CN109284689A

CN109284689A - A method of In vivo detection is carried out using gesture identification

Info

Publication number: CN109284689A
Application number: CN201810981636.7A
Authority: CN
Inventors: 张家重; 索春宝; 胡焱; 付宪瑞
Original assignee: Suzhou Wave Intelligent Software Co Ltd
Current assignee: Inspur Financial Information Technology Co Ltd
Priority date: 2018-08-27
Filing date: 2018-08-27
Publication date: 2019-01-29

Abstract

The invention discloses a kind of methods for carrying out In vivo detection using gesture identification, it is characterised in that: speculates confidence level unit including collecting training data unit, cascade resolver unit, Face datection unit, living body；The collecting training data unit acquisition various gestures movement, and every kind of gesture action command is given, it is saved with jpg picture format and creates corresponding folder；Collected movement is trained by the cascade resolver unit, is trained to each of collecting training data unit file, and a cascade resolver is generated after each file training；The Face datection unit is used to detect the face of identification object, then carries out the identification of instruction gesture；The instruction gesture that the living body speculates that confidence level unit makes identification object is given a mark, and is then given a mark according to the recognition result of gesture, is obtained living body confidence packets and conclusion；Effectively avoid photo, video deception, target lose the disadvantages of.

Description

A method of In vivo detection is carried out using gesture identification

Technical field

The present invention relates to a kind of methods for carrying out In vivo detection using gesture identification, belong to face identification system technology neck Domain.

Background technique

With the rapid development of artificial intelligence technology, application field also constantly expands, and recognition of face is as artificial intelligence One of the important technology in field also enters into rapidly people's lives, such as mobile payment, authentication, the intelligence with recognition of face Energy robot etc..Each face identification system can face the deceptions row such as photo, video, model when carrying out face alignment For, the especially application in financial field, In vivo detection is that face recognition technology is marched toward the big obstacle of higher level one.

Now widely used In vivo detection mode mainly has that In vivo detection based on biological characteristic, movement is matched based on instruction The In vivo detection of conjunction and the In vivo detection based on special installation etc., the detection based on biological characteristic are mainly distinctive according to human body Biological detection, such as living body finger print detection are the temperature for detecting finger, perspire, the information such as electric conductivity；Living body iris detects Iridodonesis feature, pupil are detected to shrinkage expansion response feature of visible light source intensity etc.；The work of cooperation is acted based on instruction Physical examination survey refers to that requirement detected person makes corresponding instruction action, then judges to be to whether the motion detection made prepares No is living body, such as face such as turns left, turns right, opening one's mouth, blinking at the combination of some movement or multiple movements；Based on special installation In vivo detection refer to and directly detected whether with special face acquisition equipment as living body, such as lived using depth camera Physical examination is surveyed.

But above-mentioned several In vivo detection modes the disadvantages of still having a possibility that being attacked and poor experience property, it can conclude The In vivo detection of cooperation is acted based on instruction, detected object is needed accurately to make phase for the following: (1) poor user experience It should instruct, this often requires user more, not only to make instruction action, it is also necessary to pass through motion detection, it may be necessary to flower The time taken is long, such as face turns left, the angle of deflection, holding time bad setting, in fact it could happen that user turns left dynamic Work is too fast, and system does not capture left-hand rotation movement, it is necessary to which user cooperates again；Blink detection, due to blink Movement it is smaller, movement is than very fast, it is difficult to capture, it may be necessary to which repeated detection, it is irritated that these can all cause user to generate, body The property tested is poor；(2) equipment requirement is high, and there are many limitations；In vivo detection based on biological characteristic may need to be equipped with infrared Camera just may be implemented, such as carry out iris detection, and common camera is unable to reach requirement；Living body inspection based on special installation Surveying is even more to need fitting depth camera, these equipment requirements are unable to satisfy in fields such as mobile payments；(3) there are safety Hidden danger is easy to be spoofed；In vivo detection based on physiological characteristic may be cheated by other living bodies, such as carry out fingerprint detection When, it is substituted for the living body of another non-detected person, acts the In vivo detection of cooperation based on instruction, it may be by photo, video Deng deception, such as the photo of a certain movement is made, makes a series of actions video；In vivo detection based on special installation, may When can be cheated by faceform, such as be detected using depth camera, the headform that a 3D printing goes out may be regarded as It is living body.

Summary of the invention

For the above technical problems, the purpose of the present invention is: propose it is a kind of utilize gesture identification carry out living body The method of detection promotes the accuracy and user experience of face In vivo detection, reduces cheated possibility to the greatest extent.

The technical solution of the invention is as follows is achieved: a method of In vivo detection being carried out using gesture identification, Confidence level unit is speculated including collecting training data unit, cascade resolver unit, Face datection unit, living body；The training Data acquisition unit acquires various gestures movement, and gives every kind of gesture action command, is saved with jpg picture format and creates correspondence File；Collected movement is trained by the cascade resolver unit, to each of collecting training data unit File is trained, and a cascade resolver is generated after each file training；The Face datection unit is used to detect It identifies the face of object, then carries out the identification of instruction gesture；The living body speculates the finger that confidence level unit makes identification object It enables gesture give a mark, is then given a mark according to the recognition result of gesture, obtain living body confidence packets and conclusion.

Due to the application of the above technical scheme, compared with the prior art, the invention has the following advantages:

A kind of method carrying out In vivo detection using gesture identification of the invention, this method is on the basis for detecting face On, judge whether it is living body using the corresponding gesture that detected object is made, gesture motion have movement is clear, be easy to capture, The features such as detection, and can be easy to stop than movements such as blink, rotary heads and be identified for more time, simultaneously operation instruction is random Generation, diversity can effectively avoid by force the disadvantages of photo, video deception, target loss.

Specific embodiment

The present invention addressed below.

A kind of method carrying out In vivo detection using gesture identification of the present invention, including collecting training data unit, It cascades resolver unit, Face datection unit, living body and speculates confidence level unit；The collecting training data unit acquires a variety of hands Gesture movement, and every kind of gesture action command is given, corresponding folder is saved and created with jpg picture format, and the movement of acquisition is as far as possible Comprehensively, varied, each movement will carry out multiple angular positive-negative face acquisitions, such as stretching index finger and middle finger show " V " Movement, will acquire " V " towards multiple angles, each angular positive-negative will acquire, each movement is stored in a file.

Collected movement is trained by the cascade resolver unit, to each of collecting training data unit File is trained, and a cascade resolver, the cascade resolver that such as " V " movement generates are generated after each file training It is named as " victory.xml ", and so on, generate multiple cascade resolvers.

Face datection is the basis of all In vivo detections, carries out the method for In vivo detection also in inspection using gesture identification It measures and carries out on the basis of face, after successfully being detected face, then carry out the identification of instruction gesture, the Face datection unit For detect identification object face, then carry out instruction gesture identification.

The instruction gesture that the living body speculates that confidence level unit makes identification object is given a mark, then according to gesture Recognition result is given a mark, and living body confidence packets and conclusion are obtained；Instruction gesture is made to identification object to give a mark, it is first First, different according to the complexity of gesture, assignment is carried out for each gesture, more complicated gesture assignment is higher；Then basis The recognition result of gesture is given a mark；In order to realize gesture identification carry out In vivo detection ease for use and uniformity, it is specified that each Group gesture is made of 5 gesture motions, includes the gesture of five score values in each group, score value is respectively 0.1,0.2,0.3,0.4, 0.5, each gesture identification result top score is 100 points, and the full marks of each group in this way gesture are all 150 points, according to last Score calculate living body confidence level.For example, it is desired to which object to be identified makes set, this group includes 5 movements, respectively 1-5 finger is stretched out, the assignment of five movements is respectively 0.1,0.2,0.3,0.4,0.5, each gesture identification result highest 100 points are scored at, being multiplied then to sum with the marking of its recognition result with the score value of each gesture obtains the gesture knowledge of the group Then other score is divided with the score divided by full marks 150, obtain living body confidence level, and with the In vivo detection confidence level threshold that sets Value, which compares, draws a conclusion.

The objective environments such as gesture identification will receive the installation of light, camera in practical application scene, object is blocked because Therefore the influence of element needs to carry out a large amount of simulation test in advance in practical applications, is suitble to based on the actual application requirements The threshold value of In vivo detection confidence level under the scene.

Before detection, collecting training data unit acquires various gestures movement, and cascade resolver unit completes each movement Training and name, and set according to objective environment factor the threshold value of In vivo detection confidence level, when detection, the inspection of Face datection unit The face for surveying living body is identified, the identification of instruction gesture is carried out after detection face success, and detection face failure, which then returns, not to be examined It measures the result of face and prompts to detect again, be detected personnel according to instruction gesture and complete corresponding act and until living body can Confidence score, the threshold value comparison of obtained score and setting obtain In vivo detection conclusion.

A kind of method carrying out In vivo detection using gesture identification of the invention, this method is on the basis for detecting face On, judge whether it is living body using the corresponding gesture that detected object is made, gesture motion have movement is clear, be easy to capture, The features such as detection, and can be easy to stop than movements such as blink, rotary heads and be identified for more time, simultaneously operation instruction is random Generation, diversity can effectively avoid by force the disadvantages of photo, video deception, target loss；The In vivo detection that this method proposes is credible The calculation method of degree is comprehensive one group of gesture motion to judge living body, and more previous single action judges that the method for living body increases It is credible.

The above embodiments merely illustrate the technical concept and features of the present invention, and its object is to allow person skilled in the art Scholar can understand the contents of the present invention and be implemented, and it is not intended to limit the scope of the present invention, it is all according to the present invention Equivalent change or modification made by Spirit Essence, should be covered by the scope of protection of the present invention.

Claims

1. a kind of method for carrying out In vivo detection using gesture identification, it is characterised in that: including collecting training data unit, cascade Resolver unit, Face datection unit, living body speculate confidence level unit；The collecting training data unit acquisition various gestures are dynamic Make, and give every kind of gesture action command, is saved with jpg picture format and create corresponding folder；The cascade resolver unit Collected movement is trained, each of collecting training data unit file is trained, each file A cascade resolver is generated after folder training；The Face datection unit is used to detect the face of identification object, then is instructed The identification of gesture；The instruction gesture that the living body speculates that confidence level unit makes identification object is given a mark, then according to hand The recognition result of gesture is given a mark, and living body confidence packets and conclusion are obtained.