Latest News 

Product Info
Company News
Activity info

 Home News 
Each big giant layout face recognition, IR LED factory have meat to eat!
Publisher:waimao Published:2016/9/5 9:47:49 Read:6102 【Font:L M S

Face recognition or stand out from the crowd


From fingerprint identification to iris recognition, biometrics adopted by more and more consumer electronics manufacturers, everybody about the technology into the white-hot, but technology makes little sense to compare, the market can tell.




According to forward-looking industry research institute, during the six years from 2007 to 2013, biometric technologies global market size of the average annual growth of 21.7%, the most in the global industry growth rate of less than 5% of a rare.Biometric technologies in 2015 global market scale will reach $13 billion, will reach $25 billion in 2020, five years the average annual growth of about 14%.



From 2015 to 2015, the industry are: increase the size of the market segment fingerprint (73.3%), voice (100%) and face (166.6%), iris (100%), other (140%).Many face recognition in biological recognition technology on the growth in the first place, is expected to face recognition technology to 2020 the size of the market will rise to $2.4 billion.We expect that the intelligent terminal penetration under the condition of face recognition, the size of the market may be larger than expected.


Five biometric each have characteristics, but from the market share, after the fingerprint identification is the most likely to stand out is expected to face recognition.Business point of view, pay treasure, brokers, Banks of financial services have been in the last year a large number of accounts, transfer, payment, etc., using facial recognition in China merchants bank, for example, face recognition can realize transfer of more than 500000 mobile end, visible for its security.


1. The fingerprint identification is the most widely used biometric technology, mature technology and low cost, are widely used in the attendance, entrance guard and other identification.But easy to copy, and after the impact of the fingerprint recognition accuracy.


2. The feature of iris recognition by using the human eye image of iris (rings, wrinkles, spots, corona) template form characteristics, through comparing the characteristic parameters of complete recognition.High recognition accuracy, this method is not easy to imitate but related equipment is expensive.


3. The voice recognition through the analysis of the unique characteristics of voice authentication, the scope of its equipment from big, easy to install, but the recognition accuracy is low, may be deceived by the recording, and vulnerable to background noise, such as health, emotional factors.


4. The signature recognition is based on the behavior characteristics of biological recognition technology, through the analysis of handwriting, pressure, writing speed for authentication.But the signature copy can be high, and the signature tool, emotions, etc all can cause interference to the signature recognition.


5. Compared with other biometric facial recognition, advantage lies in the following characteristics: natural, not being noticed.Naturalness is the recognition method is used to identify the individual with the humanity by use of the biological characteristics of the same, human also by observing more face, voice and other information on other individuals to distinguish and confirm.Therefore, the fingerprint recognition, iris recognition are not natural.Does not detect the characteristics of the resistance, the recognition method is not easy to make people use visible light can get face image information, and the fingerprint recognition or iris recognition using electronic pressure sensor or infrared fingerprint, iris image, identity in the process of gathering information that may be counterfeit.



 Giants are layout


Any technological innovation can know ahead of time in the patent and acquisitions, face recognition into an intelligent terminal is not just a guess, giants face recognition has long been active layout.


1. The company has successively acquired Polar Rose, Prime Sense, Perceptio, Faceshift, Emotient, face recognition technology companies such as Turi;


2. Samsung on December 29, 2010 to apply for face recognition device, algorithm and machine readable media patents;On June 19, 2004, patent application for equipment and the algorithm of image recognition feature extraction;


3. The company to apply for at least 10 patents of face recognition related, such as, determine face in face images, and identification method of terminal;


4. Google (Google) has acquired company PittPatt and Viewdle face recognition system, until June of 2016, a total of to apply for 21 face recognition related patents;


5. Facebook acquisition of Tel Aviv has launched in 2014, facial recognition software DeepFace;


6. Amazon (Amazon), Microsoft (Microsoft) are respectively applied for 7 and 6 patents of face recognition.



How to implement face recognition


Face recognition is mainly divided into two parts of the face detection and face than in.Its working process is:


1. The image acquisition: through sampling sensors, such as camera face image;


2. Face positioning and extraction, and then to deal with data collected, eliminate the noise in the data gathering and environmental factors, to extract the characteristics of the samples to characterization of personal identity information;


3. The character comparison: put the feature information and comparing the information is already available in the database;


4. Output: according to the comparison of similar degree to determine whether the match.



The current face recognition market solutions include: 2 d, 3 d recognition, thermal recognition, on the market at present the mainstream recognition scheme is to use the 2 d of cameras.2 d face recognition is based on the graphic image recognition method, but due to the person's face is not flat, so 2 d recognition in will exist in the process of 3 d information of complanation projection face feature information loss. 



 3 d recognition using 3 d face stereo modeling method, can keep effective information to the greatest extent.So the algorithm of 3 d face recognition than 2 d algorithm is more reasonable and has higher precision.Thermal sensor identification technology using a three layer BP (back propagation) feed-forward neural network as classifier, at the same time you use thermal information will not be hairstyle, breathing and other environmental factors influence the key facial geometric information, such as the Angle of the bridge of the nose, cheek area, and so on, to enhance the identification accuracy.


Looked from the present development, mainly divided into commercial systems, mainstream software and algorithm, we believe that the integrated application of terminal equipment needs a complete set of solutions, big players has advantages in this respect.Component parts of the face recognition mainly involves: the software part for the database, algorithm;Hardware part for camera module, integrated devices, sensors, chip, IC, hardware interface circuit, LCD display, storage, etc.;System service providers, and integration, software and hardware manufacturers.The algorithm as the core industrial chain link, at the same time is also the highest technical barriers.Face recognition from the current domestic design company, mainly provide camera algorithm, etc., the value of the hardware can be ignored.



 Infrared leds narrow-band filter is expected to become the core elements


Traditional face recognition technology is mainly based on visible light image of face recognition, but this way is difficult to overcome the defect of the near infrared face recognition system can completely solve the problem of environment light.Traditional recognition can be in the light changes, to identify the effect will be dropped sharply, unable to meet the needs of actual system.Meet side light when you take a photo, for example, the phenomenon of "face" of Yin and Yang, it may not be able to correctly identify.


Solution to the problem of illumination three-dimensional image with face recognition, and thermal imaging facial recognition.But now the two technologies is not yet mature, recognition effect.Based on near infrared image face recognition of core technologies and systems, in different light conditions, to capture is not affected by environment illumination change of near infrared human face image, combined with advanced algorithm, can achieve high recognition rate.


 Near infrared face recognition includes two parts: active face near infrared imaging device and the corresponding light has nothing to do face recognition algorithm.Use strength higher than that of ambient light active imaging, near infrared light source with corresponding band optical filter, can get face image that is independent of the environment face image will only change as people and the distance of the camera and monotonous changes.


Image on this specific feature extraction method, such as: Local Binary Pattern (Local Binary Pattern, LBP) characteristics, can further remove the image of monotonous change, fully and light has nothing to do with the characteristics of the expression.In recent years, near infrared face recognition in real life has many applications, such as: "shenzhen - Hong Kong automated customs clearance system for biological passport", "macau - zhuhai automated customs clearance system for

biological passport", "Beijing airport T3 terminal building automated

customs clearance system", etc., have achieved good effect.


Active face near infrared imaging device for face recognition is not affected by the ambient light, the high quality of face image, the so-called high quality include: image brightness, uniformity, contrast the right, there is no excessive exposure, etc.Active face near infrared imaging equipment generally includes the following units:


In the corresponding active near infrared band intensity is higher than the ambient light of light source, is usually 850 nm and 940 nm infrared LED high power;


To receive near-infrared light camera, usually in the form of CCD image sensor.CCD has small volume, light weight, small distortion, low power consumption, low voltage driver, impact resistant, the advantages of strong vibration resistance, resistance to electromagnetic interference, so they are widely used in all kinds of image acquisition system.In face recognition system of CCD is basically a silicon substrate, the spectral response range of 400 nm to 1100 nm, the scope is narrow band filter spectral range to consider;


Narrowband filter,  in camera lens, allows the near-infrared light through filtering the ambient light at the same time.Is mainly used to isolate the interference light, through the signal light, fully highlighted the useful information, reduce the interference information, and lay a foundation for subsequent image processing and recognition.


Technology on infrared LED narrow-band filter is expected to become the core factor.At present some solutions using isolation visible light through infrared infrared glass as a filter, isolated from the ordinary infrared glass, however, is only visible and ultraviolet light, and not a part of the isolation interference light at infrared wavelengths.So you want a good anti-jamming effect, narrow-band filter must be used.


Selection of narrow-band filters need to consider multiple optical indicators, including bandwidth, center wavelength, cutoff wavelength, by the depth, the peak transmittance, product thickness and so on.From the recent camera and the development of AR, domestic optics has become the main supplier of reaction of domestic optical power enough consumer electronics and the demand of the special display, narrow-band filter module international big customers will still be according to the domestic suppliers.



Print || Close
Home News
Friendly Links:

Copyright @ ShenZhen Kena Co.,Ltd All Rights Reserved
Tel:86-755-27168119 Fax:86-755-27229986 Address: BaoAn District, ShenZhen City, GuangDong Province. CHINA