CN107077617B - Fingerprint extraction method and device - Google Patents
Fingerprint extraction method and device Download PDFInfo
- Publication number
- CN107077617B CN107077617B CN201780000029.1A CN201780000029A CN107077617B CN 107077617 B CN107077617 B CN 107077617B CN 201780000029 A CN201780000029 A CN 201780000029A CN 107077617 B CN107077617 B CN 107077617B
- Authority
- CN
- China
- Prior art keywords
- pixel point
- gaussian
- pixel
- pixel value
- value
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
- G06V40/1347—Preprocessing; Feature extraction
Abstract
The embodiment of the invention relates to the field of fingerprint identification, and discloses a fingerprint extraction method and device. The fingerprint extraction method of the invention comprises the following steps: when finger touch is detected, acquiring an induction pixel value of each pixel point in a fingerprint induction area; identifying the corresponding grain characteristics of each pixel point according to the induction pixel value of each pixel point and a preset Gaussian mixture background model of each pixel point; and generating a fingerprint image of the finger according to the corresponding texture characteristics of each pixel point. The invention also provides a fingerprint extraction device; according to the invention, the fingerprint image is extracted based on the mixed Gaussian background model, so that the high-quality fingerprint image can be extracted.
Description
Technical Field
The embodiment of the invention relates to the technical field of fingerprint identification, in particular to a fingerprint extraction method and a fingerprint extraction device.
Background
Nowadays, with the rapid development of intelligent terminal equipment, a fingerprint identification technology is widely applied to electronic equipment such as mobile phones and tablet computers, and fingerprint identification becomes one of the main modes of mobile phone unlocking, mobile payment and the like, so that great convenience is brought to the life of people; the fingerprint extraction is an important link in the fingerprint identification technology, and the accuracy of the fingerprint identification is directly influenced by the quality of the fingerprint extraction.
In the process of implementing the invention, the inventor finds that at least the following problems exist in the prior art:
in the prior art, fingerprint extraction generally reads sensing data twice through a fingerprint sensing chip, wherein one time is a value of a pixel corresponding to the chip in a finger non-touch state, and the other time is a value of a pixel corresponding to the chip in a finger touch state, and a fingerprint image is obtained by comparing differences of the two data. Conventionally, the fingerprint image is obtained by subtracting the sensed images twice. However, due to the influence of external factors such as temperature, the sensing data read by the fingerprint sensing chip is unstable, so that the fingerprint image obtained by subtracting the sensing data twice in the conventional method contains a lot of noise, and even the fingerprint image cannot be obtained.
Disclosure of Invention
The embodiment of the invention aims to provide a fingerprint extraction method and a fingerprint extraction device, which can extract a high-quality fingerprint image based on a mixed Gaussian background model.
In order to solve the above technical problem, an embodiment of the present invention provides a fingerprint extraction method, including: when finger touch is detected, acquiring an induction pixel value of each pixel point in a fingerprint induction area; identifying the corresponding grain characteristics of each pixel point according to the induction pixel value of each pixel point and a preset Gaussian mixture background model of each pixel point; and generating a fingerprint image of the finger according to the corresponding texture characteristics of each pixel point.
An embodiment of the present invention further provides a fingerprint extraction device, including: the pixel value acquisition module is used for acquiring the sensing pixel value of each pixel point in the fingerprint sensing area when finger touch is detected; the grain characteristic identification module is used for identifying the corresponding grain characteristic of each pixel point according to the induction pixel value of each pixel point and a preset Gaussian mixture background model of each pixel point; and the fingerprint image generation module is used for generating a fingerprint image of the finger according to the corresponding line characteristics of each pixel point.
Compared with the prior art, the method has the advantages that the induction pixel value of each pixel point in the fingerprint induction area when the finger touches the fingerprint induction area and the preset Gaussian mixture background model of each pixel point are obtained, then the texture characteristics corresponding to each pixel point are identified, the texture characteristics are arranged according to the pixel points, and the fingerprint image of the finger is generated; the mixed Gaussian background model has a good description effect on unstable pixels, and the mixed Gaussian background model is used for extracting high-quality fingerprint images.
In addition, in the mixed gaussian background model according to the induced pixel value of every pixel and every pixel of predetermineeing, discern the corresponding line characteristic of every pixel, specifically include: for each pixel point, judging whether the induction pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point; and when the induction pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point, identifying the grain characteristics of the pixel point as concave grain. The embodiment provides a specific identification mode; namely, the induction pixel value of the pixel point of the concave grain is matched with the mixed Gaussian background model of the preset pixel point, and the concave grain of the fingerprint image is obtained according to the induction pixel value.
In addition, in identifying the line characteristics corresponding to each pixel point according to the sensing pixel value of each pixel point and the preset Gaussian mixture background model of each pixel point, the method further comprises the following steps: when the induction pixel value of the pixel point is not matched with the Gaussian mixture background model of the pixel point, judging whether the induction pixel value of the pixel point meets the preset convex-pattern path matching condition or not; when the induction pixel value of the pixel point meets the convex pattern path matching condition, identifying the grain characteristics of the pixel point as a convex pattern path; and when the induction pixel value of the pixel point does not meet the convex grain matching condition, identifying the grain characteristics of the pixel point as concave grain. Namely, when the induction pixel value of the pixel point is not matched with the preset Gaussian mixture background model of the pixel point but meets the preset convex pattern path matching condition, the texture characteristic of the pixel point is considered as a convex pattern path, otherwise, the texture characteristic of the pixel point is a concave pattern path; the embodiment further improves the specific identification mode; therefore, the convex lines and the concave lines of the fingerprint image can be simultaneously obtained so as to obtain a complete fingerprint image.
In addition, the mixed Gaussian background model comprises a plurality of Gaussian components which are sequentially arranged; judging whether the induction pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point, and specifically: sequentially comparing the induction pixel value of the pixel point with a plurality of Gaussian components, and judging whether a Gaussian component matched with the induction pixel value of the pixel point exists or not; when a Gaussian component matched with the induction pixel value of the pixel point exists, judging that the induction pixel value of the pixel point is matched with a Gaussian mixture background model of the pixel point; and when the Gaussian component matched with the induction pixel value of the pixel point does not exist, judging that the induction pixel value of the pixel point is not matched with the mixed Gaussian background model of the pixel point. The embodiment provides a specific way for judging whether the sensing pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point; the Gaussian mixture background model comprises a plurality of Gaussian components and can effectively describe the multimodal state of the induction pixel value.
In addition, compare the response pixel value of pixel and a plurality of gaussian component in proper order, judge whether exist one with the gaussian component of the response pixel value matching of pixel, specifically include: calculating the difference value between the induction pixel value of the pixel point and the sample mean value in each Gaussian component, and obtaining the difference value with the minimum absolute value as a matching parameter; judging whether the matching parameter is smaller than or equal to a preset first threshold value; when the matching parameter is smaller than or equal to a preset first threshold value, judging that a Gaussian component matched with the induction pixel value of the pixel point exists; when the matching parameter is larger than a first threshold value, judging that a Gaussian component matched with the induction pixel value of the pixel point does not exist; the convex line matching conditions include: the sensing pixel value of the pixel point is larger than a preset second threshold value; the second threshold is greater than the first threshold. The embodiment further refines the specific way of judging whether to match; the minimum difference value in the absolute value of the difference between the sample mean value in the Gaussian component and the sensing pixel value of the pixel is used as a matching parameter, and the accuracy of judgment is further improved.
In addition, after generating the fingerprint image of the finger according to the corresponding line characteristics of each pixel point, the method further comprises the following steps: and when no finger touch is detected, updating the Gaussian mixture background model of each pixel point. In this embodiment, the gaussian mixture background model of each preset pixel point is continuously updated, so as to ensure that the gaussian mixture background model is updated in time when the environment changes.
In addition, the preset mode of the gaussian background model of each pixel specifically includes: creating a Gaussian mixture model of the pixel points; the Gaussian mixture model comprises a plurality of Gaussian components which are sequentially arranged; according to the basic pixel values of the pixel points acquired for multiple times, the Gaussian mixture model is subjected to multiple learning and updating; the basic pixel value of the pixel point is obtained when no finger touches the pixel point; normalizing the weights of a plurality of Gaussian components in the mixed Gaussian model after repeated learning and updating; and selecting a plurality of Gaussian components from the plurality of Gaussian components after normalization processing according to a preset selection rule to form a mixed Gaussian background model. The embodiment provides a specific implementation mode of a preset mixed Gaussian background model, and the actual design requirements are met.
In addition, the learning update method specifically includes: comparing the basic pixel value of the pixel point with a plurality of Gaussian components which are sequentially arranged, and judging whether a Gaussian component matched with the basic pixel value of the pixel point exists or not; when a Gaussian component matched with the basic pixel value of the pixel point exists, updating the weight of the Gaussian component according to a preset weight increment, and updating the sample mean and the sample variance of the Gaussian component according to the basic pixel value of the pixel point; and reordering the plurality of Gaussian components according to a preset ordering rule. The embodiment provides a specific implementation mode of learning updating; namely, when a gaussian component matched with the basic pixel value of the pixel point exists, the weight of the gaussian component is higher, the weight of the gaussian component is increased according to a preset increment, and the sample mean-square error and the sample variance of the gaussian component are updated, so that the gaussian component in the mixed gaussian background model is more reasonably arranged.
In addition, before reordering the plurality of gaussian components according to the preset ordering rule, the method further comprises: when the Gaussian component matched with the basic pixel value of the pixel point does not exist, deleting the Gaussian component ranked at the last in the Gaussian mixture model; adding a new Gaussian component in the Gaussian mixture model according to the basic pixel value of the pixel point; and updating the weight of the Gaussian components except the newly added Gaussian component in the mixed Gaussian model according to the preset weight decrement. The embodiment further improves the learning updating mode; that is, when there is no gaussian component matching the basic pixel value of the pixel, it indicates that the gaussian component in the mixed gaussian background model needs to be updated, and at this time, a new gaussian component created according to the basic pixel value of the pixel is added, and a gaussian component ranked at the end is deleted, thereby ensuring the accuracy of the gaussian background model.
Drawings
One or more embodiments are illustrated by way of example in the accompanying drawings, which correspond to the figures in which like reference numerals refer to similar elements and which are not to scale unless otherwise specified.
Fig. 1 is a detailed flowchart of a fingerprint extraction method according to a first embodiment of the present invention;
fig. 2 is a specific flowchart for identifying a texture feature corresponding to each pixel point according to a sensed pixel value of each pixel point and a preset gaussian background model of each pixel point according to a second embodiment of the present invention;
FIG. 3 is a specific flowchart of a preset method of a Gaussian mixture background model for each pixel according to a second embodiment of the present invention;
fig. 4 is a detailed flowchart of a fingerprint extraction method according to a third embodiment of the present invention;
fig. 5 is a block schematic diagram of a fingerprint extraction device according to a fourth embodiment of the present invention;
fig. 6 is a schematic block diagram of a texture feature recognition module according to a fifth embodiment of the present invention;
fig. 7 is a block schematic diagram of a fingerprint extraction device according to a sixth embodiment of the present invention;
fig. 8 is a block diagram illustrating a model presetting module according to a sixth embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, it will be appreciated by those of ordinary skill in the art that numerous technical details are set forth in order to provide a better understanding of the present application in various embodiments of the present invention. However, the technical solution claimed in the present application can be implemented without these technical details and various changes and modifications based on the following embodiments.
The first embodiment of the invention relates to a fingerprint extraction method, which is applied to terminal equipment such as a smart phone. The specific flow of the fingerprint extraction method is shown in fig. 1.
Step 101, judging whether the finger touches.
In particular, a sensor in the terminal device may detect whether a finger touches, and the sensor may be a pressure sensor or other sensor that can detect finger pressure or touch.
102, acquiring an induction pixel value of each pixel point in the fingerprint induction area.
Specifically, when the terminal device detects that a finger touches the terminal device, the fingerprint sensor collects a sensing pixel value corresponding to each pixel point in the fingerprint sensing area at the moment. More specifically, the data collected by the fingerprint sensor is usually an M × N matrix, and each element in the matrix corresponds to a pixel value collected by the fingerprint sensor for each pixel.
And 103, identifying the corresponding texture characteristics of each pixel point according to the induction pixel value of each pixel point and a preset Gaussian mixture background model of each pixel point.
Specifically, the sensing pixel values of all pixel points acquired by the fingerprint sensor in a finger touch state and a finger non-touch state are different, and the preset Gaussian mixture background model of each pixel point is the sensing pixel value of each pixel point acquired in the finger non-touch state; in this embodiment, the sensing pixel value of each pixel point collected in the state where the finger is not touched may be referred to as a basic pixel value corresponding to each pixel point; wherein each pixel point corresponds to a mixed gaussian background model.
Therefore, for each pixel point, the induction pixel value acquired under the finger touch state is compared and analyzed with the Gaussian mixture background model, and the corresponding line characteristic of each pixel point can be obtained.
And 104, generating a fingerprint image of the finger according to the corresponding texture characteristics of each pixel point.
Specifically, after the texture features of each pixel point pair are obtained, the texture features are arranged according to the pixel points, so that the fingerprint image of the finger can be obtained.
How to acquire a fingerprint image of a finger is further specified as follows: after the texture characteristics corresponding to each pixel point are obtained, the convex ridge is set as a mark bit '1', the mark bit is represented by a point with a darker color, the concave ridge is set as a mark bit '0', and the mark bit is represented by a point with a lighter color, so that the texture characteristics of each pixel point can be represented by '1' and '0', and then the corresponding color points are used for replacement, and then the fingerprint image of the finger can be obtained.
Compared with the prior art, the method has the advantages that the induction pixel value of each pixel point in the fingerprint induction area when the finger touches the fingerprint induction area and the preset Gaussian mixture background model of each pixel point are obtained, then the texture characteristics corresponding to each pixel point are identified, the texture characteristics are arranged according to the pixel points, and the fingerprint image of the finger is generated; the mixed Gaussian background model has a good description effect on unstable pixels, and the mixed Gaussian background model is used for extracting high-quality fingerprint images.
The second embodiment of the present invention relates to a fingerprint extraction method, and this embodiment is a refinement of the first embodiment, and mainly includes: in the second embodiment of the present invention, in step 103 in the first embodiment: and identifying the corresponding grain characteristics of each pixel point according to the induction pixel value of each pixel point and a preset Gaussian mixture background model of each pixel point, and specifically explaining.
In this embodiment, a specific process for identifying the texture features corresponding to each pixel point according to the induced pixel value of each pixel point and the preset gaussian background model of each pixel point is shown in fig. 2.
And step 1031, judging whether the induction pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point. If yes, go to step 1032; if not, go to step 1033.
In this step, a specific flow of the preset mode of the gaussian background model of the pixel point is shown in fig. 3.
In practice, in the basic pixel values of the pixel points collected by the fingerprint sensor, the value of a certain pixel point j is expressed as XjX may bejThe pixel point is expressed as a gaussian mixture model composed of M gaussian components, that is, the gaussian mixture model corresponding to the pixel point is composed of M gaussian components, wherein the gaussian components can be expressed by the probability of the pixel point, and the formula is as follows:
wherein, ω iskRepresenting the weight, μ, of the kth Gaussian component in the Gaussian mixture modelkAnd σkMean and standard deviation of the kth Gaussian component, η (X), respectivelyj,μk,σk) Is a gaussian probability density function expressed as:
the formula (1) can represent the probability of occurrence of the basic pixel value of the pixel point corresponding to the collected pixel j point at a certain time.
Creating a Gaussian mixture model of a pixel point, firstly determining the number M (selected according to the data range of a sensor and can be 3-5) of Gaussian distributions of the Gaussian mixture model, and initializing the same weight omega for each Gaussian distributioninitAnd a larger standard deviation σinitAnd uniformly initializing the mean values of different Gaussian distributions into pixel values of corresponding pixel points in the first frame of sensing data.
In practice, a learning stage is performed from the basic pixel value of the pixel point acquired by the second frame sensor, the parameters of different Gaussian mixture models corresponding to each pixel are continuously learned and updated, and the sensor acquires 1000-2000 frames of images in the whole learning process. It should be noted that, in this embodiment, the number of frames of the image collected by the sensor is not limited, and may fluctuate according to the change of the basic pixel value of the pixel point collected by the sensor.
In the present embodiment, the learning update method specifically includes:
firstly, comparing the basic pixel value of the pixel point with a plurality of Gaussian components which are arranged in sequence, and judging whether a Gaussian component matched with the basic pixel value of the pixel point exists.
And when a Gaussian component matched with the basic pixel value of the pixel point exists, updating the weight of the Gaussian component according to a preset weight increment, and updating the sample mean and the sample variance of the Gaussian component according to the basic pixel value of the pixel point. It should be noted that the basic pixel value of the pixel point is obtained when no finger touches the pixel point.
In practice, for the basic pixel values of the pixel points newly input into the fingerprint sensor, it is checked whether each basic pixel value matches with M gaussian distributions in the corresponding gaussian mixture model, and if the absolute value of the difference between the basic pixel value and the mean value of one of the gaussian component samples is smaller than a first threshold, the basic pixel value is considered to match with the gaussian distribution. At this time, if the gaussian component in the gaussian mixture model is matched, the gaussian component needs to be updated, the weight is increased according to the preset weight increment, and the sample mean and the sample variance of the gaussian component are updated by using the current basic pixel value. While the remaining gaussian components remain unchanged.
In this embodiment, a specific manner of updating the sample mean of the gaussian component by using the current basic pixel value may be represented by the following formula: (1-P) Uk+PXj(ii) a Wherein, UkRepresents the mean of the samples before update, XjRepresenting the sensed pixel value; p represents the degree of matching of the base pixel value with the gaussian component. In practice, the degree of matching may be based on the difference between the base pixel value and the mean of the gaussian component samples to calculate the transformation; the smaller the difference, the higher the matching degree, and the larger the difference, the smaller the matching degree. It should be noted that updating the sample variance of the gaussian component with the current pixel value is also based on thisThe principle is not described herein.
When the Gaussian component matched with the basic pixel value of the pixel point does not exist, deleting the Gaussian component ranked at the last in the Gaussian mixture model; adding a new Gaussian component in the Gaussian mixture model according to the basic pixel value of the pixel point; and updating the weight of the Gaussian components except the newly added Gaussian component in the mixed Gaussian model according to the preset weight decrement.
In practice, for the basic pixel values of the pixel points in the newly input fingerprint sensor, if each basic pixel value is not matched with the M gaussian distributions in the corresponding gaussian mixture model, the last ranked gaussian component in the gaussian mixture model is deleted, and a new gaussian component is created to replace the basic pixel value. Taking the mean value of newly created Gaussian components as the basic pixel value, standard deviation and weight initialization value omega of the pixel pointinitAnd σinit. The mean value and the variance of the rest Gaussian components are unchanged, and the weight is reduced according to the preset weight.
Then, after the updating process of the gaussian components is completed, the gaussian components are reordered according to a preset ordering rule. The preset ordering rule follows the front row of components with large weight and small variance, and the back row with small weight and large variance.
And step 203, normalizing the weight values of the Gaussian components in the mixed Gaussian model after multiple learning and updating.
Specifically, after a plurality of learning updates, the weights of the gaussian components in the gaussian model change, and the sum of the weights of the gaussian components in the gaussian model may not be equal to 1 (greater than 1 or less than 1), so that the weights of the gaussian components after a plurality of updates need to be normalized.
And 204, selecting a plurality of Gaussian components from the plurality of Gaussian components after normalization processing according to a preset selection rule to form a mixed Gaussian background model.
Specifically, Gaussian components are selected according to the arrangement sequence, the weights of the selected Gaussian components are accumulated, and when the accumulated weight reaches a preset value, the selection is stopped. For example, the mixture gaussian model after normalization includes 5 gaussian components, the weights of the 5 gaussian components arranged in sequence are 0.35, 0.25, 0.20, 0.15, and 0.05, respectively, and if the set preset value is 0.8 and the weights of the first three gaussian components are accumulated to reach the preset value, the first three gaussian components are selected to form the mixture gaussian background model.
It should be noted that, the preset value is not limited in this embodiment, and may be set according to experience or requirements; generally, the preset value is set to be less than 1 (when the preset value is equal to 1, all gaussian components in the gaussian mixture model are selected to form the gaussian mixture background model).
In this embodiment, the gaussian background model includes a plurality of gaussian components arranged in sequence, and when there is a gaussian component matching the sensed pixel value of the pixel, it is considered that the sensed pixel value of the pixel matches the gaussian background model of the pixel, otherwise, it is considered that the sensed pixel value of the pixel does not match the gaussian background model of the pixel.
In this embodiment, in determining whether there is a gaussian component matching the sensing pixel value of the pixel, first, a difference between the sensing pixel value of the pixel and a sample mean value in each gaussian component is calculated, and a difference with the smallest absolute value is obtained as a matching parameter.
In practice, a first threshold value is preset, when the matching parameter is smaller than or equal to the preset first threshold value, a Gaussian component matched with the induction pixel value of the pixel point is judged to exist, and the induction pixel value of the pixel point is considered to be matched with the Gaussian mixture background model of the pixel point; when the matching parameter is larger than a first threshold value, judging that a Gaussian component matched with the induction pixel value of the pixel point does not exist;
and step 1032, identifying the texture characteristics of the pixel points as concave texture.
Specifically, the line characteristic of fingerprint divide into concave line and burr way, and the response value that fingerprint concave line and burr way correspond under the finger touch state is also different, and the response pixel value of the pixel that concave line corresponds is matchd with the response pixel value of pixel in the mixture gaussian background model of predetermineeing, consequently, when the response pixel value of judging the pixel matches with the mixture gaussian background model of pixel, then discerns the line characteristic of pixel for concave line way.
And 1033, judging whether the induction pixel value of the pixel point meets a preset convex-pattern-path matching condition. If yes, go to step 1034; if not, go to step 1032.
Specifically, when the sensing pixel value of the pixel point is not matched with the Gaussian mixture background model of the pixel point, if the sensing pixel value of the pixel point meets the preset convex-pattern-path matching condition, the grain characteristics of the pixel point are identified as convex-pattern paths, otherwise, the grain characteristics of the pixel point are identified as concave-pattern paths.
In practice, the second threshold is preset. And when the matching parameter is greater than a preset second threshold value, determining that the induction pixel value of the pixel point meets the convex-pattern circuit matching condition. Wherein the second threshold is greater than the first threshold.
Specifically, if the sensed pixel value of the pixel point satisfies the convex pattern path matching condition, the texture feature of the pixel point is identified as the convex pattern path.
The embodiment provides a preset mode of a mixed Gaussian background model of pixel points and a specific identification mode of the corresponding grain characteristics of each pixel point, and is a perfect description of the first embodiment so as to meet the actual design requirements.
A third embodiment of the present invention relates to a fingerprint extraction method, and this embodiment is an improvement of the first embodiment, and mainly improves the following: and when no finger touch is detected, updating the Gaussian mixture background model of each pixel point.
The specific flow of the fingerprint extraction method provided by this embodiment is shown in fig. 4.
Step 301 corresponds to substantially the same step 101, and step 303 to step 305 correspond to substantially the same step 102 to step 104, which are not described herein again; the difference is that step 302 is added in the present embodiment, which is specifically explained as follows:
In this embodiment, the gaussian background model of each pixel point may be updated periodically, for example, the gaussian background model of each pixel point may be updated every four hours in the daytime.
It should be noted that, the process of updating the gaussian background model of the pixel point is as follows: firstly, performing one or more times of learning and updating on a mixed Gaussian background model according to a basic pixel value of a currently collected pixel point; then, normalizing the weight values of a plurality of Gaussian components in the mixed Gaussian background model after one or more times of learning and updating; and finally, selecting a plurality of Gaussian components from the plurality of Gaussian components after normalization processing according to a preset selection rule to form a new Gaussian mixture background model. The specific implementation process is specifically described in the second embodiment, and is not described herein again.
Compared with the first embodiment, when no finger touch is detected, the gaussian background model of each pixel point is updated, and the gaussian background model of the pixel point is updated in time, so that the accuracy of the gaussian background model of the pixel point is ensured when the environment changes.
The steps of the methods in the first to third embodiments of the present invention are divided for clarity, and the implementation may be combined into one step or split some steps into multiple steps, and all of the steps are within the scope of the present patent as long as the same logical relationship is included; it is within the scope of the patent to add insignificant modifications to the algorithms or processes or to introduce insignificant design changes to the core design without changing the algorithms or processes.
A fourth embodiment of the present invention relates to a fingerprint extraction device applied to an electronic device, such as a mobile phone. As shown in fig. 5, the fingerprint extraction device includes: the device comprises a pixel value acquisition module 1, a grain characteristic identification module 2 and a fingerprint image generation module 3.
The pixel value acquisition module 1 is used for acquiring the sensing pixel value of each pixel point in the fingerprint sensing area when finger touch is detected;
the grain characteristic identification module 2 is used for identifying the corresponding grain characteristic of each pixel point according to the induction pixel value of each pixel point and a preset Gaussian mixture background model of each pixel point;
the fingerprint image generation module 3 is used for generating a fingerprint image of the finger according to the corresponding line characteristics of each pixel point.
It should be understood that this embodiment is an example of the apparatus corresponding to the first embodiment, and may be implemented in cooperation with the first embodiment. The related technical details mentioned in the first embodiment are still valid in this embodiment, and are not described herein again in order to reduce repetition. Accordingly, the related-art details mentioned in the present embodiment can also be applied to the first embodiment.
Compared with the prior art, the fingerprint extraction device provided by the embodiment has the advantages that the induction pixel value of each pixel point in the fingerprint induction area when the finger touches the fingerprint extraction device is obtained, the preset mixed Gaussian background model of each pixel point is obtained, then the corresponding line characteristics of each pixel point are identified, and the line characteristics are arranged according to the pixel points to generate the fingerprint image of the finger; the mixed Gaussian background model has a good description effect on unstable pixels, and the mixed Gaussian background model is used for extracting high-quality fingerprint images.
A fifth embodiment of the present invention relates to a fingerprint extraction device, and the present embodiment is a refinement of the fourth embodiment, and mainly includes: in the fifth embodiment of the present invention, a specific module included in the vein feature recognition module 2 is further described.
In this embodiment, as shown in fig. 6, the fingerprint feature recognition module 2 of the fingerprint extraction device includes a first matching unit 21 and a second matching unit 22.
The first matching unit 21 is configured to determine whether the induced pixel value of each determination pixel matches the gaussian background model of the pixel, and when the induced pixel value of the pixel matches the gaussian background model of the pixel, the first matching unit 21 identifies the texture feature of the pixel as a concave texture.
The second matching unit 22 is configured to determine whether the induced pixel value of the pixel meets a preset condition for matching the relief circuit when the first matching unit 21 determines that the induced pixel value of the pixel is not matched with the gaussian background model of the pixel. When the sensed pixel value of the pixel point satisfies the convex grain matching condition, the second matching unit 22 identifies the grain characteristics of the pixel point as a convex grain; when the sensed pixel value of the pixel does not satisfy the convex grain matching condition, the second matching unit 22 identifies the grain feature of the pixel as a concave grain.
In practice, the mixed gaussian background model includes a plurality of gaussian components arranged in sequence, and the first matching unit 21 is specifically configured to compare the induced pixel value of the pixel point with the plurality of gaussian components in sequence, and determine whether there is a gaussian component matching the induced pixel value of the pixel point; when there is a gaussian component matching the induced pixel value of the pixel, the first matching unit 21 determines that the induced pixel value of the pixel matches the gaussian background model of the pixel; when there is no gaussian component matching the induced pixel value of the pixel, the first matching unit 21 determines that the induced pixel value of the pixel is not matched with the gaussian background model of the pixel.
In practice, the first matching unit 21 includes: a calculating subunit and a judging subunit.
The calculating subunit is configured to calculate a difference between an induced pixel value of the pixel and a sample mean value in each gaussian component, and obtain a difference with a smallest absolute value as a matching parameter corresponding to the induced pixel value of the pixel. And the judging subunit is used for judging whether the matching parameter is smaller than or equal to a preset first threshold value.
When the matching parameter is smaller than or equal to the preset first threshold, the judging subunit judges that a Gaussian component matched with the sensing pixel value of the pixel point exists, and when the matching parameter is larger than the preset first threshold, the judging subunit judges that the Gaussian component matched with the sensing pixel value of the pixel point does not exist. The convex line matching conditions include: the matching parameter is larger than a preset second threshold value; wherein the second threshold is greater than the first threshold.
In the present embodiment, the first threshold value and the second threshold value are not limited at all and may be set as needed.
Since the second embodiment corresponds to the present embodiment, the present embodiment can be implemented in cooperation with the second embodiment. The related technical details mentioned in the second embodiment are still valid in this embodiment, and the technical effects that can be achieved in the second embodiment can also be achieved in this embodiment, and are not described herein again in order to reduce the repetition. Accordingly, the related-art details mentioned in the present embodiment can also be applied to the second embodiment.
The embodiment provides a specific composition of the texture feature recognition module, can complete the presetting of the Gaussian mixture background model of the pixel points and the recognition of the texture feature corresponding to each pixel point, and is a perfect explanation of the fourth embodiment to meet the actual design requirements.
A sixth embodiment of the present invention relates to a fingerprint extraction device, which is an improvement of the fourth embodiment, wherein: referring to fig. 7, the fingerprint extracting apparatus further includes a model presetting module 4.
As shown in fig. 7, the model presetting module 4 is configured to update the gaussian background model of each pixel point when no finger touch is detected.
In practice, as shown in fig. 8, the model presetting module 4 includes a creating unit 41, a learning updating unit 42, a normalization processing unit 43 and a selecting unit 44.
The creating unit 41 is configured to create a gaussian mixture model of the pixel point; the Gaussian mixture model comprises a plurality of Gaussian components which are sequentially arranged;
the learning updating unit 42 is configured to perform multiple learning updates on the gaussian mixture model according to the basic pixel values of the pixel points obtained multiple times; it should be noted that, the basic pixel value of the pixel point is obtained by the pixel value obtaining module 1 when no finger touches the pixel point;
the normalization processing unit 43 is configured to perform normalization processing on the weights of the multiple gaussian components in the updated gaussian mixture model after multiple learning;
and the selecting unit 44 is configured to select a plurality of gaussian components from the plurality of gaussian components after the normalization processing according to a preset selecting rule, so as to form a mixed gaussian background model.
In practice, the learning update unit 42 includes a matching subunit, a first update subunit, a sorting subunit, and a second update subunit.
The matching subunit is used for sequentially comparing the basic pixel value of the pixel point with a plurality of Gaussian components which are sequentially arranged, and judging whether a Gaussian component matched with the basic pixel value of the pixel point exists or not;
the first updating subunit is used for updating the weight of the Gaussian component according to a preset weight increment when the matching subunit judges that the Gaussian component matched with the basic pixel value of the pixel exists, and updating the sample mean and the sample variance of the Gaussian component according to the basic pixel value of the pixel;
the sorting subunit is used for re-sorting the Gaussian components according to a preset sorting rule;
the second updating subunit is used for deleting the Gaussian component arranged at the last in the Gaussian mixture model when the matching subunit judges that the Gaussian component matched with the basic pixel value of the pixel does not exist; adding a new Gaussian component in the Gaussian mixture model according to the basic pixel value of the pixel point; and updating the weight of the Gaussian components except the newly added Gaussian component in the mixed Gaussian model according to the preset weight decrement.
Since the third embodiment corresponds to the present embodiment, the present embodiment can be implemented in cooperation with the third embodiment. The related technical details mentioned in the third embodiment are still valid in this embodiment, and the technical effects that can be achieved in the third embodiment can also be achieved in this embodiment, and are not described herein again in order to reduce the repetition. Accordingly, the related-art details mentioned in the present embodiment can also be applied to the third embodiment.
Compared with the fourth embodiment, the fingerprint extraction device provided by the embodiment updates the Gaussian mixture background model of each pixel point when no finger touch is detected, and updates the Gaussian mixture background model of the pixel point in time so as to ensure the accuracy of the Gaussian mixture background model of the pixel point when the environment changes.
It should be noted that, all the modules related to the fourth to sixth embodiments of the present invention are logic modules, and in practical applications, one logic unit may be one physical unit, may also be a part of one physical unit, and may also be implemented by a combination of multiple physical units. In addition, in order to highlight the innovative part of the present invention, elements that are not so closely related to solving the technical problems proposed by the present invention are not introduced in the present embodiment, but this does not indicate that other elements are not present in the present embodiment.
Those skilled in the art can understand that all or part of the steps in the method of the foregoing embodiments may be implemented by a program to instruct related hardware, where the program is stored in a storage medium and includes several instructions to enable a device (which may be a single chip, a chip, etc.) or a processor (processor) to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
It will be understood by those of ordinary skill in the art that the foregoing embodiments are specific examples for carrying out the invention, and that various changes in form and details may be made therein without departing from the spirit and scope of the invention in practice.
Claims (16)
1. A fingerprint extraction method, comprising:
when finger touch is detected, acquiring an induction pixel value of each pixel point in a fingerprint induction area;
identifying a line characteristic corresponding to each pixel point according to the induction pixel value of each pixel point and a preset Gaussian mixture background model of each pixel point;
generating a fingerprint image of the finger according to the line characteristics corresponding to each pixel point;
the Gaussian mixture background model of each pixel point is an induction pixel value of each pixel point collected under the condition that a finger is not touched; after the fingerprint image of the finger is generated according to the texture features corresponding to each pixel point, the method further includes: and when no finger touch is detected, updating the Gaussian mixture background model of each pixel point.
2. The fingerprint extraction method according to claim 1, wherein the identifying, according to the sensed pixel value of each pixel point and a preset gaussian background model of each pixel point, the texture feature corresponding to each pixel point specifically comprises:
for each pixel point, judging whether the induction pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point;
and when the induction pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point, identifying the grain characteristics of the pixel point as concave grain lines.
3. The fingerprint extraction method according to claim 2, wherein in the identifying of the texture feature corresponding to each pixel point according to the sensed pixel value of each pixel point and a preset gaussian background model of each pixel point, the method further comprises:
when the induction pixel value of the pixel point is not matched with the Gaussian mixture background model of the pixel point, judging whether the induction pixel value of the pixel point meets a preset convex-pattern path matching condition or not;
when the induction pixel value of the pixel point meets the convex grain matching condition, identifying the grain characteristics of the pixel point as a convex grain;
and when the induction pixel value of the pixel point does not meet the convex grain matching condition, identifying the grain characteristics of the pixel point as the concave grain.
4. The fingerprint extraction method according to claim 3, wherein the Gaussian mixture background model comprises a plurality of Gaussian components arranged in sequence; the step of judging whether the induction pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point specifically comprises the following steps:
comparing the sensing pixel value of the pixel point with the plurality of Gaussian components in sequence, and judging whether a Gaussian component matched with the sensing pixel value of the pixel point exists or not;
when a Gaussian component matched with the induction pixel value of the pixel point exists, judging that the induction pixel value of the pixel point is matched with a Gaussian mixture background model of the pixel point; and when the Gaussian component matched with the induction pixel value of the pixel point does not exist, judging that the induction pixel value of the pixel point is not matched with the Gaussian mixture background model of the pixel point.
5. The fingerprint extraction method according to claim 4, wherein the sequentially comparing the sensing pixel value of the pixel point with the plurality of gaussian components to determine whether there is one gaussian component matching the sensing pixel value of the pixel point specifically comprises:
calculating the difference value between the induction pixel value of the pixel point and the sample mean value in each Gaussian component, and obtaining the difference value with the minimum absolute value as a matching parameter;
judging whether the matching parameter is smaller than or equal to a preset first threshold value;
when the matching parameter is smaller than or equal to a preset first threshold value, judging that a Gaussian component matched with the sensing pixel value of the pixel point exists; when the matching parameter is larger than the first threshold value, judging that no Gaussian component matched with the sensing pixel value of the pixel point exists;
the convex line matching conditions include: the matching parameter is larger than a preset second threshold value; the second threshold is greater than the first threshold.
6. The fingerprint extraction method according to claim 1, wherein the preset mode of the gaussian mixture background model of each pixel point specifically includes:
creating a Gaussian mixture model of the pixel points; the Gaussian mixture model comprises a plurality of Gaussian components which are sequentially arranged;
according to the basic pixel values of the pixel points acquired for multiple times, the Gaussian mixture model is subjected to multiple learning and updating; the basic pixel value of the pixel point is obtained when no finger touches the pixel point;
normalizing the weights of the Gaussian components in the Gaussian mixture model after multiple learning and updating;
and selecting a plurality of Gaussian components from the plurality of Gaussian components after normalization processing according to a preset selection rule to form the Gaussian mixture background model.
7. The fingerprint extraction method according to claim 6, wherein the learning update mode specifically comprises:
comparing the basic pixel value of the pixel point with the plurality of Gaussian components which are sequentially arranged, and judging whether a Gaussian component matched with the basic pixel value of the pixel point exists or not;
when a Gaussian component matched with the basic pixel value of the pixel point exists, updating the weight of the Gaussian component according to a preset weight increment, and updating the sample mean and the sample variance of the Gaussian component according to the basic pixel value of the pixel point;
and reordering the Gaussian components according to a preset ordering rule.
8. The fingerprint extraction method according to claim 7, further comprising, before said reordering said plurality of gaussian components according to a preset ordering rule:
when the Gaussian component matched with the basic pixel value of the pixel point does not exist, deleting the Gaussian component ranked at the last in the Gaussian mixture model;
adding a new Gaussian component in the Gaussian mixture model according to the basic pixel value of the pixel point;
and updating the weight of the Gaussian components except the newly added Gaussian component in the Gaussian mixture model according to a preset weight decrement.
9. A fingerprint extraction device, comprising:
the pixel value acquisition module is used for acquiring the sensing pixel value of each pixel point in the fingerprint sensing area when finger touch is detected;
the line characteristic identification module is used for identifying the line characteristic corresponding to each pixel point according to the induction pixel value of each pixel point and a preset Gaussian mixture background model of each pixel point;
the fingerprint image generation module is used for generating a fingerprint image of the finger according to the line characteristics corresponding to each pixel point;
and the model presetting module is used for updating the Gaussian mixture background model of each pixel point when no finger touch is detected.
10. The fingerprint extraction device of claim 9, wherein the texture feature recognition module comprises:
the first matching unit is used for judging whether the induction pixel value of each judged pixel point is matched with the Gaussian mixture background model of the pixel point or not;
when the induction pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point, the first matching unit identifies the grain characteristics of the pixel point as concave grain lines.
11. The fingerprint extraction device of claim 10, wherein the texture feature recognition module further comprises:
the second matching unit is used for judging whether the induction pixel value of the pixel point meets a preset convex road matching condition or not when the first matching unit judges that the induction pixel value of the pixel point is not matched with the mixed Gaussian background model of the pixel point;
when the induction pixel value of the pixel point meets the convex grain matching condition, the second matching unit identifies the grain characteristics of the pixel point as a convex grain path; and when the induction pixel value of the pixel point does not meet the convex grain matching condition, the second matching unit identifies the grain characteristics of the pixel point as the concave grain.
12. The fingerprint extraction apparatus according to claim 11, wherein the gaussian mixture background model comprises several gaussian components arranged in sequence;
the first matching unit is specifically configured to sequentially compare the sensing pixel values of the pixel points with the plurality of gaussian components, and determine whether there is one gaussian component matching the sensing pixel values of the pixel points;
when a Gaussian component matched with the induction pixel value of the pixel point exists, the first matching unit judges that the induction pixel value of the pixel point is matched with the Gaussian mixture background model of the pixel point; when the Gaussian component matched with the induction pixel value of the pixel point does not exist, the first matching unit judges that the induction pixel value of the pixel point is not matched with the Gaussian mixture background model of the pixel point.
13. The fingerprint extraction device according to claim 12, wherein the first matching unit includes:
the calculating subunit is used for calculating a difference value between the sensing pixel value of the pixel point and the sample mean value in each Gaussian component, and acquiring a difference value with the smallest absolute value as a matching parameter corresponding to the sensing pixel value of the pixel point;
the judging subunit is used for judging whether the matching parameter is smaller than or equal to a preset first threshold value;
when the matching parameter is smaller than or equal to a preset first threshold value, the judging subunit judges that a Gaussian component matched with the sensing pixel value of the pixel point exists; when the matching parameter is larger than a preset first threshold value, the judging subunit judges that no Gaussian component matched with the sensing pixel value of the pixel point exists;
the convex line matching conditions include: the sensing pixel value of the pixel point is larger than a preset second threshold value; the second threshold is greater than the first threshold.
14. The fingerprint extraction apparatus according to claim 9, wherein the model presetting module comprises:
the creating unit is used for creating a Gaussian mixture model of the pixel points; the Gaussian mixture model comprises a plurality of Gaussian components which are sequentially arranged;
the learning updating unit is used for performing multiple learning updating on the Gaussian mixture model according to the basic pixel values of the pixel points obtained multiple times; the basic pixel value of the pixel point is obtained through a pixel value obtaining module when no finger touches the pixel point;
the normalization processing unit is used for performing normalization processing on the weights of the Gaussian components in the Gaussian mixture model after multiple learning and updating;
and the selecting unit is used for selecting a plurality of Gaussian components from the plurality of Gaussian components after normalization processing according to a preset selecting rule so as to form the Gaussian mixture background model.
15. The fingerprint extraction device according to claim 14, wherein the learning update unit specifically includes:
the matching subunit is used for sequentially comparing the basic pixel value of the pixel point with the plurality of Gaussian components which are sequentially arranged and judging whether a Gaussian component matched with the basic pixel value of the pixel point exists or not;
the first updating subunit is used for updating the weight of the Gaussian component according to a preset weight increment when the matching subunit judges that the Gaussian component matched with the basic pixel value of the pixel exists, and updating the sample mean and the sample variance of the Gaussian component according to the basic pixel value of the pixel;
and the sequencing subunit is used for reordering the Gaussian components according to a preset sequencing rule.
16. The fingerprint extraction device according to claim 15, wherein the learning update unit further includes:
the second updating subunit is used for deleting the Gaussian component arranged at the last in the Gaussian mixture model when the matching subunit judges that the Gaussian component matched with the basic pixel value of the pixel does not exist; adding a new Gaussian component in the Gaussian mixture model according to the basic pixel value of the pixel point; and updating the weight of the Gaussian components except the newly added Gaussian component in the Gaussian mixture model according to a preset weight decrement.
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/CN2017/072712 WO2018137226A1 (en) | 2017-01-25 | 2017-01-25 | Fingerprint extraction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107077617A CN107077617A (en) | 2017-08-18 |
| CN107077617B true CN107077617B (en) | 2020-08-28 |
Family
ID=59613826
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201780000029.1A Active CN107077617B (en) | 2017-01-25 | 2017-01-25 | Fingerprint extraction method and device |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN107077617B (en) |
| WO (1) | WO2018137226A1 (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107831468A (en) * | 2017-10-19 | 2018-03-23 | 天津大学 | A kind of fingerprinting localization algorithm based on affine propagation clustering |
| CN108124483B (en) * | 2017-12-01 | 2022-06-07 | 深圳市汇顶科技股份有限公司 | Fingerprint image enhancement method and fingerprint image module |
| CN113537196B (en) * | 2021-07-21 | 2023-04-07 | 拉扎斯网络科技（上海）有限公司 | Picture identification method and device, computer equipment and computer readable storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101686338A (en) * | 2008-09-26 | 2010-03-31 | 索尼株式会社 | System and method for partitioning foreground and background in video |
| CN102184404A (en) * | 2011-04-29 | 2011-09-14 | 汉王科技股份有限公司 | Method and device for acquiring palm region in palm image |
| CN102393905A (en) * | 2011-07-13 | 2012-03-28 | 哈尔滨工程大学 | Extraction method of venous mode texture on back of hand |
| CN104866616A (en) * | 2015-06-07 | 2015-08-26 | 中科院成都信息技术股份有限公司 | Method for searching monitor video target |
| CN105844265A (en) * | 2016-06-07 | 2016-08-10 | 广东欧珀移动通信有限公司 | Fingerprint image processing method and device |
| CN105894701A (en) * | 2016-04-05 | 2016-08-24 | 江苏电力信息技术有限公司 | Large construction vehicle identification and alarm method for preventing external damage to transmission lines |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8325993B2 (en) * | 2009-12-23 | 2012-12-04 | Lockheed Martin Corporation | Standoff and mobile fingerprint collection |
| CN104463132B (en) * | 2014-12-17 | 2018-09-11 | 浙江维尔科技股份有限公司 | A kind of fingerprint identification method and device |
| CN105095867A (en) * | 2015-07-21 | 2015-11-25 | 哈尔滨多智科技发展有限公司 | Rapid dynamic face extraction and identification method based deep learning |
-
2017
- 2017-01-25 CN CN201780000029.1A patent/CN107077617B/en active Active
- 2017-01-25 WO PCT/CN2017/072712 patent/WO2018137226A1/en active Application Filing
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101686338A (en) * | 2008-09-26 | 2010-03-31 | 索尼株式会社 | System and method for partitioning foreground and background in video |
| CN102184404A (en) * | 2011-04-29 | 2011-09-14 | 汉王科技股份有限公司 | Method and device for acquiring palm region in palm image |
| CN102393905A (en) * | 2011-07-13 | 2012-03-28 | 哈尔滨工程大学 | Extraction method of venous mode texture on back of hand |
| CN104866616A (en) * | 2015-06-07 | 2015-08-26 | 中科院成都信息技术股份有限公司 | Method for searching monitor video target |
| CN105894701A (en) * | 2016-04-05 | 2016-08-24 | 江苏电力信息技术有限公司 | Large construction vehicle identification and alarm method for preventing external damage to transmission lines |
| CN105844265A (en) * | 2016-06-07 | 2016-08-10 | 广东欧珀移动通信有限公司 | Fingerprint image processing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107077617A (en) | 2017-08-18 |
| WO2018137226A1 (en) | 2018-08-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110163300B (en) | Image classification method and device, electronic equipment and storage medium | |
| CN103313018A (en) | Registration determination device, control method thereof, and electronic apparatus | |
| CN110717881A (en) | Wafer defect identification method and device, storage medium and terminal equipment | |
| CN108460346B (en) | Fingerprint identification method and device | |
| US9842279B2 (en) | Data processing method for learning discriminator, and data processing apparatus therefor | |
| CN111881289B (en) | Training method of classification model, and detection method and device of data risk class | |
| CN107077617B (en) | Fingerprint extraction method and device | |
| US20200257885A1 (en) | High speed reference point independent database filtering for fingerprint identification | |
| CN107958230A (en) | Facial expression recognizing method and device | |
| CN113408634A (en) | Model recommendation method and device, equipment and computer storage medium | |
| CN112927061A (en) | User operation detection method and program product | |
| CN111382791B (en) | Deep learning task processing method, image recognition task processing method and device | |
| CN111445021B (en) | Learning method, learning apparatus, and computer-readable recording medium | |
| CN110135428B (en) | Image segmentation processing method and device | |
| CN111784665A (en) | OCT image quality assessment method, system and device based on Fourier transform | |
| CN111783812A (en) | Method and device for identifying forbidden images and computer readable storage medium | |
| CN111222558A (en) | Image processing method and storage medium | |
| CN111428064A (en) | Small-area fingerprint image fast indexing method, device, equipment and storage medium | |
| CN108345943B (en) | Machine learning identification method based on embedded coding and contrast learning | |
| CN112633369B (en) | Image matching method and device, electronic equipment and computer-readable storage medium | |
| CN115840831A (en) | Commodity retrieval method, device, equipment and storage medium | |
| CN113569070A (en) | Image detection method and device, electronic equipment and storage medium | |
| CN113269176B (en) | Image processing model training method, image processing device and computer equipment | |
| CN112446428B (en) | Image data processing method and device | |
| CN110309127B (en) | Data processing method and device and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |