CN105718918B - Aerial signature authentication method based on video - Google Patents
Aerial signature authentication method based on video Download PDFInfo
- Publication number
- CN105718918B CN105718918B CN201610059700.7A CN201610059700A CN105718918B CN 105718918 B CN105718918 B CN 105718918B CN 201610059700 A CN201610059700 A CN 201610059700A CN 105718918 B CN105718918 B CN 105718918B
- Authority
- CN
- China
- Prior art keywords
- signature
- registration
- track
- authentication
- pursuit path
- 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
Classifications
-
- 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
- G06V40/30—Writer recognition; Reading and verifying signatures
-
- 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
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/28—Recognition of hand or arm movements, e.g. recognition of deaf sign language
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Collating Specific Patterns (AREA)
Abstract
The present invention provides a kind of aerial signature authentication method based on video, and this method carries out feature extraction to the signature for the signature and registration that need to be authenticated, and respectively obtains the track characteristic of the track characteristic for needing authentication signature and registration signature;The certification to realize signature is matched to the track characteristic of the track characteristic and registration signature that need authentication signature again.The present invention is based on the aerial signature authentication methods of video can effectively improve the accuracy of identification of signature, to improve the accuracy rate and validity of signature authentication.
Description
Technical field
The present invention relates to image processing and analysis technical fields, more specifically to a kind of aerial label based on video
Name authentication method.
Background technique
In recent years, with the fast development of biometrics identification technology and extensive use, user is to biological characteristic authentication
Technology proposes more higher requirements, so that there are some new development trends in biological characteristic authentication, it is such as stronger anti-fake
Property, higher accuracy of identification and higher user's acceptance.
Although on-line signature certification has been rapidly developed and applied, there are also deficiencies, are mainly reflected in following
Several aspects:
(1) stroke and number of words signed are limited, therefore are difficult to accurately hold when choosing has representational feature;
(2) in different times, changed by different writing implements and psychology physiological situation in the case where all
It may cause the inconsistent of signature front and back, therefore compared with other biological characteristics such as fingerprint, stability is poor, and accuracy rate is inclined
It is low;
(3) the form domination signed takes and forges so that some static natures of signature are easily stolen, therefore antifalsification
It is affected;
(4) condition is depended on but also it has that application is partially narrow based on board.
Above-mentioned these problems are all that on-line signature certification faces improved place.
Summary of the invention
It is an object of the invention to overcome shortcoming and deficiency in the prior art, a kind of aerial signature based on video is provided
Authentication method, this method can effectively improve the accuracy of identification of signature, to improve the accuracy rate and validity of signature authentication.
In order to achieve the above object, the technical scheme is that: a kind of aerial label based on video
Name authentication method, it is characterised in that: feature extraction is carried out to the signature for the signature and registration that need to be authenticated, label need to be authenticated by respectively obtaining
The track characteristic of name and the track characteristic of registration signature;Again to the track characteristic of the track characteristic and registration signature that need authentication signature
Matched the certification to realize signature;
Wherein, it carries out feature extraction to the signature of registration to refer to: by the signature video of acquisition registration, and to infusing in video
The finger tip of volume carries out detection and tracking, to obtain the signature track generated when finger is write in the sky;Then to the signature of registration
Track is pre-processed, and extracts the signature track characteristic of registration;
It carries out feature extraction to the signature that need to authenticate to refer to: by acquiring the signature video that need to be authenticated, and to needing in video
The finger tip of certification carries out detection and tracking, to obtain the signature track generated when finger is write in the sky;Then to need to authenticate
Signature track is pre-processed, and extracts the signature track characteristic that need to be authenticated.
In the above scheme, the signature of registration and the signature that need to be authenticated are all made of identical processing method, then and certifiable
The consistency matched, to effectively improve the accuracy of identification of signature.
To the finger tip registered in video carry out detection and tracking and to needed in video the finger tip of certification carry out detection and with
Track is all made of improved TLD algorithm, respectively obtains the pursuit path of registration signature and needs the pursuit path of authentication signature.
The improved TLD algorithm the following steps are included:
The first step reads video file or directly reads camera real-time pictures, outline three finger tip kneading portions, completes mesh
Mark frame initialization;
Second step tracks finger tip target using TLD track algorithm is improved: time contextual information being added during tracking,
It is adaptively adjusted the size of region of search according to the ratio between image size and target frame size, and uses TLD in region of search
Algorithm detection and tracking finger tip target;
Third step sets frame number threshold value T3, add up the frame number D that finger tip target is not detected in region of search, by frame number D
With frame number threshold value T3Compare to judge whether tracking fails: if D > T3, then failure is tracked, sounds a warning and returns to the first step again
Tracking;Otherwise, the track for saving signature respectively obtains the pursuit path of registration signature and needs the pursuit path of authentication signature.
As described above, improved TLD algorithm is the TLD track algorithm based on time context, suitably reduce detection,
Following range come avoid it is unnecessary detection and part background influence, and then improve algorithm tracking velocity and algorithm standard
True rate.Many experiments are also shown, and by improved TLD algorithm, big in intensity of illumination, environment is simple, the feelings that finger tip quickly moves
Under condition, the preferable effect all taken when inclined for the upper and lower, left and right of finger tip has strong robustness.
The signature track of registration is pre-processed the following steps are included:
The first step, setting speed threshold value T1, calculate in the pursuit path of registration signature and obtain the speed V of head and the tail tracing point1,
By speed V1With threshold speed T1Compare to judge whether the tracing point is wrong tracing point, to obtain the tracking rail of registration signature
Mark one: if V1<T1, then the tracing point is judged for wrong tracing point and is deleted;Otherwise retain the tracing point;
Second step, set distance threshold value T2, calculate the distance L of every two tracing point in the pursuit path one of registration signature1,
It will distance L1With distance threshold T2Compare to determine whether tracking exception or loss of data: if L1> T2, then judge registration signature
Pursuit path one tracks exception or loss of data, and the result of front and back two o'clock interpolation arithmetic is substituted the erroneous point, obtains registration label
The pursuit path two of name;Otherwise, judge that the pursuit path one of registration signature tracks the pursuit path for normally obtaining registration signature
Two;
Third step carries out dimension normalization to the position coordinates of the pursuit path two of registration signature, obtains registration signature
Track characteristic.
The signature track that need to be authenticated is pre-processed the following steps are included:
The first step, setting speed threshold value T1, calculate in the pursuit path for needing authentication signature and obtain the speed of head and the tail tracing point
V2, by speed V2With threshold speed T1Compare to judge whether the tracing point is wrong tracing point, come obtain needing authentication signature with
Track track one: if V2<T1, then the tracing point is judged for wrong tracing point and is deleted;Otherwise retain the tracing point;
Second step, set distance threshold value T2, calculate the distance of every two tracing point in the pursuit path one for needing authentication signature
L2, will distance L2With distance threshold T2Compare to determine whether tracking exception or loss of data: if L2>T2, then judge that label need to be authenticated
The pursuit path one of name tracks exception or loss of data, and the result of front and back two o'clock interpolation arithmetic is substituted the erroneous point, is needed
The pursuit path two of authentication signature;Otherwise, judgement needs the pursuit path one of authentication signature to track normally, obtains needing authentication signature
Pursuit path two;
Third step carries out dimension normalization to the position coordinates for the pursuit path two for needing authentication signature, obtains that label need to be authenticated
The track characteristic of name.
In actually signature, the problems such as due to improved TLD algorithm and system reaction time, if before rigid beginning and end
A dry track data point is largely overlapped and shakes, and seriously affects recognition effect, therefore use speed threshold by the above-mentioned first step
Value T1The data point of these mistakes is removed, tracing point and fine jitter tracing point is overlapped to remove head and the tail, head and the tail tracing point is obtained
The speed arrived is less than threshold speed T1Tracing point removal, thus obtain registration signature pursuit path one and need authentication signature
Pursuit path one.
When track algorithm is run, it is possible to because of the reasons such as hardware and video frame losing, lead to partial data point loss or different
Often, error in data is caused.Second step of the present invention uses distance threshold algorithm, calculates the distance between every two tracing point, setting
One biggish threshold value judges the erroneous point being likely to occur and is corrected to obtain two He of pursuit path of registration signature with interpolation algorithm
Need the pursuit path two of authentication signature.
In view of the starting of signature and extreme point are different, recognition effect is influenced, is needed through third step to registration signature
The position coordinates of pursuit path two are normalized and the position coordinates for the pursuit path two for needing authentication signature are normalized,
Signature size will be registered and need authentication signature scaled as identical size, to eliminate size shadow caused by signatures match precision
It rings.
It is to the certification for needing track characteristic of the track characteristic of authentication signature with registration signature to be matched to realize signature
Refer to: by the combination of dynamic time warping and fast fourier transform algorithm, come to the track characteristic for needing authentication signature and
The track characteristic of registration signature is matched, and realizes the certification of signature.
Motion processes different from on-line signature, that new authentication method proposed by the present invention is signed in finger tip pinch chalaza
In, without based on utilizable pressure in board signature process and the information such as pen state of rising and falling, but contain longer
Handwriting trace and other characteristic informations such as different presentation direction, therefore contain more abundant characteristic information.For
These features, the present invention is based on dynamic time programming (DTW) algorithm and Fast Fourier Transform (FFT) (FFT) algorithm to aerial label
Name is authenticated.
The combination by dynamic time warping and fast fourier transform algorithm, come to the rail for needing authentication signature
The track characteristic of mark feature and registration signature is matched, and realizes the certification of signature, comprising the following steps:
The first step chooses the track characteristic of N number of registration signature and calculates average length, according to the track of N number of registration signature
The average length of feature carries out interpolation arithmetic to the track characteristic of N number of registration signature, obtains mean trajectory feature as matching mould
Plate;Wherein, 2 < N≤registration total sample number;
Second step calculates separately the track characteristic of N number of registration signature and the DTW distance and FFT distance of matching template, and
The average value one of DTW distance and the average value two and variance two of variance one and FFT distance is calculated;Then according to average
Value one, variance one, average value two and variance two establish joint gaussian probability model;
Third step calculates the DTW distance and FFT distance of any track characteristic for needing authentication signature and matching template, then
It substitutes into joint gaussian probability model and calculates matching probability;Matching probability is greater than to the track characteristic for needing authentication signature of given threshold
It is determined as actual signature, is otherwise judged to forging a signature.
Wherein, DTW distance refers to the similarity between the coordinates for two signatures being calculated using DTW algorithm, FFT away from
From the co sinus vector included angle value referred to 32 dimension Fourier coefficients before being obtained after two X, Y coordinates signed progress FFT.
Signature Authentication System, which generally requires to sign by registration, obtains matching template, because the length of signature is often different,
It needs to carry out interpolation arithmetic and obtains the coordinate sequence of equal length, one matching template of average acquisition is carried out to each coordinate points.
The template obtained by this average calculating operation can avoid some registration signature from signing with other registrations well, and gap is excessive to be caused
Adverse effect.
Because the length of the signature and matching template that need to authenticate also tends to difference, the distance of two signatures cannot be carried out directly
It calculates.DTW algorithm can find the corresponding points that an optimal trajectory matches the different sequence of two length, calculate summation corresponding points
Euclidean distance as DTW distance.Time-domain signal can be converted to frequency-region signal by Fourier transformation, take preceding 32 coefficients can be with
The co sinus vector included angle of the former signature of expression well, 32 dimensional feature vectors of two signatures can be used as the distance metric of FFT.
DTW and fft algorithm obtain mutually over time and space respectively from the distance between two angle calculation signatures of time domain and frequency domain
Supplement is to improve algorithm performance.
Compared with prior art, the invention has the advantages that with the utility model has the advantages that the present invention is based on the aerial signatures of video
Authentication method can effectively improve the accuracy of identification of signature, to improve the accuracy rate and validity of signature authentication.
Detailed description of the invention
Fig. 1 is the flow diagram of the aerial signature authentication method the present invention is based on video;
Fig. 2 (1)-Fig. 2 (8) is the effect picture tracked after improved TLD algorithm;
Fig. 3 is the trajectory diagram of finger tip tracking result;
Fig. 4 is the trajectory diagram of tape error tracing point;
Specific embodiment
The present invention is described in further detail with specific embodiment with reference to the accompanying drawing.
Embodiment
As shown in Figure 1, a kind of aerial signature authentication method based on video of the present invention, is to the signature and registration that need to be authenticated
Signature carry out feature extraction, respectively obtain need authentication signature track characteristic and registration signature track characteristic;Again to need to recognize
The track characteristic of signed certificate name and the track characteristic of registration signature are matched the certification to realize signature;
Wherein, it carries out feature extraction to the signature of registration to refer to: by the signature video of acquisition registration, and to infusing in video
The finger tip of volume carries out detection and tracking, to obtain the signature track generated when finger is write in the sky;Then to the signature of registration
Track is pre-processed, and extracts the signature track characteristic of registration;
It carries out feature extraction to the signature that need to authenticate to refer to: by acquiring the signature video that need to be authenticated, and to needing in video
The finger tip of certification carries out detection and tracking, to obtain the signature track generated when finger is write in the sky;Then to need to authenticate
Signature track is pre-processed, and extracts the signature track characteristic that need to be authenticated.
It is above-mentioned that detection and tracking is carried out to the finger tip registered in video and the finger tip for needing certification in video is detected
And tracking, it is all made of improved TLD algorithm, the pursuit path of registration signature is respectively obtained and needs the pursuit path of authentication signature.
Wherein, improved TLD algorithm the following steps are included:
The first step reads video file or directly reads camera real-time pictures, outline three finger tip kneading portions, completes mesh
Mark frame initialization;
Second step tracks finger tip target using TLD track algorithm is improved: time contextual information being added during tracking,
It is adaptively adjusted the size of region of search according to the ratio between image size and target frame size, and uses TLD in region of search
Algorithm detection and tracking finger tip target;
Third step sets frame number threshold value T3, add up the frame number D that finger tip target is not detected in region of search, by frame number D
With frame number threshold value T3Compare to judge whether tracking fails: if D > T3, then failure is tracked, sounds a warning and returns to the first step again
Tracking;Otherwise, the track for saving signature respectively obtains the pursuit path of registration signature and needs the pursuit path of authentication signature.
Improved TLD algorithm is the TLD track algorithm based on time context, suitably reduces detection, following range
To avoid unnecessary detection and the influence of part background, and then the accuracy rate of the tracking velocity of raising algorithm and algorithm.This
Invention is shown using effect picture such as Fig. 2 (the 1)-Fig. 2 (2) tracked after improved TLD algorithm, and the track of finger tip tracking result
Figure is as shown in Figure 3.Many experiments are also shown, and by improved TLD algorithm, big in intensity of illumination, environment is simple, and finger tip is quickly transported
In the case where dynamic, the preferable effect all taken when inclined for the upper and lower, left and right of finger tip has strong robustness.
The present invention signature track of registration is pre-processed the following steps are included:
The first step, setting speed threshold value T1, calculate in the pursuit path of registration signature and obtain the speed V of head and the tail tracing point1,
By speed V1With threshold speed T1Compare to judge whether the tracing point is wrong tracing point, to obtain the tracking rail of registration signature
Mark one: if V1<T1, then the tracing point is judged for wrong tracing point and is deleted;Otherwise retain the tracing point;
Second step, set distance threshold value T2, calculate the distance L of every two tracing point in the pursuit path one of registration signature1,
It will distance L1With distance threshold T2Compare to determine whether tracking exception or loss of data: if L1> T2, then judge registration signature
Pursuit path one tracks exception or loss of data, and the result of front and back two o'clock interpolation arithmetic is substituted the erroneous point, obtains registration label
The pursuit path two of name;Otherwise, judge that the pursuit path one of registration signature tracks the pursuit path for normally obtaining registration signature
Two;
Third step carries out dimension normalization to the position coordinates of the pursuit path two of registration signature, obtains registration signature
Track characteristic.
The signature track that need to be authenticated is pre-processed the following steps are included:
The first step, setting speed threshold value T1, calculate in the pursuit path for needing authentication signature and obtain the speed of head and the tail tracing point
V2, by speed V2With threshold speed T1Compare to judge whether the tracing point is wrong tracing point, come obtain needing authentication signature with
Track track one: if V2<T1, then the tracing point is judged for wrong tracing point and is deleted;Otherwise retain the tracing point;
Second step, set distance threshold value T2, calculate the distance of every two tracing point in the pursuit path one for needing authentication signature
L2, will distance L2With distance threshold T2Compare to determine whether tracking exception or loss of data: if L2>T2, then judge that label need to be authenticated
The pursuit path one of name tracks exception or loss of data, and the result of front and back two o'clock interpolation arithmetic is substituted the erroneous point, is needed
The pursuit path two of authentication signature;Otherwise, judgement needs the pursuit path one of authentication signature to track normally, obtains needing authentication signature
Pursuit path two;
Third step carries out dimension normalization to the position coordinates for the pursuit path two for needing authentication signature, obtains that label need to be authenticated
The track characteristic of name.
In actually signature, the problems such as due to improved TLD algorithm and system reaction time, if before rigid beginning and end
A dry track data point is largely overlapped and shakes, and seriously affects recognition effect (as shown in figure 4, the tracing point with overlapping in figure
Or the tracing point of fine jitter), therefore threshold speed T is used by the above-mentioned first step1The data point of these mistakes is removed, is come
Removal head and the tail are overlapped tracing point and fine jitter tracing point, and the speed that head and the tail tracing point is obtained is less than threshold speed T1Track
Point removal, to obtain the pursuit path one of registration signature and need the pursuit path one of authentication signature.
When track algorithm is run, it is possible to because of the reasons such as hardware and video frame losing, lead to partial data point loss or different
Often, error in data is caused.Second step of the present invention uses distance threshold algorithm, calculates the distance between every two tracing point, setting
One biggish threshold value judges the erroneous point being likely to occur and is corrected to obtain two He of pursuit path of registration signature with interpolation algorithm
Need the pursuit path two of authentication signature.
In view of the starting of signature and extreme point are different, recognition effect is influenced, is needed through third step to registration signature
The position coordinates of pursuit path two are normalized and the position coordinates for the pursuit path two for needing authentication signature are normalized,
Signature size will be registered and need authentication signature scaled as identical size, to eliminate size shadow caused by signatures match precision
It rings.
The present invention matches the track characteristic of the track characteristic and registration signature that need authentication signature to realize signature
Certification refers to: by the combination of dynamic time warping and fast fourier transform algorithm, come to the track for needing authentication signature
The track characteristic of feature and registration signature is matched, and realizes the certification of signature.
Motion processes different from on-line signature, that new authentication method proposed by the present invention is signed in finger tip pinch chalaza
In, without based on utilizable pressure in board signature process and the information such as pen state of rising and falling, but contain longer
Handwriting trace and other characteristic informations such as different presentation direction, therefore contain more abundant characteristic information.For
These features, the present invention is based on dynamic time programming (DTW) algorithm and Fast Fourier Transform (FFT) (FFT) algorithm to aerial label
Name is authenticated.
Above by the combination of dynamic time warping and fast fourier transform algorithm, come to the rail for needing authentication signature
The track characteristic of mark feature and registration signature is matched, and realizes the certification of signature, comprising the following steps:
The first step chooses the track characteristic of five registration signatures and calculates average length, according to the rail of five registration signatures
The average length of mark feature carries out interpolation arithmetic to the track characteristic of five registration signatures, obtains mean trajectory feature as matching
Template;
Second step calculates separately the track characteristic of five registration signatures and the DTW distance and FFT distance of matching template, and
The average value one of DTW distance and the average value two and variance two of variance one and FFT distance is calculated;Then according to average
Value one, variance one, average value two and variance two establish joint gaussian probability model;
Third step calculates the DTW distance and FFT distance of any track characteristic for needing authentication signature and matching template, then
It substitutes into joint gaussian probability model and calculates matching probability;The track characteristic for needing authentication signature by matching probability greater than 50% determines
For actual signature, otherwise it is judged to forging a signature.
Wherein, DTW distance refers to the similarity between the coordinates for two signatures being calculated using DTW algorithm, FFT away from
From the co sinus vector included angle value referred to 32 dimension Fourier coefficients before being obtained after two X, Y coordinates signed progress FFT.
Signature Authentication System, which generally requires to sign by registration, obtains matching template, because the length of signature is often different,
It needs to carry out interpolation arithmetic and obtains the coordinate sequence of equal length, one matching template of average acquisition is carried out to each coordinate points.
The template obtained by this average calculating operation can avoid some registration signature from signing with other registrations well, and gap is excessive to be caused
Adverse effect.
Because the length of the signature and matching template that need to authenticate also tends to difference, the distance of two signatures cannot be carried out directly
It calculates.DTW algorithm can find the corresponding points that an optimal trajectory matches the different sequence of two length, calculate summation corresponding points
Euclidean distance as DTW distance.Time-domain signal can be converted to frequency-region signal by Fourier transformation, take preceding 32 coefficients can be with
The co sinus vector included angle of the former signature of expression well, 32 dimensional feature vectors of two signatures can be used as the distance metric of FFT.
DTW and fft algorithm obtain mutually over time and space respectively from the distance between two angle calculation signatures of time domain and frequency domain
Supplement is to improve algorithm performance.
The present invention is based on the aerial signature authentication methods of video can be applied to mobile phone, palm handheld terminal (such as palm electricity
Brain), the processor with video camera and the computer with video camera etc..It applies when on mobile phone, mobile phone can be passed through
Camera acquires the signature video of registration and the signature video that need to be authenticated, and user can recognize using installation present invention signature on mobile phone
The APP software of card method realizes the certification of signature.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (5)
1. a kind of aerial signature authentication method based on video, it is characterised in that: to the signature of the signature and registration that need to authenticate into
Row feature extraction respectively obtains the track characteristic of the track characteristic for needing authentication signature and registration signature;Again to needing authentication signature
The track characteristic of track characteristic and registration signature is matched the certification to realize signature;
Wherein, it carries out feature extraction to the signature of registration referring to: by the signature video of acquisition registration, and to registering in video
Finger tip carries out detection and tracking, to obtain the signature track generated when finger is write in the sky;Then to the signature track of registration
It is pre-processed, and extracts the signature track characteristic of registration;
It carries out feature extraction to the signature that need to authenticate to refer to: by acquiring the signature video that need to be authenticated, and to needing to authenticate in video
Finger tip carry out detection and tracking, to obtain the signature track generated when finger is write in the sky;Then to the signature that need to be authenticated
Track is pre-processed, and extracts the signature track characteristic that need to be authenticated;
The track characteristic of the track characteristic and registration signature that need authentication signature is matched to realize that the certification of signature refers to: logical
The combination of dynamic time warping and fast fourier transform algorithm is crossed, to sign to the track characteristic and registration that need authentication signature
The track characteristic of name is matched, and realizes the certification of signature;
The combination by dynamic time warping and fast fourier transform algorithm, come special to the track for needing authentication signature
The track characteristic that registration is signed of seeking peace is matched, and realizes the certification of signature, comprising the following steps:
The first step chooses the track characteristic of N number of registration signature and calculates average length, according to the track characteristic of N number of registration signature
Average length to it is N number of registration signature track characteristic carry out interpolation arithmetic, obtain mean trajectory feature as matching template;Its
In, 2 < N < registration total sample number;
Second step calculates separately the track characteristic of N number of registration signature and the DTW distance and FFT distance of matching template, and calculates
To the average value one and variance one of DTW distance and the average value two and variance two of FFT distance;Then according to average value one, side
Poor one, average value two and variance two establish joint gaussian probability model;
Third step calculates the DTW distance and FFT distance of any track characteristic for needing authentication signature and matching template, then substitutes into
Joint gaussian probability model calculates matching probability;The track characteristic for needing authentication signature that matching probability is greater than given threshold is determined
For actual signature, otherwise it is judged to forging a signature.
2. the aerial signature authentication method according to claim 1 based on video, it is characterised in that: to what is registered in video
Finger tip carries out detection and tracking and carries out detection and tracking to the finger tip for needing certification in video, is all made of improved TLD algorithm,
It respectively obtains the pursuit path of registration signature and needs the pursuit path of authentication signature.
3. the aerial signature authentication method according to claim 2 based on video, it is characterised in that: the improved TLD
Algorithm the following steps are included:
The first step reads video file or directly reads camera real-time pictures, outline three finger tip kneading portions, completes target frame
Initialization;
Second step tracks finger tip target using TLD track algorithm is improved: time contextual information being added during tracking, according to
The ratio between image size and target frame size are adaptively adjusted the size of region of search, and TLD algorithm is used in region of search
Detection and tracking finger tip target;
Third step sets frame number threshold value T3, add up the frame number D that finger tip target is not detected in region of search, by frame number D and frame
Number threshold value T3Compare to judge whether tracking fails: if D > T3, then track failure, sound a warning and return the first step again with
Track;Otherwise, the track for saving signature respectively obtains the pursuit path of registration signature and needs the pursuit path of authentication signature.
4. the aerial signature authentication method according to claim 3 based on video, it is characterised in that: to the signature rail of registration
Mark pre-processed the following steps are included:
The first step, setting speed threshold value T1, calculate in the pursuit path of registration signature and obtain the speed V of head and the tail tracing point1, will be fast
Spend V1With threshold speed T1Compare to judge whether the tracing point is wrong tracing point, to obtain the pursuit path one of registration signature:
If V1<T1, then the tracing point is judged for wrong tracing point and is deleted;Otherwise retain the tracing point;
Second step, set distance threshold value T2, calculate the distance L of every two tracing point in the pursuit path one of registration signature1, will be away from
From L1With distance threshold T2Compare to determine whether tracking exception or loss of data: if L1>T2, then judge the tracking of registration signature
Track one tracks exception or loss of data, and the result of front and back two o'clock interpolation arithmetic is substituted the erroneous point, obtains registration signature
Pursuit path two;Otherwise, judge that the pursuit path one of registration signature tracks the pursuit path two for normally obtaining registration signature;
Third step carries out dimension normalization to the position coordinates of the pursuit path two of registration signature, obtains the track of registration signature
Feature.
5. the aerial signature authentication method according to claim 3 based on video, it is characterised in that: to the signature that need to be authenticated
Track pre-processed the following steps are included:
The first step, setting speed threshold value T1, calculate in the pursuit path for needing authentication signature and obtain the speed V of head and the tail tracing point2, will
Speed V2With threshold speed T1Compare to judge whether the tracing point is wrong tracing point, to obtain needing the tracking rail of authentication signature
Mark one: if V2<T1, then the tracing point is judged for wrong tracing point and is deleted;Otherwise retain the tracing point;
Second step, set distance threshold value T2, calculate the distance L of every two tracing point in the pursuit path one for needing authentication signature2, will
Distance L2With distance threshold T2Compare to determine whether tracking exception or loss of data: if L2>T2, then judgement needs authentication signature
Pursuit path one tracks exception or loss of data, and the result of front and back two o'clock interpolation arithmetic is substituted the erroneous point, obtains needing to authenticate
The pursuit path two of signature;Otherwise, judgement needs the pursuit path one of authentication signature to track normally, obtains the tracking for needing authentication signature
Track two;
Third step carries out dimension normalization to the position coordinates for the pursuit path two for needing authentication signature, obtains needing authentication signature
Track characteristic.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610059700.7A CN105718918B (en) | 2016-01-28 | 2016-01-28 | Aerial signature authentication method based on video |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610059700.7A CN105718918B (en) | 2016-01-28 | 2016-01-28 | Aerial signature authentication method based on video |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105718918A CN105718918A (en) | 2016-06-29 |
CN105718918B true CN105718918B (en) | 2019-07-16 |
Family
ID=56154263
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610059700.7A Active CN105718918B (en) | 2016-01-28 | 2016-01-28 | Aerial signature authentication method based on video |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105718918B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107944336A (en) * | 2017-10-11 | 2018-04-20 | 中国科学院自动化研究所 | Handwriting signature verification system based on cloud computing |
CN109448189A (en) * | 2018-10-29 | 2019-03-08 | 极鱼(北京)科技有限公司 | Self-service bank's management method, self-service bank and self-service bank's system |
CN109409316B (en) * | 2018-11-07 | 2022-04-01 | 极鱼(北京)科技有限公司 | Over-the-air signature method and device |
CN113158887A (en) * | 2021-04-19 | 2021-07-23 | 福建捷宇电脑科技有限公司 | Electronic signature authentication method and equipment for improving identification accuracy of electronic signature |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104573451A (en) * | 2013-10-29 | 2015-04-29 | 深圳市腾讯计算机系统有限公司 | User verification method and device and electronic device |
CN104992192A (en) * | 2015-05-12 | 2015-10-21 | 浙江工商大学 | Visual motion tracking telekinetic handwriting system |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9437009B2 (en) * | 2011-06-20 | 2016-09-06 | University Of Southern California | Visual tracking in video images in unconstrained environments by exploiting on-the-fly context using supporters and distracters |
-
2016
- 2016-01-28 CN CN201610059700.7A patent/CN105718918B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104573451A (en) * | 2013-10-29 | 2015-04-29 | 深圳市腾讯计算机系统有限公司 | User verification method and device and electronic device |
CN104992192A (en) * | 2015-05-12 | 2015-10-21 | 浙江工商大学 | Visual motion tracking telekinetic handwriting system |
Non-Patent Citations (2)
Title |
---|
"一种改进的TLD动态手势跟踪算法";张毅等;《机器人》;20151130;第37卷(第6期);第754-759页 |
形状匹配方法研究与展望;周瑜等;《自动化学报》;20120630;第38卷(第6期);第889-908页 |
Also Published As
Publication number | Publication date |
---|---|
CN105718918A (en) | 2016-06-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108537867B (en) | Video rendering method and device according to user limb movement | |
CN105718918B (en) | Aerial signature authentication method based on video | |
CN110232311B (en) | Method and device for segmenting hand image and computer equipment | |
US6944315B1 (en) | Method and apparatus for performing scale-invariant gesture recognition | |
US20110044513A1 (en) | Method for n-wise registration and mosaicing of partial prints | |
WO2019071664A1 (en) | Human face recognition method and apparatus combined with depth information, and storage medium | |
Malassiotis et al. | Personal authentication using 3-D finger geometry | |
JP2018508888A (en) | System and method for performing fingerprint-based user authentication using an image captured using a mobile device | |
CN111079791A (en) | Face recognition method, face recognition device and computer-readable storage medium | |
CN110287918B (en) | Living body identification method and related product | |
Fang et al. | A novel video-based system for in-air signature verification | |
WO2017092296A1 (en) | Gesture unlocking method and apparatus, and mobile terminal | |
US20200302218A1 (en) | Fast curve matching for tattoo recognition and identification | |
Boyer et al. | Introduction to the special issue on recent advances in biometric systems [guest editorial] | |
KR20170092533A (en) | A face pose rectification method and apparatus | |
CN109858433B (en) | Method and device for identifying two-dimensional face picture based on three-dimensional face model | |
Uriarte-Antonio et al. | Vascular biometrics based on a minutiae extraction approach | |
US20210034895A1 (en) | Matcher based anti-spoof system | |
US20190188443A1 (en) | Information processing apparatus, biometric authentication method, and recording medium having recorded thereon biometric authentication program | |
CN102831388B (en) | Based on real-time characteristic point detecting method and the system of the moving shape model of expansion | |
US20210319204A1 (en) | Fingerprint Matching Method And Apparatus, Electronic Equipment And Readable Storage Medium | |
CN108255352B (en) | Multi-touch implementation method and system in projection interaction system | |
JP6349817B2 (en) | Alignment apparatus, alignment method, and computer program for alignment | |
WO2024045454A1 (en) | Target identification method, storage medium and device | |
CN115171223A (en) | Online signature identification method and system based on unconstrained dynamic time programming |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |