,用于获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
    模型训练模块,用于利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
    图像识别模块,用于当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
  9. 一种电子设备,其中,所述电子设备包括:
    至少一个处理器;以及,
    与所述至少一个处理器通信连接的存储器;其中,
    所述存储器存储有可被所述至少一个处理器执行的计算机程序,所述计算机程序被所述至少一个处理器执行,以使所述至少一个处理器能够执行如下步骤:
    获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;
    获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;
    汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
    对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
    利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
    当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;
    利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
  10. 如权利要求9所述的电子设备,其中,所述利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型,包括:
    构建第一损失函数;
    基于所述第一损失函数,利用所述第一训练图像集对所述生成对抗网络模型进行生成器与鉴别器的交替迭代训练;
    当所述第一损失函数的值达到第一预设阈值时,停止训练,得到所述数据增强模型。
  11. 如权利要求9所述的电子设备,其中,所述对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集,包括:
    对所述第一目标样本图像集中的所有图像进行平移翻转及颜色调整操作,得到扩增图像集;
    汇总所述扩增图像集及所述第一目标样本图像集;
    对汇总后的图像集中的图像进行标签区域标记,得到所述第二目标样本图像集。
  12. 如权利要求11所述的电子设备,其中,所述利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型,包括:
    特征提取步骤:根据预设的卷积池化次数,对所述第二目标样本图像集进行卷积池化操作,得到特征集;
    损失计算步骤:利用预设的激活函数对所述特征集进行计算得到预测值,获取所述第二目标样本图像集中每张图像对应的所述标签区域的标签值,根据所述预测值及所述标签值,利用预构建的第二损失函数进行计算,得到损失值;
    训练判断步骤:对比所述损失值与第二预设阈值的大小,当所述损失值大于或等于所述第二预设阈值时,返回所述特征提取步骤;或者当所述损失值小于所述第二预设阈值时,停止训练,得到所述图像识别模型。
  13. 如权利要求9所述的电子设备,其中,所述利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量,包括:
    利用所述图像识别模型对所述待识别图像进行图像识别;
    提取完成图像识别后所述图像识别模型中全连接层的输出值,得到所述特征向量。
  14. 如权利要求9所述的电子设备,其中,所述利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果,包括:
    计算所述特征向量与所述图像特征向量库中每个图像特征向量的相似度,得到对应的相似度值;
    汇总所有的所述相似度值,得到相似度值集;
    根据所述相似度值集进行筛选比对,得到所述识别结果。
  15. 如权利要求13所述的电子设备,其中,所述根据所述相似度值集进行筛选比对,得到所述识别结果,包括:
    若所述相似度值集中存在相似度值大于或等于第三预设阈值,则所述识别结果为识别通过;或者
    若所述相似度值集中不存在相似度值大于或等于所述第三预设阈值,则所述识别结果为识别不通过。
  16. 一种计算机可读存储介质,存储有计算机程序,其中,所述计算机程序被处理器执行时实现如下步骤:
    获取第一训练图像集,利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型;
    获取第一样本图像集,利用所述数据增强模型对所述第一样本图像集进行数据增强,得到第二样本图像集;
    汇总所述第一样本图像集及所述第二样本图像集,得到第一目标样本图像集;
    对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集;
    利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型;
    当接收到待识别生物特征图像时,利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量;
    利用所述特征向量在预设的图像特征向量库中进行比对识别,得到识别结果。
  17. 如权利要求16所述的计算机可读存储介质,其中,所述利用所述第一训练图像集对预构建的生成对抗网络模型进行训练,得到数据增强模型,包括:
    构建第一损失函数;
    基于所述第一损失函数,利用所述第一训练图像集对所述生成对抗网络模型进行生成器与鉴别器的交替迭代训练;
    当所述第一损失函数的值达到第一预设阈值时,停止训练,得到所述数据增强模型。
  18. 如权利要求16所述的计算机可读存储介质,其中,所述对所述第一目标样本图像集进行数据扩增,得到第二目标样本图像集,包括:
    对所述第一目标样本图像集中的所有图像进行平移翻转及颜色调整操作,得到扩增图像集;
    汇总所述扩增图像集及所述第一目标样本图像集;
    对汇总后的图像集中的图像进行标签区域标记,得到所述第二目标样本图像集。
  19. 如权利要求18所述的计算机可读存储介质,其中,所述利用所述第二目标样本图像集对预构建的深度学习网络模型进行训练,得到图像识别模型,包括:
    特征提取步骤:根据预设的卷积池化次数,对所述第二目标样本图像集进行卷积池化操作,得到特征集;
    损失计算步骤:利用预设的激活函数对所述特征集进行计算得到预测值,获取所述第二目标样本图像集中每张图像对应的所述标签区域的标签值,根据所述预测值及所述标签值,利用预构建的第二损失函数进行计算,得到损失值;
    训练判断步骤:对比所述损失值与第二预设阈值的大小,当所述损失值大于或等于所述第二预设阈值时,返回所述特征提取步骤;或者当所述损失值小于所述第二预设阈值时,停止训练,得到所述图像识别模型。
  20. 如权利要求16所述的计算机可读存储介质,其中,所述利用所述图像识别模型对所述待识别生物特征图像进行特征提取,得到特征向量,包括:
    利用所述图像识别模型对所述待识别图像进行图像识别;
    提取完成图像识别后所述图像识别模型中全连接层的输出值,得到所述特征向量。
PCT/CN2021/097072 2020-11-23 2021-05-30 生物特征图像识别方法、装置、电子设备及可读存储介质 WO2022105179A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011322604.X 2020-11-23
CN202011322604.XA CN112396005A (zh) 2020-11-23 2020-11-23 生物特征图像识别方法、装置、电子设备及可读存储介质

Publications (1)

Publication Number Publication Date
WO2022105179A1 true WO2022105179A1 (zh) 2022-05-27

Family

ID=74606952

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/097072 WO2022105179A1 (zh) 2020-11-23 2021-05-30 生物特征图像识别方法、装置、电子设备及可读存储介质

Country Status (2)

Country Link
CN (1) CN112396005A (zh)
WO (1) WO2022105179A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115035463A (zh) * 2022-08-09 2022-09-09 阿里巴巴(中国)有限公司 行为识别方法、装置、设备和存储介质
CN116052141A (zh) * 2023-03-30 2023-05-02 北京市农林科学院智能装备技术研究中心 作物生育期的识别方法、装置、设备及介质

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112396005A (zh) * 2020-11-23 2021-02-23 平安科技(深圳)有限公司 生物特征图像识别方法、装置、电子设备及可读存储介质
CN114549928A (zh) * 2022-02-21 2022-05-27 平安科技(深圳)有限公司 图像增强处理方法、装置、计算机设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108921123A (zh) * 2018-07-17 2018-11-30 重庆科技学院 一种基于双重数据增强的人脸识别方法
CN110889457A (zh) * 2019-12-03 2020-03-17 深圳奇迹智慧网络有限公司 样本图像分类训练方法、装置、计算机设备和存储介质
CN111666994A (zh) * 2020-05-28 2020-09-15 平安科技(深圳)有限公司 样本图像数据增强方法、装置、电子设备及存储介质
US20200335086A1 (en) * 2019-04-19 2020-10-22 Behavioral Signal Technologies, Inc. Speech data augmentation
CN112396005A (zh) * 2020-11-23 2021-02-23 平安科技(深圳)有限公司 生物特征图像识别方法、装置、电子设备及可读存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108921123A (zh) * 2018-07-17 2018-11-30 重庆科技学院 一种基于双重数据增强的人脸识别方法
US20200335086A1 (en) * 2019-04-19 2020-10-22 Behavioral Signal Technologies, Inc. Speech data augmentation
CN110889457A (zh) * 2019-12-03 2020-03-17 深圳奇迹智慧网络有限公司 样本图像分类训练方法、装置、计算机设备和存储介质
CN111666994A (zh) * 2020-05-28 2020-09-15 平安科技(深圳)有限公司 样本图像数据增强方法、装置、电子设备及存储介质
CN112396005A (zh) * 2020-11-23 2021-02-23 平安科技(深圳)有限公司 生物特征图像识别方法、装置、电子设备及可读存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
QIAN YANMIN; HU HU; TAN TIAN: "Data augmentation using generative adversarial networks for robust speech recognition", SPEECH COMMUNICATION, ELSEVIER SCIENCE PUBLISHERS , AMSTERDAM, NL, vol. 114, 19 August 2019 (2019-08-19), NL , pages 1 - 9, XP085889819, ISSN: 0167-6393, DOI: 10.1016/j.specom.2019.08.006 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115035463A (zh) * 2022-08-09 2022-09-09 阿里巴巴(中国)有限公司 行为识别方法、装置、设备和存储介质
CN115035463B (zh) * 2022-08-09 2023-01-17 阿里巴巴(中国)有限公司 行为识别方法、装置、设备和存储介质
CN116052141A (zh) * 2023-03-30 2023-05-02 北京市农林科学院智能装备技术研究中心 作物生育期的识别方法、装置、设备及介质
CN116052141B (zh) * 2023-03-30 2023-06-27 北京市农林科学院智能装备技术研究中心 作物生育期的识别方法、装置、设备及介质

Also Published As

Publication number Publication date
CN112396005A (zh) 2021-02-23

Similar Documents

Publication Publication Date Title
WO2022105179A1 (zh) 生物特征图像识别方法、装置、电子设备及可读存储介质
WO2021232594A1 (zh) 语音情绪识别方法、装置、电子设备及存储介质
EP3869385B1 (en) Method for extracting structural data from image, apparatus and device
WO2021189855A1 (zh) 基于ct序列的图像识别方法、装置、电子设备及介质
WO2022048209A1 (zh) 车牌识别方法、装置、电子设备及存储介质
WO2021151270A1 (zh) 图像结构化数据提取方法、装置、设备及存储介质
WO2022141859A1 (zh) 图像检测方法、装置、电子设备及存储介质
WO2022156066A1 (zh) 文字识别方法、装置、电子设备及存储介质
WO2021151313A1 (zh) 证件鉴伪方法、装置、电子设备及存储介质
CN113157927B (zh) 文本分类方法、装置、电子设备及可读存储介质
CN112699775A (zh) 基于深度学习的证件识别方法、装置、设备及存储介质
CN112528616B (zh) 业务表单生成方法、装置、电子设备及计算机存储介质
JP7364709B2 (ja) 機械学習および自然言語処理を利用したワクチン接種データの抽出および確認
WO2021068613A1 (zh) 面部识别方法、装置、设备及计算机可读存储介质
CN113157739B (zh) 跨模态检索方法、装置、电子设备及存储介质
CN113434542B (zh) 数据关系识别方法、装置、电子设备及存储介质
CN114708461A (zh) 基于多模态学习模型的分类方法、装置、设备及存储介质
WO2022227192A1 (zh) 图像分类方法、装置、电子设备及介质
CN112528903B (zh) 人脸图像获取方法、装置、电子设备及介质
CN112560855B (zh) 图像信息提取方法、装置、电子设备及存储介质
WO2022222228A1 (zh) 文本不良信息识别方法、装置、电子设备及存储介质
CN112580505B (zh) 网点开关门状态识别方法、装置、电子设备及存储介质
CN113419951B (zh) 人工智能模型优化方法、装置、电子设备及存储介质
WO2022227191A1 (zh) 非主动活体检测方法、装置、电子设备及存储介质
WO2022141867A1 (zh) 语音识别方法、装置、电子设备及可读存储介质

Legal Events

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

Ref document number: 21893334

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: 21893334

Country of ref document: EP

Kind code of ref document: A1