WO2021244414A1 - 基于光谱与多波段融合的人脸识别监控系统及识别方法 - Google Patents

基于光谱与多波段融合的人脸识别监控系统及识别方法 Download PDF

Info

Publication number
WO2021244414A1
WO2021244414A1 PCT/CN2021/096673 CN2021096673W WO2021244414A1 WO 2021244414 A1 WO2021244414 A1 WO 2021244414A1 CN 2021096673 W CN2021096673 W CN 2021096673W WO 2021244414 A1 WO2021244414 A1 WO 2021244414A1
Authority
WO
WIPO (PCT)
Prior art keywords
face
image
spectrum
infrared
spectral
Prior art date
Application number
PCT/CN2021/096673
Other languages
English (en)
French (fr)
Inventor
任玉
聂刚
周浩
刘禹辰
刘晓慧
王朔
张永生
蔡红星
姚治海
Original Assignee
吉林求是光谱数据科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 吉林求是光谱数据科技有限公司 filed Critical 吉林求是光谱数据科技有限公司
Publication of WO2021244414A1 publication Critical patent/WO2021244414A1/zh
Priority to US17/555,996 priority Critical patent/US11443550B2/en

Links

Images

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
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/161—Detection; Localisation; Normalisation
    • G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/56—Extraction of image or video features relating to colour
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/58—Extraction of image or video features relating to hyperspectral data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • 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
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • 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
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/168—Feature extraction; Face representation
    • 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
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/168—Feature extraction; Face representation
    • G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
    • 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
    • G06V40/40—Spoof detection, e.g. liveness detection
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/10—Cameras or camera modules comprising electronic image sensors; Control thereof for generating image signals from different wavelengths
    • H04N23/11—Cameras or camera modules comprising electronic image sensors; Control thereof for generating image signals from different wavelengths for generating image signals from visible and infrared light wavelengths
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/20—Cameras or camera modules comprising electronic image sensors; Control thereof for generating image signals from infrared radiation only
    • H04N23/21—Cameras or camera modules comprising electronic image sensors; Control thereof for generating image signals from infrared radiation only from near infrared [NIR] radiation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/50—Constructional details
    • H04N23/54—Mounting of pick-up tubes, electronic image sensors, deviation or focusing coils
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/70—Circuitry for compensating brightness variation in the scene
    • H04N23/74—Circuitry for compensating brightness variation in the scene by influencing the scene brightness using illuminating means
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N25/00—Circuitry of solid-state image sensors [SSIS]; Control thereof
    • H04N25/10—Circuitry of solid-state image sensors [SSIS]; Control thereof for transforming different wavelengths into image signals
    • H04N25/11—Arrangement of colour filter arrays [CFA]; Filter mosaics
    • H04N25/13—Arrangement of colour filter arrays [CFA]; Filter mosaics characterised by the spectral characteristics of the filter elements
    • H04N25/131—Arrangement of colour filter arrays [CFA]; Filter mosaics characterised by the spectral characteristics of the filter elements including elements passing infrared wavelengths
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N25/00—Circuitry of solid-state image sensors [SSIS]; Control thereof
    • H04N25/10—Circuitry of solid-state image sensors [SSIS]; Control thereof for transforming different wavelengths into image signals
    • H04N25/11—Arrangement of colour filter arrays [CFA]; Filter mosaics
    • H04N25/13—Arrangement of colour filter arrays [CFA]; Filter mosaics characterised by the spectral characteristics of the filter elements
    • H04N25/135—Arrangement of colour filter arrays [CFA]; Filter mosaics characterised by the spectral characteristics of the filter elements based on four or more different wavelength filter elements
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N5/00—Details of television systems
    • H04N5/30—Transforming light or analogous information into electric information
    • H04N5/33—Transforming infrared radiation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10048—Infrared image

Definitions

  • the invention belongs to the field of optics, relates to the technical field of face recognition and monitoring, and specifically relates to a face recognition monitoring system and recognition method based on spectrum and multi-band fusion.
  • Video surveillance has always played an important role in the security field.
  • face recognition technology has entered the field of surveillance with the most friendly biometric authentication technology, which dynamically captures human facial images through a camera, and then compares and recognizes it.
  • the core of face recognition technology lies in the use of the azimuth relationship of the facial organs and feature parts or the skeleton characteristics of the human body posture to form the recognition parameters, which are recognized and calculated with the parameters in the template database to finally give a judgment, for example: CN 110796079 A public "Method and system for multi-camera visitor recognition based on face depth features and human body local depth features".
  • Chinese patent CN110874588A discloses "method and device for dynamically optimizing the influence of light in face recognition", and proposes a method of setting RGGB camera and infrared camera binocular camera, using infrared camera to perform face recognition processing to avoid the influence of light conditions.
  • Chinese patent CN108090477A discloses "a method and device for face recognition based on multi-spectral fusion” also proposes to use RGGB images and infrared images for face recognition. Although the use of infrared images can solve the problem of poor light, it still does not play an effective role in identifying real and fake faces. In addition, infrared cameras are relatively expensive and not suitable for promotion.
  • Chinese patent CN202887210U discloses "a multi-spectral face recognition system".
  • the system uses a multi-spectral imaging system to obtain the spectral feature information of a living body, and uses the spectral reflectance of the facial skin to identify true and false faces, avoiding the process of face recognition Some of the disguised deceptions in.
  • the high-precision spectrometer used in the identification system is expensive and still not suitable for promotion. Therefore, it is necessary to develop a face recognition monitoring system that is inexpensive, can effectively recognize true and false faces, and recognizes accurately.
  • the first object of the present invention is to provide a face recognition monitoring system based on spectrum and multi-band fusion.
  • the system divides the silicon-based detector into multiple units, each unit has three regions, The spectral modulation film layers of different bands are etched separately on the area, so that the reflection multi-spectral data of the face in the visible-near-infrared light band, the near-infrared band face image and the RGGB face image can be obtained through a single camera.
  • Reduce the cost of the monitoring system and when the system recognizes the face, it can use the spectral data to identify whether it is real human skin, and then combine the near-infrared band face image and the RGGB face image for further recognition.
  • the dual recognition effectively ensures that the person is The accuracy of face recognition improves recognition security, and solves the drawbacks of the current monitoring system's face recognition misjudgment.
  • a face recognition monitoring system based on spectrum and multi-band fusion includes a spectrum camera, a face spectrum image acquisition module, a face spectrum image data preprocessing module, a face spectrum image database, and a face spectrum image recognition module;
  • the spectroscopic camera includes an optical lens and a silicon-based detector.
  • the silicon-based detector includes a photoelectric conversion substrate and a filter film arranged on the photoelectric conversion substrate;
  • the filter film includes N units, each unit Including three areas, specifically the visible spectrum sensing area, the near-infrared spectrum image sensing area and the RGGB image acquisition area, N units covering all pixels on the photoelectric conversion substrate;
  • the visible spectrum sensing area is used to distinguish the incident visible light spectrum; the visible spectrum sensing area is formed by splicing n materials with known and different light transmittances through coating and etching one by one.
  • the visible spectrum sensing area Single-layer structure, including T 1 , T 2 ??T n sub-units, each sub-unit covers M pixels on the photoelectric conversion substrate, where M is greater than or equal to 1, visible spectrum sensing area and near-infrared spectrum image
  • the sensing area and the RGGB image acquisition area constitute a periodic structure;
  • the transmission bands of the near-infrared spectral image sensing area are divided into 780-850nm, 850-950nm, 950-1050nm, 1050-1200nm, which are used to detect the deep layer feature image of the near-infrared light band with a wavelength greater than 780nm.
  • Four materials with different near-infrared transmittances are formed by coating and etching one by one.
  • the sensing area of the near-infrared spectrum image is a single-layer structure, including four subunits of Tir1, Tir2, Tir3, and Tir4. Each subunit Cover M pixels on the photoelectric conversion substrate, where M is greater than or equal to 1;
  • the RGGB image acquisition area is used for the human face color image acquisition in the monitoring area.
  • the RGGB image acquisition area includes four sub-units in three colors of transmission red, green, green and blue, and each sub-unit covers M pixels on the photoelectric conversion substrate , Where M is greater than or equal to 1;
  • the photoelectric conversion substrate is used to convert the light signal passing through the visible spectrum sensing area into an electrical signal, and form a digital signal or code output after amplification and analog-to-digital conversion; convert the light signal passing through the near-infrared spectrum image sensing area into The electrical signal is amplified and converted into a digital signal or coded output; the optical signal passing through the RGGB image acquisition area is converted into an electrical signal, and then amplified and converted into a digital signal or coded output;
  • the face spectrum image acquisition module is used to reverse the face spectrum data and image data according to the incident light signal intensity information and the corresponding photoelectric conversion substrate pixel position information, including the reflection multispectral data of the face in the visible-near infrared band ,
  • the near-infrared waveband face image and the RGGB face image; the near-infrared waveband face image and the RGGB face image are acquired by image inversion, and the inversion method of the multispectral data is shown in formula (1)
  • S is the intensity value of the optical signal output by the photoelectric conversion substrate
  • I is the incident spectrum, which is the signal to be solved
  • T is the spectral transmittance of the filter film, including the spectral transmittance of the filter film in the visible light band and near infrared
  • ⁇ is the quantum efficiency of the photoelectric conversion substrate
  • ⁇ is the incident wavelength
  • the face spectral image data preprocessing module is used to denoise the face images acquired in the near-infrared spectral image sensing area and the RGGB image acquisition area, and to reduce the noise of the RGGB face image and the near-infrared band face Image feature extraction, the multi-spectral image acquired in the visible spectrum sensing area and the multi-spectral image acquired in the near-infrared spectrum sensing area, extract the multi-spectral data of the face, and combine the multi-dimensional facial feature data with the face spectrum Compare images and multispectral data in the image database;
  • the face spectral image database is used to store the image feature values and spectral data of the N x characteristic faces to be searched, and the multi-spectral data of a variety of mask materials, so as to facilitate the comparison with the face images and their spectra appearing in the monitoring area Comparison;
  • the face spectral image recognition module is used to combine the visible-near-infrared band reflection multi-spectral data of the face, the preprocessed near-infrared band face image, and the RGGB face image data with the information in the face spectral image database comparing.
  • the system also includes a supplemental light source whose emission band includes visible light and near-infrared light, which is used to monitor people in the surveillance area at night or under low-light conditions to enhance deep-level image collection.
  • a supplemental light source whose emission band includes visible light and near-infrared light, which is used to monitor people in the surveillance area at night or under low-light conditions to enhance deep-level image collection.
  • the optical lens can automatically adjust the focal length according to the target distance in the monitoring area.
  • the silicon-based detector is a CMOS image sensor or a CCD image sensor.
  • the method for preparing the visible spectrum sensing area is as follows: select n kinds of polyimide filter film materials with different spectral transmittances, and first coat the first filter film on the photoelectric conversion substrate Material, and then coat an etching layer, according to the corresponding relationship with the photoelectric conversion substrate pixels, the necessary places are reserved, and the unnecessary places are etched away; then the second filter film material is coated, and then coated An etching layer, according to the corresponding relationship with the pixels of the photoelectric conversion substrate, keep the necessary places and etch away the unnecessary places; cycle in turn until all n kinds of filter film materials are coated on the photoelectric conversion substrate
  • the preparation method of the near-infrared spectrum image sensing area is the same as the preparation method of the visible spectrum sensing area, and the near-infrared spectroscopy image sensing area is coated with four kinds of filter film materials.
  • a microlens array is prepared on the filter film, and the microlens array is coated with a layer of transparent photoresist for preparing the microlens array on the filter film, and then direct laser writing is used. Or the mask lithography method etches the unnecessary part, and the remaining part constitutes a microlens array.
  • Each microlens on the microlens array corresponds to the pixel of the photoelectric conversion substrate one-to-one, and is used to transfer the incident light beam Converge.
  • the second object of the present invention is to provide a recognition method for a face recognition monitoring system based on spectrum and multi-band fusion, which includes the following steps:
  • Step S1 When the person to be tested enters the monitoring area, the optical lens of the spectrum camera automatically focuses, so that the face of the inspector is clearly imaged; The face of multiple inspectors is clearly imaged. When the image is not clear due to insufficient light, the supplementary light source is turned on;
  • Step S2 Start the silicon-based detector and the face spectrum image acquisition module, the visible spectrum sensing area and the near-infrared spectrum image sensing area on the silicon-based detector collect the reflected multispectral data of the target person’s face, and the near-infrared spectrum image sensing area Collect face image information in the near-infrared band, and collect face color image information in the RGGB image collection area;
  • Step S3 Compare the multi-spectral data obtained in step S2 with the face spectrum data in the face spectrum image database.
  • the living body tissue in the database is compared with the multispectral data of multiple mask materials, and the comparison result shows that it matches the living body tissue, then go to step S4; if the comparison result shows that it matches the mask material, then the monitoring area and the police alarm are both Sounds, the test ends; if there is no matching result between the two, it is possible that the person under test may change the facial features through makeup, and then proceed to step S4';
  • Step S4 After the comparison result in step S3 shows the living body tissue, the face spectral image data preprocessing module will perform denoising processing on the N face color images acquired in the RGGB image acquisition area during the dynamic process of the person. Feature extraction is performed on the noisy image to obtain facial feature data. The face spectral image recognition module compares the facial feature data with the facial image feature values in the face spectral image database. If there is no matching data in the comparison result, then Delete the acquired RGGB face image; if there is matching data, the alarm with the police in the monitoring area will sound, and the detection will end;
  • Step S4' In step S3, the comparison result shows non-living tissues.
  • the face spectral image data preprocessing module performs denoising processing on the N near-infrared face images acquired during the dynamic process of the person, and will reduce the noise. Feature extraction is performed on the noisy image to obtain facial feature data.
  • the face spectral image recognition module compares the facial feature data with the facial image feature values in the face spectral image database. If there is no matching data in the comparison result, then Delete the acquired face images in the near-infrared band; if there is matching data, both the alarms in the surveillance area and the police will sound, and the detection will end.
  • the face recognition monitoring system provided by the present invention divides the silicon-based detector into a plurality of units, each unit has three areas, and each area is etched on the spectral modulation film of different bands, so that one camera can be used.
  • the face recognition monitoring system integrates visible-near-infrared light band spectral data, near-infrared light band images and RGGB images to recognize faces to judge the authenticity of the face, and exclude masking by means such as makeup
  • the real face realizes more accurate recognition of human faces, greatly improves the accuracy of monitoring, prevention and early warning, and effectively prevents criminals from using silica gel and other materials to imitate human face masks.
  • the visible spectrum modulation film layer and the near-infrared spectrum modulation film layer of the silicon-based detector in the face recognition monitoring system provided by the present invention adopt a single-layer structure, and the formed photoelectric sensor has a simple structure, small volume and thin thickness (for Micron level), light weight, high spectral resolution, and spatial resolution, high accuracy, and fast detection speed. It can be integrated into any existing monitoring device to achieve spectrum extraction and The high-precision imaging function makes the extracted face clearer and more accurate.
  • the method for identifying the true and false face patterns of the integrated image and spectral information of the present invention uses the existing silicon-based detector coating method to achieve more accurate face recognition, and the photoelectric sensor obtains the unique characteristics of human skin. Reflectance spectrum information and face image information are used to form a low-cost, ultra-convenient face recognition system using a comparison method.
  • Fig. 1 is a schematic diagram of the monitoring system for real and false face recognition of the present invention.
  • Fig. 2 is a schematic diagram of the structure of the filter film unit of the present invention.
  • Fig. 3 is a schematic diagram of the structure of the filter film of the present invention.
  • Fig. 4 is a schematic diagram of the structure of the silicon-based photoelectric sensor of the present invention.
  • Fig. 5 is a schematic diagram of the structure of the silicon-based detector of the present invention.
  • Fig. 6 is a schematic diagram of the method for identifying true and false faces of the present invention.
  • Fig. 7 is a schematic diagram of multi-spectral data of the visible light waveband living tissue and the silicone mask of the present invention.
  • Fig. 8 is a schematic diagram of multi-spectral data of living tissues and silicone masks in the near-infrared light waveband of the present invention.
  • Embodiment 1 Face recognition monitoring system based on spectrum and multi-band fusion
  • the present invention provides a face recognition monitoring system based on spectrum and multi-band fusion.
  • the system includes a spectrum camera 1, a supplementary light source (a wide-band white light source, which may be a bromine tungsten lamp), and a face spectral image Acquisition module 2, face spectral image data preprocessing module 3, face spectral image database 4, face spectral image recognition module 5;
  • a spectrum camera 1 a supplementary light source (a wide-band white light source, which may be a bromine tungsten lamp)
  • a face spectral image Acquisition module 2 face spectral image data preprocessing module 3
  • face spectral image database 4 face spectral image recognition module 5;
  • the spectroscopic camera includes an optical lens and a silicon-based detector.
  • the optical lens can automatically adjust the focal length according to the target distance in the monitoring area.
  • the silicon-based detector is a CMOS image sensor, and the detection band is 200-1200nm.
  • the silicon-based detection The device includes a photoelectric conversion substrate 101 and a filter film 102 arranged on the photoelectric conversion substrate.
  • the filter film 102 has a thickness of 10 ⁇ m, a size of 4 mm ⁇ 4 mm, and a data collection time of 1 ms; the filter film 102 includes N
  • Each unit includes three areas, specifically the visible spectrum sensing area 6, the near-infrared spectrum image sensing area 7, and the RGGB image acquisition area 8.
  • N units cover all the pixels on the photoelectric conversion substrate, as shown in Figure 2.
  • the visible-near-infrared light waveband reflection multispectral data of the face, the near-infrared waveband face image and the RGGB face image are collected through three sensing areas;
  • the visible spectrum sensing area 6 is used to distinguish the incident visible light spectrum; the visible spectrum sensing area is made of 8 polyimide materials with known and different light transmittances through coating, etching and splicing one by one.
  • the visible spectrum sensing area is a single-layer structure, including T 1 , T 2 ??T 8 sub-units, each sub-unit covers 1 pixel on the photoelectric conversion substrate, each sub-unit for the photoelectric conversion substrate image Elements (pixels) have different spectral transmittances, and their spectral transmittances change, the spectral modulation characteristics on different pixels in each sub-unit are different, and the combination can restore and invert the spectral line pattern of the incident light to realize the spectrum Spectroscopy function; visible spectrum sensing area, near-infrared spectrum image sensing area and RGGB image acquisition area form a periodic structure;
  • the transmission wavelength bands of the near-infrared spectral image sensing area 7 are divided into 780-850nm, 850-950nm, 950-1050nm, 1050-1200nm, which are used to detect deep facial feature images in the near-infrared light band with a wavelength greater than 780nm, excluding the use of makeup
  • the near-infrared spectrum image sensing area is composed of four materials with known and different near-infrared transmittances through coating and etching one by one.
  • the near-infrared spectrum image sensing area is Single-layer structure, including four sub-units Tir1, Tir2, Tir3, and Tir4, each sub-unit covering 1 pixel on the photoelectric conversion substrate;
  • the RGGB image acquisition area 8 is used to acquire human face color images in the monitoring area; the RGGB image acquisition area includes four subunits in three colors of transmission red, green, green, and blue, and each subunit covers one of the photoelectric conversion substrates. Pixel
  • the filter film 102 is provided with a microlens array 103, and the microlens array 103 is coated on the filter film 102.
  • the layer is used to prepare the transparent photoresist (epoxy resin material) of the microlens array, and then use laser direct writing or mask photolithography to etch the unnecessary part, and the remaining part constitutes the microlens array 103,
  • Each microlens on the microlens array corresponds to the pixel of the photoelectric conversion substrate one-to-one, and is used to converge the incident light beam.
  • the total thickness of the microlens array 103 and the filter film 102 is 30 ⁇ m, and the size is 4mm ⁇ 4mm.
  • the resolution is 10nm, and the data acquisition time is 1ms;
  • the photoelectric conversion substrate is used to convert the light signal passing through the visible spectrum sensing area 6 into an electrical signal, and form a digital signal or code output after amplification and analog-to-digital conversion; the light signal passing through the near-infrared spectrum image sensing area 7 Converted into electrical signals, after amplification and analog-to-digital conversion, to form a digital signal or coded output; convert the optical signal passing through the RGGB image acquisition area 8 into an electrical signal, and form a digital signal or coded output after amplification and analog-to-digital conversion;
  • the supplementary light source whose light source emission band includes visible light and near-infrared light, is used to monitor personnel in the monitoring area at night or in a low-light environment and to enhance deep-level image collection;
  • the face spectral image acquisition module is used to reverse the face spectral data and image data according to the incident light signal intensity information and the corresponding photoelectric conversion substrate pixel position information, including the multi-spectral data reflected by the face in the visible-near infrared band , Near-infrared waveband face image and RGGB face image; the near-infrared waveband face image and RGGB face image are acquired by image inversion (according to the spectrum on each pixel corresponding to the known spectral transmittance information , To correct the intensity value of the light signal on the corresponding pixel, the correction method is to divide the intensity value of the light signal on the pixel by the spectral transmittance value on the pixel; the combination of all the pixels can invert the image information);
  • the inversion method of multispectral data is shown in formula (1)
  • S is the intensity value of the optical signal output by the photoelectric conversion substrate
  • I is the incident spectrum, which is the signal to be solved
  • T is the spectral transmittance of the filter film, including the spectral transmittance T 1 of the filter film in the visible light band.
  • the combination of 8 pixels in the visible spectrum sensing area and 4 pixels in the near-infrared spectrum image sensing area can invert and calculate the visible incident spectral values and near-infrared spectra of the 8 pixels on the visible spectrum sensing area Near-infrared spectrum values of 4 pixels on the image sensing area;
  • the face spectral image data preprocessing module 3 is used to perform noise reduction processing on the face images acquired in the near-infrared spectral image sensing area and the RGGB image acquisition area (multiple images are taken in a short period of time, and the multiple images are divided by pixel. Take the average, for example, take 10 pictures, take the average of the first pixel of each picture, take the average of the second pixel....
  • the multi-spectral image acquired in the visible spectrum sensing area and the multi-spectral image acquired in the near-infrared spectrum sensing area are extracted from the multi-spectral data of the face, and the multi-dimensional human Compare the facial feature data with the image and multispectral data in the face spectral image database 4;
  • the face spectral image database 4 is used to store the image feature values and spectral data of the N x characteristic faces to be searched, as well as the multi-spectral data of a variety of mask materials, so as to facilitate the comparison with the face images and their appearance in the monitoring area. Spectral comparison;
  • the face spectral image recognition module 5 is used to combine the visible-near infrared band reflection multi-spectral data of the face, the preprocessed near-infrared band face image and the RGGB face image data with the face spectral image database 4 Information to compare.
  • the preparation method of the visible spectrum sensing area of the present invention is: selecting 8 kinds of polyimide filter film materials with different spectral transmittances, first coating the first filter film material on the photoelectric conversion substrate, and then coating Covered with an etching layer, according to the corresponding relationship with the photoelectric conversion substrate pixels, the necessary places are reserved, and the unneeded places are etched away; then the second filter film material is applied, and then an etching layer is applied According to the corresponding relationship with the pixels of the photoelectric conversion substrate, the necessary places are kept, and the unnecessary places are etched away; cycle in turn until all n kinds of filter film materials are coated on the photoelectric conversion substrate, the above 8 kinds After the film materials are coated and etched one by one, a complete film is finally formed, and each film includes T 1 , T 2 ... T 8 subunits; preparation of the near-infrared spectrum image sensing area
  • the method is the same as the preparation method of the visible spectrum sensing area, and the near-infrared spectrum image sensing area is coated
  • Embodiment 2 A recognition method of a face recognition monitoring system based on spectrum and multi-band fusion
  • the present invention provides a recognition method for a face recognition monitoring system based on spectrum and multi-band fusion, including the following steps:
  • Step S1 When a person under test enters the monitoring area, the spectral camera locks the face target, and the optical lens automatically focuses to make the face of the inspector clearly imaged.
  • the spectral camera locks multiple face targets at the same time.
  • the optical lens automatically focuses, so that the faces of multiple inspectors can be clearly imaged.
  • the supplementary light source is turned on;
  • Step S2 Start the silicon-based detector and the face spectrum image acquisition module, the visible spectrum sensing area and the near-infrared spectrum image sensing area on the silicon-based detector collect the visible light and near-infrared spectrum data reflected by the target person’s face,
  • the near-infrared spectral image sensing area collects face image information in the near-infrared band at the same time, and the RGGB image acquisition area collects face color image information.
  • the time used for multi-spectral data in the infrared band is 1ms;
  • Step S3 Compare the spectral data of the 10 pixels obtained in step S2 with the facial spectral data in the facial spectral image database.
  • the multi-spectral data obtained from the visible spectral sensing area and the near-infrared spectral image sensing area are compared with
  • the living body tissue in the face spectrum image database is compared with the multi-spectral data of various mask materials.
  • the database face spectrum and the silicone mask spectrum are shown in Figure 7 and Figure 8.
  • the square line spectrum in the figure is the linearity of the living skin spectrum.
  • the circular spectral line is the linearity of the silicone mask spectrum.
  • the comparison method is to set the discrimination threshold in advance by the system, and then calculate the discrimination between the face spectrum and the face spectrum in the database and the silicone mask spectrum.
  • step S4 When the face spectrum is compared with the face in the database When the spectral discrimination is less than or equal to the threshold, the comparison result appears to match the living tissue, then step S4; when the face spectrum and the silica gel mask spectral discrimination in the database are less than or equal to the threshold, the comparison result appears to match the mask material, Then both the alarms in the monitoring area and the police sounded, and the detection ends; if there is no matching result between the two, it is possible that the person under test may change the facial features through makeup, then proceed to step S4';
  • Step S4 After the comparison result in step S3 shows the living body tissue, the face spectral image data preprocessing module performs noise reduction processing on the 5 face color images acquired in the RGGB image acquisition area during the dynamic process of the person, and after noise reduction Perform feature extraction on the image to obtain facial feature data.
  • the face spectral image recognition module compares the facial feature data with the facial image feature values in the face spectral image database. If the comparison result does not match the data, it will delete all the facial features.
  • the acquired RGGB face image if there is matching data, an alarm with the police in the monitoring area will sound and the detection will end;
  • Step S4' In step S3, the comparison result shows non-living tissues.
  • the face spectral image data preprocessing module performs noise reduction processing on the 5 near-infrared face images acquired during the dynamic process of the person. Feature extraction is performed on the noisy image to obtain facial feature data.
  • the face spectral image recognition module compares the facial feature data with the facial image feature values in the face spectral image database. If there is no matching data in the comparison result, then Delete the acquired face images in the near-infrared band; if there is matching data, both the alarms in the surveillance area and the police will sound, and the detection will end.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Toxicology (AREA)
  • Studio Devices (AREA)
  • Image Processing (AREA)

Abstract

本发明涉及一种基于光谱与多波段融合的人脸识别监控系统及识别方法，该系统包括光谱摄像头、人脸光谱图像采集模块、人脸光谱图像数据预处理模块、人脸光谱图像数据库、人脸光谱图像识别模块；所述光谱摄像头包括光学镜头和硅基探测器，所述硅基探测器包括光电转换基底、设置在光电转换基底上面的滤光薄膜；所述滤光薄膜包括N个单元，每个单元包括三个区域，具体为可见光谱感应区、近红外光谱图像感应区和RGGB图像采集区，N个单元覆盖光电转换基底上的所有像素点；该系统实现通过一个摄像头即可获取人脸在可见-近红外光波段的反射多光谱数据、近红外波段人脸图像和RGGB人脸图像的目的，大大降低监控系统的成本，实现了人脸更精准的识别。

Description

基于光谱与多波段融合的人脸识别监控系统及识别方法 技术领域
本发明属于光学领域，涉及人脸识别与监控相关技术领域，具体涉及一种基于光谱与多波段融合的人脸识别监控系统及识别方法。
背景技术
视频监控在安全领域一直扮演着重要的角色。随着科技的发展，人脸识别技术以最为友好的生物特征认证技术，通过摄像头动态捕捉人面部图像，再进行比对识别的方法进入了监控领域。人脸识别技术的核心在于，利用面部各器官及特征部位的方位关系或人体姿态骨架特征，形成识别参数，其与模板数据库中的参数进行识别计算，最终给出判断，例如：CN 110796079 A公开的“基于人脸深度特征和人体局部深度特征的多相机访客识别的方法及系统”。但随着一些高清3D面具出现和通过化妆更改面容特征等不法欺诈手段的出现(这里我们统称真假人脸)，使这种生物识别技术失效，安全性面临了巨大的挑战。即使监控系统嵌入了人脸识别活体检测技术，通过抬头、低头、或念一串文字等面部动作，来判定检测对象是否为活体，也无法做到对视频中真假人脸的实时智能分析。另外，现有的RGGB摄像头成像很大程度上受光线环境的影响，在光线条件恶劣的情况下无法实现人脸、活体检测。
中国专利CN110874588A公开了“动态优化人脸识别中光线影响的方法与装置”，提出通过设置RGGB摄像头及红外摄像头双目摄像头的方法，利用红外摄像头进行人脸识别处理，避免光线条件影响。中国专利CN108090477A公开的“一种基于多光谱融合的人脸识别方法与装置”同样提出利用RGGB图像和红外图像进行人脸识别。虽然利用红外图像这种方案能够解决光线不好的问题，但在识别真假人脸上仍然起不到有效作用，加上红外摄像头价格相对昂贵，不适于推广。
中国专利CN202887210U公开了“一种多光谱人脸识别系统”，该系统利用多光谱成像系统获取生物体的光谱特征信息，利用人脸皮肤的光谱反射率识别真假人脸，避免人脸识别过程中的一些伪装欺骗手段。但是该识别系统所用的高精度光谱仪价格昂贵，仍不适于推广。因此，有必要开发一种价格低廉、能够有效进行真假人脸识别，且识别准确的人脸识别监控系统。
发明内容
鉴于上述问题，本发明的第一个目的在于提供一种基于光谱与多波段融合的人脸识别监控系统，该系统通过将硅基探测器分成多个单元，每个单元有三个区域，每个区域上分别刻蚀不同波段的光谱调制膜层，实现通过一个摄像头即可获取人脸在可见-近红外光波段的反射多光谱数据、近红外波段人脸图像和RGGB人脸图像的目的，大大降低监控系统的成本；而且该系统在对人脸进行识别时，可利用光谱数据识别是否为真实人体皮肤，再结合近红外波段人脸图像和RGGB人脸图像进一步识别，双重识别有效保证了人脸识别的准确性，提高识别安全性，解决了目前监控系统人脸识别错判的弊端。
为实现上述目的，本发明具体是采用如下技术方案实现的：
一种基于光谱与多波段融合的人脸识别监控系统，该系统包括光谱摄像头、人脸光谱图像采集模块、人脸光谱图像数据预处理模块、人脸光谱图像数据库、人脸光谱图像识别模块；
其中，所述光谱摄像头包括光学镜头和硅基探测器，所述硅基探测器包括光电转换基底、设置在光电转换基底上面的滤光薄膜；所述滤光薄膜包括N个单元，每个单元包括三个区域，具体为可见光谱感应区、近红外光谱图像感应区和RGGB图像采集区，N个单元覆盖光电转换基底上的所有像素点；
所述可见光谱感应区，用于将入射的可见光光谱进行区分；可见光谱感应区是由已知且透光率不同的n种材料通过逐一涂覆、刻蚀后拼接而成，可见光谱感应区为单层结构，包括T 1、T 2......T n个子单元，每个子单元覆盖光电转换基底上的M个像素，其中M大于等于1，可见光谱感应区与近红外光谱图像感应区和RGGB图像采集区构成一个周期性结构；
所述近红外光谱图像感应区的透射波段分为780-850nm、850-950nm、950-1050nm、1050-1200nm，用于探测波长大于780nm的近红外光波段的深层面部特征图像，其是由已知且近红外透光率不同的四种材料通过逐一涂覆、刻蚀后拼接而成，近红外光谱图像感应区为单层结构，包括Tir1、Tir2、Tir3、Tir4四个子单元，每个子单元覆盖光电转换基底上的M个像素，其中M大于等于1；
所述RGGB图像采集区，用于监控区人脸彩色图像采集，RGGB图像采集区包括透射红、绿、绿、蓝三种颜色的四个子单元，每个子单元覆盖光电转换基 底上的M个像素，其中M大于等于1；
所述光电转换基底用于将透过可见光谱感应区的光信号转换为电信号，经过放大及模数转换后形成数字信号或者编码输出；将透过近红外光谱图像感应区的光信号转换为电信号，经过放大及模数转换后形成数字信号或者编码输出；将透过RGGB图像采集区的光信号转换为电信号，经过放大及模数转换后形成数字信号或者编码输出；
所述人脸光谱图像采集模块，用于根据入射光信号强度信息和对应光电转换基底像素位置信息，反演出人脸光谱数据和图像数据，包括人脸在可见-近红外波段的反射多光谱数据、近红外波段人脸图像和RGGB人脸图像；所述近红外波段人脸图像和RGGB人脸图像采用图像反演方式进行获取，所述多光谱数据的反演方式如公式(1)所示
S i＝∫I(λ)T i(λ)η(λ)d，   (1)
其中，S为光电转换基底输出的光信号强度值，I为入射光谱，是待求解信号，T为滤光薄膜的光谱透过率，包括可见光波段的滤光薄膜的光谱透过率和近红外光波段的滤光薄膜的光谱透过率，η为光电转换基底的量子效率，λ为入射波长；
所述人脸光谱图像数据预处理模块，用于将近红外光谱图像感应区和RGGB图像采集区获取的人脸图像进行降噪处理，并将降噪后的RGGB人脸图像和近红外波段人脸图像进行特征提取，在可见光谱感应区获取的多光谱图像以及在近红外光谱感应区获取的多光谱图像中，对人脸的多光谱数据进行提取，将多维的人面部特征数据与人脸光谱图像数据库中图像及多光谱数据进行比对；
所述人脸光谱图像数据库，用于储存所要搜寻的N x个特征人脸的图像特征数值和光谱数据，以及多种假面具材料多光谱数据，便于与监控区域内出现的人脸图像及其光谱比对；
所述人脸光谱图像识别模块，用于将人脸可见-近红外波段的反射多光谱数据、预处理后的近红外波段人脸图像和RGGB人脸图像数据与人脸光谱图像数据库中的信息进行对比。
作为本发明的优选，该系统还包括补光光源，所述补光光源发射波段包括可 见光和近红外光，用于夜晚或光线不足环境下对监控区的人员进行监控，加强深层面部图像采集。
作为本发明的优选，所述光学镜头在监控区域内可根据目标距离自动调节焦距。
作为本发明的优选，所述硅基探测器为CMOS图像传感器或CCD图像传感器。
作为本发明的优选，所述可见光谱感应区的制备方法为：选择n种光谱透过率不同的聚酰亚胺类滤光薄膜材料，先在光电转换基底上涂覆第一种滤光薄膜材料，再涂覆一层刻蚀层，根据与光电转换基底像素的对应关系，将需要的地方保留，将不需要的地方刻蚀掉；之后涂覆第二种滤光薄膜材料，再涂覆一层刻蚀层，根据与光电转换基底像素的对应关系，将需要的地方保留，将不需要的地方刻蚀掉；依次循环，直至将n种滤光薄膜材料全部涂覆到光电转换基底上；所述近红外光谱图像感应区的制备方法与可见光谱感应区的制备方法相同，近红外光谱图像感应区涂覆四种滤光薄膜材料。
作为本发明的优选，所述滤光薄膜上制备有微透镜阵列，所述微透镜阵列是在滤光薄膜上面涂覆一层用于制备微透镜阵列的透明光刻胶，然后采用激光直写或者掩膜光刻的方法刻蚀不需要的部分，保留下来的部分便构成了微透镜阵列，微透镜阵列上的每个微透镜与光电转换基底的像素一一对应，用于将入射光光束进行汇聚。
本发明的第二个目的在于提供一种基于光谱与多波段融合的人脸识别监控系统的识别方法，包括以下步骤：
步骤S1：待测人员进入监控区域内，光谱摄像头的光学镜头自动对焦，使检测人员面部清晰成像；当待测人员为多人时，光谱摄像头同时锁定多个人脸目标，光学镜头自动对焦，使多名检测人员面部清晰成像，当光线不足导致不能清晰成像时，补光光源开启；
步骤S2：启动硅基探测器和人脸光谱图像采集模块，硅基探测器上的可见光谱感应区和近红外光谱图像感应区采集锁定目标人物面部的反射多光谱数据，近红外光谱图像感应区采集近红外波段人脸图像信息，RGGB图像采集区采集人脸彩色图像信息；
步骤S3：将步骤S2获得的多光谱数据与人脸光谱图像数据库中的人脸光谱数据进行比对，首先将可见光谱感应区和近红外光谱图像感应区获得的多光谱数据与人脸光谱图像数据库中的活体组织和多种假面具材料多光谱数据进行比对，比对结果呈现与活体组织匹配，则进行步骤S4；若比对结果呈现与假面具材料匹配，则监控区域内与警方的警报均响起，检测结束；若二者无匹配结果，则可能待测人员通过化妆更改面容特征，则进行步骤S4’；
步骤S4：步骤S3中比对结果呈现出活体组织后，人脸光谱图像数据预处理模块将人员动态过程中RGGB图像采集区获取的N张人脸彩色图像进行图像进行降噪处理，并将降噪后的图像进行特征提取，获得面部特征数据，人脸光谱图像识别模块将面部特征数据与人脸光谱图像数据库中的人脸图像特征数值进行比对，若比对结果无匹配数据时，则删除所获取的RGGB人脸图像；若有匹配数据，则监控区域内与警方的警报响起，检测结束；
步骤S4’：步骤S3中比对结果呈现出非活体组织者，人脸光谱图像数据预处理模块将人员动态过程中获取的N张近红外波段人脸图像进行图像进行降噪处理，并将降噪后的图像进行特征提取，获得面部特征数据，人脸光谱图像识别模块将面部特征数据与人脸光谱图像数据库中的人脸图像特征数值进行比对，若比对结果无匹配数据时，则删除所获取的近红外波段人脸图像；若有匹配数据，则监控区域内与警方的警报均响起，检测结束。
本发明的优点及积极效果是：
1、本发明提供的人脸识别监控系统通过将硅基探测器分成多个单元，每个单元有三个区域，每个区域上分别刻蚀不同波段的光谱调制膜层，实现通过一个摄像头即可获取人脸在可见-近红外光波段的反射多光谱数据、近红外波段人脸图像和RGGB人脸图像的目的，大大降低监控系统的成本。
2、本发明提供的人脸识别监控系统综合可见-近红外光波段光谱数据、近红外光波段图像和RGGB图像对人脸进行识别，来判断人脸的真假，并排除通过化妆等手段遮掩真实面容，实现了人脸更精准的识别，大大提高了监控预防、预警的准确性，有效避免不法分子用硅胶等材料仿造人脸面具蒙混过关。
3、本发明提供的人脸识别监控系统中硅基探测器的可见光谱调制膜层、近红外光谱调制膜层采用的是单层结构，构成的光电传感器结构简单、体积小、厚 度薄(为微米量级)、重量轻、具有较高光谱分辨率的同时具有空间分辨率、精准度高、检测速度快，可集成在现有任何监控装置内，即可实现对光谱的提取，还可实现高精度的成像功能，使提取的人脸更加清晰，准确率更高。
4、本发明综合图像和光谱信息的人脸真假模式识别方法，利用现有的硅基探测器镀膜的方法，实现了人脸更精准的识别，通过光电传感器获得能够反应出人体皮肤特有的反射光谱信息和人脸图像信息，利用比对方法构成低成本、超便捷式的人脸识别系统。
附图说明
图1为本发明真假人脸识别监控系统的原理图。
图2为本发明滤光薄膜单元的结构示意图。
图3为本发明滤光薄膜的结构示意图。
图4为本发明硅基光电传感器的结构示意图。
图5为本发明硅基探测器的结构示意图。
图6为本发明真假人脸识别方法示意图。
图7为本发明可见光波段活体组织和硅胶面具的多光谱数据示意图。
图8为本发明近红外光波段活体组织和硅胶面具的多光谱数据示意图。
具体实施方式
在下面的描述中，出于说明的目的，为了提供对一个或多个实施例的全面理解，阐述了许多具体细节。然而，很明显，也可以在没有这些具体细节的情况下实现这些实施例。在其它例子中，为了便于描述一个或多个实施例，公知的结构和设备以方框图的形式示出。
实施例1基于光谱与多波段融合的人脸识别监控系统
参阅图1，本发明提供的一种基于光谱与多波段融合的人脸识别监控系统，该系统包括光谱摄像头1、补光光源(宽波段白光光源，可以是溴钨灯)、人脸光谱图像采集模块2、人脸光谱图像数据预处理模块3、人脸光谱图像数据库4、人脸光谱图像识别模块5；
其中，所述光谱摄像头包括光学镜头和硅基探测器，光学镜头可在监控区域内根据目标距离自动调节焦距，硅基探测器选用类型为CMOS图像传感器，探测波段为200～1200nm，硅基探测器包括光电转换基底101、设置在光电转换基 底上面的滤光薄膜102，所述滤光薄膜102膜层厚度为10μm，其大小为4mm×4mm，数据采集时间为1ms；滤光薄膜102包括N个单元，每个单元包括三个区域，具体为可见光谱感应区6、近红外光谱图像感应区7和RGGB图像采集区8，N个单元覆盖光电转换基底上的所有像素点，如图2、图3所示，通过三个传感区域实现人脸可见-近红外光波段反射多光谱数据、近红外波段人脸图像和RGGB人脸图像的采集；
所述可见光谱感应区6，用于将入射的可见光光谱进行区分；可见光谱感应区是由已知且透光率不同的8种聚酰亚胺类材料通过逐一涂覆、刻蚀后拼接而成，可见光谱感应区为单层结构，包括T 1、T 2......T 8个子单元，每个子单元覆盖光电转换基底上的1个像素，每个子单元中针对光电转换基底像元(像素)具有不同的光谱透过率，其光谱透过率发生变化，则每个子单元中不同像素上的光谱调制特性不同，组合起来可以复原并反演入射光的光谱线型，实现光谱分光功能；可见光谱感应区与近红外光谱图像感应区和RGGB图像采集区构成一个周期性结构；
所述近红外光谱图像感应区7的透射波段分为780-850nm、850-950nm、950-1050nm、1050-1200nm，用于探测波长大于780nm的近红外光波段的深层面部特征图像，排除通过化妆更改面容特征为系统带来的干扰；近红外光谱图像感应区是由已知且近红外透光率不同的四种材料通过逐一涂覆、刻蚀后拼接而成，近红外光谱图像感应区为单层结构，包括Tir1、Tir2、Tir3、Tir4四个子单元，每个子单元覆盖光电转换基底上的1个像素；
所述RGGB图像采集区8，用于监控区人脸彩色图像采集；RGGB图像采集区包括透射红、绿、绿、蓝三种颜色的四个子单元，每个子单元覆盖光电转换基底上的1个像素；
如图5所示，为提高入射光能量的利用率以及传感器的开口率，所述滤光薄膜102上设置有微透镜阵列103，所述微透镜阵列103是在滤光薄膜102上面涂覆一层用于制备微透镜阵列的透明光刻胶(环氧树脂材料)，然后采用激光直写或者掩膜光刻的方法刻蚀不需要的部分，保留下来的部分便构成了微透镜阵列103，微透镜阵列上的每个微透镜与光电转换基底的像素一一对应，用于将入射光光束进行汇聚，微透镜阵列103和滤光薄膜102总厚度为30μm，其大小为 4mm×4mm，光谱分辨率为10nm，数据采集时间为1ms；
所述光电转换基底用于将透过可见光谱感应区6的光信号转换为电信号，经过放大及模数转换后形成数字信号或者编码输出；将透过近红外光谱图像感应区7的光信号转换为电信号，经过放大及模数转换后形成数字信号或者编码输出；将透过RGGB图像采集区8的光信号转换为电信号，经过放大及模数转换后形成数字信号或者编码输出；
所述补光光源，其光源发射波段包括可见光和近红外光，用于夜晚或光线不足环境下对监控区的人员进行监控和加强深层面部图像采集；
所述人脸光谱图像采集模块，用于根据入射光信号强度信息和对应光电转换基底像素位置信息，反演出人脸光谱数据和图像数据，包括人脸在可见-近红外波段反射的多光谱数据、近红外波段人脸图像和RGGB人脸图像；所述近红外波段人脸图像和RGGB人脸图像采用图像反演方式进行获取(根据每个像素上的光谱对应已知的光谱透过率信息，修正对应像素上的光信号强度值，修正方法为该像素上的光信号强度值除以该像素上的光谱透过率值；结合所有像素的组合，即可反演出图像信息)；所述多光谱数据的反演方式如公式(1)所示
S i＝∫I(λ)T i(λ)η(λ)d，   (1)
其中，S为光电转换基底输出的光信号强度值，I为入射光谱，是待求解信号，T为滤光薄膜的光谱透过率，包括可见光波段的滤光薄膜的光谱透过率T 1～T 8和近红外光波段的滤光薄膜的光谱透过率T ir1～T ir4，η为光电转换基底的量子效率，λ为入射波长；
根据光谱透过率曲线，可见光谱感应区的8个像素、近红外光谱图像感应区的4个像素组合，可反演计算出可见光谱感应区上8个像素的可见入射光谱值和近红外光谱图像感应区上4个像素的近红外光谱值；
所述人脸光谱图像数据预处理模块3，用于将近红外光谱图像感应区和RGGB图像采集区获取的人脸图像进行降噪处理(短时间内拍摄多张图像，按像素点将多张图像取平均，例如拍摄10张图，每张图的第一个像素取平均，第二个像素取平均….第i个像素取平均，完成降噪处理)，并将降噪后的RGGB人脸图像和近红外波段人脸图像进行特征提取，在可见光谱感应区获取的多光谱图 像以及在近红外光谱感应区获取的多光谱图像中，对人脸的多光谱数据进行提取，将多维的人面部特征数据与人脸光谱图像数据库4中图像及多光谱数据进行比对；
所述人脸光谱图像数据库4，用于储存所要搜寻的N x个特征人脸的图像特征数值和光谱数据，以及多种假面具材料多光谱数据，便于与监控区域内出现的人脸图像及其光谱比对；
所述人脸光谱图像识别模块5，用于将人脸可见-近红外波段的反射多光谱数据、预处理后的近红外波段人脸图像和RGGB人脸图像数据与人脸光谱图像数据库4中的信息进行对比。
本发明所述可见光谱感应区的制备方法为：选择8种光谱透过率不同的聚酰亚胺类滤光薄膜材料，先在光电转换基底上涂覆第一种滤光薄膜材料，再涂覆一层刻蚀层，根据与光电转换基底像素的对应关系，将需要的地方保留，将不需要的地方刻蚀掉；之后涂覆第二种滤光薄膜材料，再涂覆一层刻蚀层，根据与光电转换基底像素的对应关系，将需要的地方保留，将不需要的地方刻蚀掉；依次循环，直至将n种滤光薄膜材料全部涂覆到光电转换基底上，上述8种薄膜材料经过逐一的涂覆和刻蚀后，最后形成一层完整的薄膜，每个薄膜包括T 1、T 2......T 8个子单元；所述近红外光谱图像感应区的制备方法与可见光谱感应区的制备方法相同，近红外光谱图像感应区涂覆四种滤光薄膜材料。
实施例2一种基于光谱与多波段融合的人脸识别监控系统的识别方法
如图6所示，本发明提供的一种基于光谱与多波段融合的人脸识别监控系统的识别方法，包括以下步骤：
步骤S1：1名待测人员进入监控区域内，光谱摄像头锁定人脸目标，光学镜头自动对焦，使检测人员面部清晰成像，当待测人员为多人时，光谱摄像头同时锁定多个人脸目标，光学镜头自动对焦，使多名检测人员面部清晰成像，当光线不足导致不能清晰成像时，补光光源开启；
步骤S2：启动硅基探测器和人脸光谱图像采集模块，硅基探测器上的可见光谱感应区和近红外光谱图像感应区采集锁定目标人物面部反射的可见光和近红外波段的多光谱数据，近红外光谱图像感应区同时采集近红外波段人脸图像信息，RGGB图像采集区采集人脸彩色图像信息；多光谱数据采集方法为在人脸轮 廓内10个像素点进行光谱采集，采集可见光以及近红外波段的多光谱数据所用时间为1ms；
步骤S3：将步骤S2获得的10个像素点的光谱数据与人脸光谱图像数据库中的人脸光谱数据进行比对，首先将可见光谱感应区和近红外光谱图像感应区获得的多光谱数据与人脸光谱图像数据库中的活体组织和多种假面具材料多光谱数据进行比对，数据库人脸光谱与硅胶面具光谱如图7、图8所示，图中方块线形谱线为活体皮肤光谱线性，圆线形谱线为硅胶面具光谱线性，比对方法为系统提前设定区分度阈值，然后计算人脸光谱与数据库中人脸光谱与硅胶面具光谱的区分度，当人脸光谱与数据库中人脸光谱区分度小于或等于阈值时，比对结果呈现与活体组织匹配，则进行步骤S4；当人脸光谱与数据库中硅胶面具光谱区分度小于或等于阈值时，比对结果呈现与假面具材料匹配，则监控区域内与警方的警报均响起，检测结束；若二者无匹配结果，则可能待测人员通过化妆更改面容特征，则进行步骤S4’；
步骤S4：步骤S3中比对结果呈现出活体组织后，人脸光谱图像数据预处理模块将人员动态过程中RGGB图像采集区获取的5张人脸彩色图像进行降噪处理，并将降噪后的图像进行特征提取，获得面部特征数据，人脸光谱图像识别模块将面部特征数据与人脸光谱图像数据库中的人脸图像特征数值进行比对，若比对结果无匹配数据时，则删除所获取的RGGB人脸图像；若有匹配数据，则监控区域内与警方的警报响起，检测结束；
步骤S4’：步骤S3中比对结果呈现出非活体组织者，人脸光谱图像数据预处理模块将人员动态过程中获取的5张近红外波段人脸图像进行图像进行降噪处理，并将降噪后的图像进行特征提取，获得面部特征数据，人脸光谱图像识别模块将面部特征数据与人脸光谱图像数据库中的人脸图像特征数值进行比对，若比对结果无匹配数据时，则删除所获取的近红外波段人脸图像；若有匹配数据，则监控区域内与警方的警报均响起，检测结束。

Claims (7)

  1. 一种基于光谱与多波段融合的人脸识别监控系统，其特征在于，该系统包括光谱摄像头、人脸光谱图像采集模块、人脸光谱图像数据预处理模块、人脸光谱图像数据库、人脸光谱图像识别模块；
    其中，所述光谱摄像头包括光学镜头和硅基探测器，所述硅基探测器包括光电转换基底、设置在光电转换基底上面的滤光薄膜；所述滤光薄膜包括N个单元，每个单元包括三个区域，具体为可见光谱感应区、近红外光谱图像感应区和RGGB图像采集区，N个单元覆盖光电转换基底上的所有像素点；
    所述可见光谱感应区，用于将入射的可见光光谱进行区分；可见光谱感应区是由已知且透光率不同的n种材料通过逐一涂覆、刻蚀后拼接而成，可见光谱感应区为单层结构，包括T 1、T 2......T n个子单元，每个子单元覆盖光电转换基底上的M个像素，其中M大于等于1，可见光谱感应区与近红外光谱图像感应区和RGGB图像采集区构成一个周期性结构；
    所述近红外光谱图像感应区的透射波段分为780-850nm、850-950nm、950-1050nm、1050-1200nm，用于探测波长大于780nm的近红外光波段的深层面部特征图像，其是由已知且近红外透光率不同的四种材料通过逐一涂覆、刻蚀后拼接而成，近红外光谱图像感应区为单层结构，包括Tir1、Tir2、Tir3、Tir4四个子单元，每个子单元覆盖光电转换基底上的M个像素，其中M大于等于1；
    所述RGGB图像采集区，用于监控区人脸彩色图像采集，RGGB图像采集区包括透射红、绿、绿、蓝三种颜色的四个子单元，每个子单元覆盖光电转换基底上的M个像素，其中M大于等于1；
    所述光电转换基底用于将透过可见光谱感应区的光信号转换为电信号，经过放大及模数转换后形成数字信号或者编码输出；将透过近红外光谱图像感应区的光信号转换为电信号，经过放大及模数转换后形成数字信号或者编码输出；将透过RGGB图像采集区的光信号转换为电信号，经过放大及模数转换后形成数字信号或者编码输出；
    所述人脸光谱图像采集模块，用于根据入射光信号强度信息和对应光电转换基底像素位置信息，反演出人脸光谱数据和图像数据，包括人脸在可见-近红外波段的反射多光谱数据、近红外波段人脸图像和RGGB人脸图像；所述近红外波段人脸图像和RGGB人脸图像采用图像反演方式进行获取，所述多光谱数据 的反演方式如公式(1)所示
    S i＝∫I(λ)T i(λ)η(λ)dλ，  (1)
    其中，S为光电转换基底输出的光信号强度值，I为入射光谱，是待求解信号，T为滤光薄膜的光谱透过率，包括可见光波段的滤光薄膜的光谱透过率和近红外光波段的滤光薄膜的光谱透过率，η为光电转换基底的量子效率，λ为入射波长；
    所述人脸光谱图像数据预处理模块，用于将近红外光谱图像感应区和RGGB图像采集区获取的人脸图像进行降噪处理，并将降噪后的RGGB人脸图像和近红外波段人脸图像进行特征提取，在可见光谱感应区获取的多光谱图像以及在近红外光谱感应区获取的多光谱图像中，对人脸的多光谱数据进行提取，将多维的人面部特征数据与人脸光谱图像数据库中图像及多光谱数据进行比对；
    所述人脸光谱图像数据库，用于储存所要搜寻的N x个特征人脸的图像特征数值和光谱数据，以及多种假面具材料多光谱数据，便于与监控区域内出现的人脸图像及其光谱比对；
    所述人脸光谱图像识别模块，用于将人脸可见-近红外波段的反射多光谱数据、预处理后的近红外波段人脸图像和RGGB人脸图像数据与人脸光谱图像数据库中的信息进行对比。
  2. 根据权利要求1所述的一种基于光谱与多波段融合的人脸识别监控系统，其特征在于，该系统还包括补光光源，所述补光光源发射波段包括可见光和近红外光。
  3. 根据权利要求1所述的一种基于光谱与多波段融合的人脸识别监控系统，其特征在于，所述光学镜头在监控区域内可根据目标距离自动调节焦距。
  4. 根据权利要求1所述的一种基于光谱与多波段融合的人脸识别监控系统，其特征在于，所述硅基探测器为CMOS图像传感器或CCD图像传感器。
  5. 根据权利要求1所述的一种基于光谱与多波段融合的人脸识别监控系统，其特征在于，所述可见光谱感应区的制备方法为：选择n种光谱透过率不同的聚酰亚胺类滤光薄膜材料，先在光电转换基底上涂覆第一种滤光薄膜材料，再涂覆一层刻蚀层，根据与光电转换基底像素的对应关系，将需要的地方保留，将不需 要的地方刻蚀掉；之后涂覆第二种滤光薄膜材料，再涂覆一层刻蚀层，根据与光电转换基底像素的对应关系，将需要的地方保留，将不需要的地方刻蚀掉；依次循环，直至将n种滤光薄膜材料全部涂覆到光电转换基底上；所述近红外光谱图像感应区的制备方法与可见光谱感应区的制备方法相同，近红外光谱图像感应区涂覆四种滤光薄膜材料。
  6. 根据权利要求1所述的一种基于光谱与多波段融合的人脸识别监控系统，其特征在于，所述滤光薄膜上制备有微透镜阵列，所述微透镜阵列是在滤光薄膜上面涂覆一层用于制备微透镜阵列的透明光刻胶，然后采用激光直写或者掩膜光刻的方法刻蚀不需要的部分，保留下来的部分便构成了微透镜阵列，微透镜阵列上的每个微透镜与光电转换基底的像素一一对应，用于将入射光光束进行汇聚。
  7. 权利要求1所述的一种基于光谱与多波段融合的人脸识别监控系统的识别方法，其特征在于，包括以下步骤：
    步骤S1：待测人员进入监控区域内，光谱摄像头的光学镜头自动对焦，使检测人员面部清晰成像；当待测人员为多人时，光谱摄像头同时锁定多个人脸目标，光学镜头自动对焦，使多名检测人员面部清晰成像，当光线不足导致不能清晰成像时，补光光源开启；
    步骤S2：启动硅基探测器和人脸光谱图像采集模块，硅基探测器上的可见光谱感应区和近红外光谱图像感应区采集锁定目标人物面部的反射多光谱数据，近红外光谱图像感应区采集近红外波段人脸图像信息，RGGB图像采集区采集人脸彩色图像信息；
    步骤S3：将步骤S2获得的多光谱数据与人脸光谱图像数据库中的人脸光谱数据进行比对，首先将可见光谱感应区和近红外光谱图像感应区获得的多光谱数据与人脸光谱图像数据库中的活体组织和多种假面具材料多光谱数据进行比对，比对结果呈现与活体组织匹配，则进行步骤S4；若比对结果呈现与假面具材料匹配，则监控区域内与警方的警报均响起，检测结束；若二者无匹配结果，则可能待测人员通过化妆更改面容特征，则进行步骤S4’；
    步骤S4：步骤S3中比对结果呈现出活体组织后，人脸光谱图像数据预处理模块将人员动态过程中RGGB图像采集区获取的N张人脸彩色图像进行图像进行降噪处理，并将降噪后的图像进行特征提取，获得面部特征数据，人脸光谱图 像识别模块将面部特征数据与人脸光谱图像数据库中的人脸图像特征数值进行比对，若比对结果无匹配数据时，则删除所获取的RGGB人脸图像；若有匹配数据，则监控区域内与警方的警报响起，检测结束；
    步骤S4’：步骤S3中比对结果呈现出非活体组织者，人脸光谱图像数据预处理模块将人员动态过程中获取的N张近红外波段人脸图像进行图像进行降噪处理，并将降噪后的图像进行特征提取，获得面部特征数据，人脸光谱图像识别模块将面部特征数据与人脸光谱图像数据库中的人脸图像特征数值进行比对，若比对结果无匹配数据时，则删除所获取的近红外波段人脸图像；若有匹配数据，则监控区域内与警方的警报均响起，检测结束。
PCT/CN2021/096673 2020-06-05 2021-05-28 基于光谱与多波段融合的人脸识别监控系统及识别方法 WO2021244414A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/555,996 US11443550B2 (en) 2020-06-05 2021-12-20 Face recognition monitoring system based on spectrum and multi-band fusion and recognition method using same

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010503421.1A CN111611977B (zh) 2020-06-05 2020-06-05 基于光谱与多波段融合的人脸识别监控系统及识别方法
CN202010503421.1 2020-06-05

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/555,996 Continuation US11443550B2 (en) 2020-06-05 2021-12-20 Face recognition monitoring system based on spectrum and multi-band fusion and recognition method using same

Publications (1)

Publication Number Publication Date
WO2021244414A1 true WO2021244414A1 (zh) 2021-12-09

Family

ID=72200293

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/096673 WO2021244414A1 (zh) 2020-06-05 2021-05-28 基于光谱与多波段融合的人脸识别监控系统及识别方法

Country Status (3)

Country Link
US (1) US11443550B2 (zh)
CN (1) CN111611977B (zh)
WO (1) WO2021244414A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117456584A (zh) * 2023-11-13 2024-01-26 江苏创斯达智能科技有限公司 一种应用于智能保险柜的面部识别设备

Families Citing this family (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111611977B (zh) 2020-06-05 2021-10-15 吉林求是光谱数据科技有限公司 基于光谱与多波段融合的人脸识别监控系统及识别方法
CN111879724B (zh) * 2020-08-05 2021-05-04 中国工程物理研究院流体物理研究所 基于近红外光谱成像的人皮面具识别方法及系统
CN112115838B (zh) * 2020-09-11 2024-04-05 南京华图信息技术有限公司 一种热红外图像光谱融合的人脸分类方法
CN112669817B (zh) * 2020-12-25 2023-08-08 维沃移动通信有限公司 语言识别方法、装置及电子设备
CN112986179A (zh) * 2021-02-28 2021-06-18 深圳盈达机器视觉技术有限公司 电子装置及物品信息检测方法
CN115147878A (zh) * 2021-03-15 2022-10-04 北京与光科技有限公司 对象识别方法、装置和电子设备和指纹识别模组
CN113011391B (zh) * 2021-04-25 2024-04-26 秦皇岛本征晶体科技有限公司 一种用于人脸识别活体检测的光学系统及方法
US11636700B2 (en) 2021-05-21 2023-04-25 Ford Global Technologies, Llc Camera identification
US11967184B2 (en) * 2021-05-21 2024-04-23 Ford Global Technologies, Llc Counterfeit image detection
US11769313B2 (en) 2021-05-21 2023-09-26 Ford Global Technologies, Llc Counterfeit image detection
CN113566966B (zh) * 2021-06-03 2023-07-04 奥比中光科技集团股份有限公司 一种多光谱图像传感器的制造方法及其设备
CN113743343A (zh) * 2021-09-10 2021-12-03 维沃移动通信有限公司 图像信息采集模组、信息处理方法、装置和电子设备
CN113959346A (zh) * 2021-10-18 2022-01-21 苏州多感科技有限公司 位移检测模块和移动设备
CN114650360B (zh) * 2022-03-31 2023-01-17 联想(北京)有限公司 电子设备及其成像方法
CN114973490A (zh) * 2022-05-26 2022-08-30 南京大学 一种基于人脸识别的监控预警系统
CN114964376B (zh) * 2022-05-31 2023-10-20 西安工业大学 一种基于分布式光学成像的多参量测量装置及测量方法
CN114942072A (zh) * 2022-06-02 2022-08-26 广州睿芯微电子有限公司 多光谱成像芯片及对象识别系统
CN115273249A (zh) * 2022-08-01 2022-11-01 思博特集成科技(珠海横琴)有限公司 光谱三维识别方法、识别方法装置及可读存储介质
US20240129604A1 (en) * 2022-10-14 2024-04-18 Motional Ad Llc Plenoptic sensor devices, systems, and methods
CN116448689B (zh) * 2023-06-16 2023-08-29 天津博霆光电技术有限公司 基于像素微分的单光频率物体全构造检测方法及设备
CN117089817B (zh) * 2023-06-29 2024-02-13 同济大学 一种基于卡尔曼滤波数据融合的光学薄膜混合监控方法
CN117392710B (zh) * 2023-12-05 2024-03-08 杭州海康威视数字技术股份有限公司 一种图像识别系统
CN117351525B (zh) * 2023-12-05 2024-03-08 杭州海康威视数字技术股份有限公司 一种图像识别系统
CN118230395B (zh) * 2024-05-13 2024-09-17 广东电网有限责任公司 一种基于InsightFace与LIS文件管理的人脸识别方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102831400A (zh) * 2012-07-31 2012-12-19 西北工业大学 一种多光谱人脸识别方法及其系统
CN202887210U (zh) * 2012-07-31 2013-04-17 西北工业大学 一种多光谱人脸识别系统
CN104881632A (zh) * 2015-04-28 2015-09-02 南京邮电大学 高光谱人脸识别方法
WO2017208233A1 (en) * 2016-06-02 2017-12-07 Fst21 Ltd. Device and method for face identification
CN110337656A (zh) * 2019-05-27 2019-10-15 深圳市汇顶科技股份有限公司 用于人脸识别的光学传感器、装置、方法和电子设备
CN210469476U (zh) * 2019-07-30 2020-05-05 上海世科嘉车辆技术研发有限公司 一种可见光和红外光双通车载人脸识别摄像头
CN111611977A (zh) * 2020-06-05 2020-09-01 吉林求是光谱数据科技有限公司 基于光谱与多波段融合的人脸识别监控系统及识别方法

Family Cites Families (33)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6920236B2 (en) * 2001-03-26 2005-07-19 Mikos, Ltd. Dual band biometric identification system
US7027619B2 (en) * 2001-09-13 2006-04-11 Honeywell International Inc. Near-infrared method and system for use in face detection
US7602942B2 (en) * 2004-11-12 2009-10-13 Honeywell International Inc. Infrared and visible fusion face recognition system
JP4702441B2 (ja) * 2008-12-05 2011-06-15 ソニー株式会社 撮像装置及び撮像方法
US8254768B2 (en) * 2010-12-22 2012-08-28 Michael Braithwaite System and method for illuminating and imaging the iris of a person
US8600227B2 (en) * 2011-09-22 2013-12-03 Xerox Corporation Multi-filter array for a multi-resolution, multi-spectral camera
US10452894B2 (en) * 2012-06-26 2019-10-22 Qualcomm Incorporated Systems and method for facial verification
JP2014078052A (ja) * 2012-10-09 2014-05-01 Sony Corp 認証装置および方法、並びにプログラム
WO2014100783A1 (en) * 2012-12-21 2014-06-26 Flir Systems, Inc. Time spaced infrared image enhancement
CN103268499B (zh) * 2013-01-23 2016-06-29 北京交通大学 基于多光谱成像的人体皮肤检测方法
WO2014176485A1 (en) * 2013-04-26 2014-10-30 West Virginia High Technology Consortium Foundation, Inc. Facial recognition method and apparatus
US9996726B2 (en) * 2013-08-02 2018-06-12 Qualcomm Incorporated Feature identification using an RGB-NIR camera pair
CN106133796B (zh) * 2014-03-25 2019-07-16 苹果公司 用于在真实环境的视图中表示虚拟对象的方法和系统
US9495584B1 (en) * 2015-06-05 2016-11-15 Digital Signal Corporation System and method for facial recognition using images captured from a target illuminated with infrared light
US10523855B2 (en) * 2015-09-24 2019-12-31 Intel Corporation Infrared and visible light dual sensor imaging system
US9886640B1 (en) * 2016-08-08 2018-02-06 International Business Machines Corporation Method and apparatus to identify a live face image using a thermal radiation sensor and a visual radiation sensor
US10796514B2 (en) * 2017-10-13 2020-10-06 Alcatraz AI, Inc. System and method for optimizing a facial recognition-based system for controlling access to a building
US10726245B2 (en) * 2017-12-12 2020-07-28 Black Sesame International Holding Limited Secure facial authentication system using active infrared light source and RGB-IR sensor
JP2019106045A (ja) * 2017-12-13 2019-06-27 キヤノン株式会社 画像処理装置、方法及びプログラム
CN108268839A (zh) * 2018-01-05 2018-07-10 北京万相融通科技股份有限公司 一种活体验证方法及其系统
CN108090477A (zh) 2018-01-23 2018-05-29 北京易智能科技有限公司 一种基于多光谱融合的人脸识别方法与装置
US10785422B2 (en) * 2018-05-29 2020-09-22 Microsoft Technology Licensing, Llc Face recognition using depth and multi-spectral camera
US20200082160A1 (en) * 2018-09-12 2020-03-12 Kneron (Taiwan) Co., Ltd. Face recognition module with artificial intelligence models
FR3091114B1 (fr) * 2018-12-19 2021-06-11 Valeo Comfort & Driving Assistance dispositif de capture d’images et système de surveillance d’un conducteur associé
US11361525B2 (en) * 2019-06-07 2022-06-14 Fotonation Limited Electronic device for performing biometric identification
CN110796079A (zh) 2019-10-29 2020-02-14 深圳龙岗智能视听研究院 基于人脸深度特征和人体局部深度特征的多相机访客识别的方法及系统
US12067052B2 (en) * 2020-01-03 2024-08-20 AlgoLook, Inc. Air particulate classification
CN110874588B (zh) 2020-01-17 2020-04-14 南京甄视智能科技有限公司 动态优化人脸识别中光线影响的方法与装置
US11475714B2 (en) * 2020-02-19 2022-10-18 Motorola Solutions, Inc. Systems and methods for detecting liveness in captured image data
US20210287469A1 (en) * 2020-03-11 2021-09-16 Alcatraz AI, Inc. System and method for provisioning a facial recognition-based system for controlling access to a building
US20210344852A1 (en) * 2020-05-04 2021-11-04 Rebellion Photonics, Inc. Apparatuses, systems, and methods for thermal imaging
US11138734B1 (en) * 2021-02-19 2021-10-05 Vr Media Technology, Inc. Hyperspectral facial analysis system and method for personalized health scoring
CN113128481A (zh) * 2021-05-19 2021-07-16 济南博观智能科技有限公司 一种人脸活体检测方法、装置、设备及存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102831400A (zh) * 2012-07-31 2012-12-19 西北工业大学 一种多光谱人脸识别方法及其系统
CN202887210U (zh) * 2012-07-31 2013-04-17 西北工业大学 一种多光谱人脸识别系统
CN104881632A (zh) * 2015-04-28 2015-09-02 南京邮电大学 高光谱人脸识别方法
WO2017208233A1 (en) * 2016-06-02 2017-12-07 Fst21 Ltd. Device and method for face identification
CN110337656A (zh) * 2019-05-27 2019-10-15 深圳市汇顶科技股份有限公司 用于人脸识别的光学传感器、装置、方法和电子设备
CN210469476U (zh) * 2019-07-30 2020-05-05 上海世科嘉车辆技术研发有限公司 一种可见光和红外光双通车载人脸识别摄像头
CN111611977A (zh) * 2020-06-05 2020-09-01 吉林求是光谱数据科技有限公司 基于光谱与多波段融合的人脸识别监控系统及识别方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117456584A (zh) * 2023-11-13 2024-01-26 江苏创斯达智能科技有限公司 一种应用于智能保险柜的面部识别设备

Also Published As

Publication number Publication date
US20220114834A1 (en) 2022-04-14
CN111611977B (zh) 2021-10-15
CN111611977A (zh) 2020-09-01
US11443550B2 (en) 2022-09-13

Similar Documents

Publication Publication Date Title
WO2021244414A1 (zh) 基于光谱与多波段融合的人脸识别监控系统及识别方法
US11869255B2 (en) Anti-counterfeiting face detection method, device and multi-lens camera
Du et al. A prism-based system for multispectral video acquisition
CN102647941A (zh) 用于执行光电容积描记的方法和系统
CN104881632A (zh) 高光谱人脸识别方法
CN110046564B (zh) 一种多光谱活体指纹识别设备及识别方法
Shen et al. Rapid detection of camouflaged artificial target based on polarization imaging and deep learning
CN104080663B (zh) 用于识别下雨的装置和利用摄像机识别下雨的方法
Chang et al. Multispectral visible and infrared imaging for face recognition
TWI798940B (zh) 光譜成像晶片及訊息處理方法
Bourlai et al. On designing a SWIR multi-wavelength facial-based acquisition system
JP2010035560A (ja) 生体認証装置
CN110312079A (zh) 图像采集装置及其应用系统
JP2014094301A (ja) 生体認証装置及び個人認証システム
WO2021197510A1 (zh) 具有真假指纹识别功能的手机指纹识别系统和识别方法
KR101608316B1 (ko) 실외 및 실내에서의 홍채이미지 획득장치 및 방법
CN114216867A (zh) 高光谱图像采集识别装置及方法
Ngo et al. Design and implementation of a multispectral iris capture system
FR3037690A1 (fr) Procede de detection de fraude par determination de la brdf d'un objet
CN114757223A (zh) 一种无感式激光主动探测微型窃视设备系统
Tominaga et al. High-resolution imaging system for omnidirectional illuminant estimation
CN108710843A (zh) 用于考勤的面部检测方法和装置
CN212135455U (zh) 生物特征检测装置、电子设备
LU501390B1 (en) Living Face Detection and Recognition System
WO2023109874A1 (zh) 识别系统及其工作方法

Legal Events

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

Ref document number: 21816886

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21816886

Country of ref document: EP

Kind code of ref document: A1