CN111597867B - Face recognition method and system based on millimeter wave image - Google Patents
Face recognition method and system based on millimeter wave image Download PDFInfo
- Publication number
- CN111597867B CN111597867B CN201911116583.3A CN201911116583A CN111597867B CN 111597867 B CN111597867 B CN 111597867B CN 201911116583 A CN201911116583 A CN 201911116583A CN 111597867 B CN111597867 B CN 111597867B
- Authority
- CN
- China
- Prior art keywords
- feature
- image
- millimeter wave
- feature extraction
- map
- 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
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000000605 extraction Methods 0.000 claims abstract description 82
- 238000013507 mapping Methods 0.000 claims abstract description 43
- 238000012545 processing Methods 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims description 34
- 239000013598 vector Substances 0.000 claims description 18
- 238000004422 calculation algorithm Methods 0.000 claims description 16
- 238000013527 convolutional neural network Methods 0.000 claims description 13
- 230000001815 facial effect Effects 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 238000003384 imaging method Methods 0.000 abstract description 45
- 238000007689 inspection Methods 0.000 abstract description 27
- 230000003287 optical effect Effects 0.000 description 14
- 230000006870 function Effects 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 101150064138 MAP1 gene Proteins 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 101150077939 mapA gene Proteins 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 210000000697 sensory organ Anatomy 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000004753 textile Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000005428 wave function Effects 0.000 description 1
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
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
-
- 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
- G06V40/172—Classification, e.g. identification
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Human Computer Interaction (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Collating Specific Patterns (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a face recognition method and a face recognition system based on millimeter wave images, wherein the method comprises the following steps: performing size processing on the millimeter wave image to obtain a processed image with the same size as the registered image in the face image retrieval library; extracting face features of the processed image to establish feature extraction mapping; and comparing the feature extraction map with a pre-established preliminary feature map of the registered image to obtain a matched preliminary feature map. The invention can realize face recognition based on the millimeter wave image generated by the millimeter wave security inspection imaging system.
Description
Technical Field
The invention relates to the technical field of face recognition, in particular to a face recognition method based on millimeter wave images and a face recognition system based on millimeter wave images.
Background
For increasing security inspection pressure, security inspection equipment capable of rapidly identifying personal identity information is also an important requirement besides security inspection equipment capable of identifying dangerous articles carried by a human body (such as knives, guns, explosives and the like). Currently, millimeter wave security inspection imaging technology is mature, and has been widely applied to inspecting dangerous goods carried by human bodies, but various biological feature recognition technologies based on passenger identities cannot be realized technically based on millimeter wave security inspection imaging systems.
Disclosure of Invention
The technical problem solved by the technical scheme of the invention is how to carry out face recognition based on the millimeter wave image generated by the millimeter wave security imaging system.
In order to solve the technical problems, the technical scheme of the invention provides a face recognition method of millimeter wave images, wherein the millimeter wave images are obtained based on a millimeter wave security inspection imaging system, and the method comprises the following steps:
performing size processing on the millimeter wave image to obtain a processed image with the same size as the registered image in the face image retrieval library;
extracting face features of the processed image to establish feature extraction mapping;
and comparing the feature extraction map with a pre-established preliminary feature map of the registered image to obtain a matched preliminary feature map.
Optionally, the performing size processing on the millimeter wave image includes:
and carrying out difference value or super-resolution reconstruction on the millimeter wave image to realize the size processing.
Optionally, the extracting the face features of the processed image to establish a feature extraction map includes:
training a model based on a convolutional neural network algorithm in advance;
and extracting the face features of the processed image according to the training model to establish a feature extraction mapping.
Optionally, the comparing the feature extraction map with a preliminary feature map pre-established by the registration image includes:
extracting face features of the historical millimeter wave images by utilizing the historical millimeter wave images and the training model in advance so as to establish historical feature extraction mapping of the historical face features;
establishing a corresponding feature library of the historical feature extraction mapping and the preliminary feature mapping;
and comparing the current feature extraction mapping with feature vectors in the corresponding feature library, and taking the face feature with the minimum distance between the current feature extraction mapping and the feature vectors as the current matching feature.
Optionally, the comparing the feature extraction map with the preliminary feature map pre-established by the registration image is based on the following algorithm:
let U and V be the feature extraction map and the preliminary feature map pre-established by the registered image, respectively, then the objective function may be expressed as:
min{D(U(I M )，V(I Gi ))}，i＝1，2，3.......N
wherein D represents a distance function for calculating feature identities of the feature extraction map and the preliminary feature map, I M Representing the millimeter wave image to be identified at this time, S Gi Representing a registered image S in the face image search library G I takes a natural number from 1 to N, where N is a natural number greater than or equal to 1.
Optionally, the face recognition method further includes:
and obtaining the corresponding registered image or the facial features based on the matched preliminary feature mapping so as to carry out face recognition.
In order to solve the technical problem, the technical scheme of the invention also provides a face recognition system based on the millimeter wave image, and a millimeter wave security inspection imaging system based on the millimeter wave image, which comprises the following steps: the size processing module, the feature extraction module and the comparison module are connected with the millimeter wave security inspection imaging system;
the size processing module is suitable for performing size processing on the millimeter wave image to obtain a processed image with the same size as a registered image in a face image retrieval library;
the feature extraction module is suitable for extracting the face features of the processed image to establish feature extraction mapping;
the comparison module is adapted to compare the feature extraction map with a preliminary feature map pre-established for the registration image to obtain a matched preliminary feature map.
Optionally, the face recognition system further includes: a training module; the training module is suitable for training a model based on a convolutional neural network algorithm in advance, and the feature extraction module is suitable for extracting face features of the processed image according to the training model to establish a feature extraction mapping.
Optionally, the face recognition system further includes: a historical feature extraction module adapted to:
extracting face features of the historical millimeter wave images by utilizing the historical millimeter wave images and the training model in advance so as to establish historical feature extraction mapping of the historical face features;
establishing a corresponding feature library of the historical feature extraction mapping and the preliminary feature mapping;
the comparison module is further adapted to compare the current feature extraction mapping with feature vectors in the corresponding feature library, and take the face feature with the minimum distance between the current feature extraction mapping and the feature vectors as the current matching feature.
Optionally, the face recognition system further includes: an identification module; the identification module is suitable for obtaining corresponding registered images or facial features based on the matched preliminary feature mapping to perform face identification.
The technical scheme of the invention has the beneficial effects that at least:
according to the technical scheme, the face recognition can be realized on the basis of the millimeter wave image generated by the millimeter wave security imaging system and on the basis of the low-resolution face image, and the advantage of penetrability of the millimeter wave imaging to human clothes is combined, so that the application of the face recognition on the millimeter wave security imaging system can be realized, the shielding of textiles on the face can be effectively removed during recognition, and the accuracy and the effectiveness of the face recognition are improved.
According to the technical scheme, the image face features can be trained in advance, feature extraction can be performed based on the historical images, feature correspondence of the historical images and forecast feature mapping can be achieved, continuous updating of data in a face image retrieval library and a feature library can be achieved, and matching accuracy and matching efficiency of the face features are improved.
Drawings
Other features, objects and advantages of the present invention will become more apparent upon reading of the detailed description of non-limiting embodiments, given with reference to the accompanying drawings in which:
FIG. 1 is a schematic diagram of an active millimeter wave imaging security inspection instrument for collaborative human security inspection;
FIG. 2 is a schematic diagram of a three-dimensional densely-packed array imaging system;
fig. 3 is a schematic structural diagram of a non-cooperative active millimeter wave imaging mechanism;
FIG. 4 is a schematic illustration of a millimeter wave image obtained based on a millimeter wave security imaging system;
FIG. 5 is a schematic illustration of a human face portion of a millimeter wave image obtained based on a millimeter wave security imaging system;
fig. 6 is a schematic flow chart of a face recognition method based on millimeter wave images according to the technical scheme of the invention;
FIG. 7 is a schematic diagram of a process for training a convolutional neural network model using historical millimeter wave image samples and registered images and applying the convolutional neural network model to millimeter wave image feature extraction in the technical scheme of the invention;
fig. 8 is a schematic flow chart of another face recognition method based on millimeter wave images according to the technical scheme of the invention;
fig. 9 is a schematic structural diagram of a face recognition system based on millimeter wave images according to the technical scheme of the present invention;
fig. 10 is a schematic structural diagram of another face recognition system based on millimeter wave images according to the technical scheme of the present invention;
fig. 11 is a schematic structural diagram of another face recognition system based on millimeter wave images according to the technical scheme of the present invention;
fig. 12 is a schematic structural diagram of another face recognition system based on millimeter wave images according to the present invention.
Detailed Description
In order to better and clearly show the technical scheme of the invention, the invention is further described below with reference to the accompanying drawings.
The technical scheme of the invention takes a millimeter wave imaging security inspection system for inspecting dangerous goods carried by a human body as a platform, and aims at the key technology of millimeter wave human body biological feature identification in non-contact and rapid security inspection, through theoretical analysis, three-dimensional electromagnetic simulation and big data experiments, the key scientific problems of human body electromagnetic scattering mechanism, characteristic layer, fractional layer and actual physical parameter relation in an optical image and multi-mode biological feature neural network fusion model and the like in the cooperative and non-cooperative state are researched on the internal corresponding biological feature relation of an optical image and a millimeter wave image, and the key scientific problems of breaking through in the technology of a low-resolution millimeter wave face recognition algorithm are taken as means by taking various machine learning, so that the foundation is laid for further practical application of the face recognition algorithm in millimeter wave imaging security inspection.
In the prior art, because the millimeter wave frequency band self wavelength, the antenna aperture size of an actual imaging system and the real-time rapidity requirement required by security inspection are limited, the resolution of the millimeter wave human body image generally acquired is not high, the resolution is lower when reaching the face part of a human body, and the resolution can be generally only 20 multiplied by 20 or even lower, but the five sense organs of the human face can be generally resolved, and a certain recognition basis is provided. Therefore, how to identify the face under the condition of low signal-to-noise ratio compared with the optical image is a problem that the technical scheme of the invention needs to be researched. The method mainly comprises the following technical difficulties: firstly, the data dimensions are different, and only high-resolution face front images, such as identity card photos, are usually provided, and the data dimensions of the optical images are very high. And the millimeter wave low-resolution face image data has very low dimensionality, so that many face matching methods cannot be directly applied. Second is the data difference. The optical face image and the millimeter wave face image have larger gesture and illumination condition difference. In the technical scheme of the invention, all the problems can be tried to be processed in a multi-scale cross-dimensional space.
The technical scheme of the invention also relates to an inherent corresponding biological characteristic relation between the high-dimensional optical face image and the low-dimensional face image. In the prior art, the image databases used for face recognition are high-resolution high-dimensional optical images, and the millimeter wave face images are low-resolution low-dimensional face images, so that the millimeter wave face database used for face recognition is not available at present, and the millimeter wave face recognition is needed to be carried out by the existing optical face database. For the same natural person to be identified, the two images necessarily have related intrinsic biological characteristic relations, including characteristic vectors from the simplest face intrinsic characteristic points to the intra-class relations. According to the technical scheme, through the deep learning neural network, the corresponding characteristic relation between the optical high-dimensional image and the millimeter wave low-dimensional image crossing the space is researched, so that a foundation is laid for further millimeter wave face image recognition.
The main structure of the millimeter wave security inspection imaging system in the prior art, namely an active millimeter wave imaging security inspection instrument for collaborative human body security inspection is shown in fig. 1, the system combines electronic array scanning and mechanical scanning, and the millimeter wave transceiver imaging front-end linear array with 2×80 array elements is adopted, the horizontal direction is scanned by means of switching of a switch by the imaging front-end linear array, the vertical direction is scanned by the vertical mechanical movement of the imaging front-end linear array, and then multi-frequency point scanning is carried out by combining each horizontal scanning, so that air-frequency three-dimensional electromagnetic echo data of an imaging target area is obtained. From the imaging algorithm, the imaging system can be understood as a three-dimensional densely packed array imaging system as shown in fig. 2.
Referring to fig. 2, assuming that the target ("target") is a body target ("target point") on which a strong reflection point coordinates are (x, y, Z), a scanning plane ("scanned aperture") of a distributed transceiver realized by electronic switching and vertical mechanical scanning is still located at z=z 1 Is (x ', y', Z) the position ("transceiver position") of a transceiver in the scan plane 1 )。
When the system operates in a wide frequency band instead of a frequency point, the reflection characteristic function f (x, y, z) of the target is integrated to obtain the receiving response of a transceiver at the frequency point ω=kc, where k is a wave function.
The basic algorithm principle formula of the system can be obtained through transformation as follows:
as can be seen from the algorithm formula, the active millimeter wave security inspection imaging needs to acquire echo complex signals of each scanning point, namely the amplitude and the phase of the echo, so that an imaging scene image can be recovered.
Unlike the mode of combining electronic switch switching and mechanical scanning adopted by cooperative active millimeter wave imaging, in the other millimeter wave security inspection imaging system, namely, non-cooperative active millimeter wave imaging adopts a full-electronic sparse imaging array surface mechanism, and each frame of image can be acquired within 0.1 s. As shown in fig. 3, the whole full-electronic sparse array emits millimeter waves towards the target scene, then receives reflected and scattered echo signals of the target scene and performs data acquisition, and finally restores the image of the target scene through a corresponding sparse imaging algorithm. From the basic principle of imaging, the sparse array equivalent calculation is mainly restored to the dense array, so that imaging is performed according to the formula (2).
Both systems can be used for generating the millimeter wave patterns related in the technical scheme of the invention, namely, the technical scheme of the invention can be applied to both millimeter wave security inspection imaging systems so as to realize face recognition.
As described above, the face images generally used for identity authentication are high-resolution optical face images, but the millimeter wave images (as shown in FIG. 4) obtained when the millimeter wave security imaging system is used for security imaging of passengers are not high in resolution, and particularly to the face parts (as shown in FIG. 5) of the millimeter wave images, the resolution is possibly lower, and is generally only about 20×20. In practical application, it is difficult to obtain a millimeter wave face image directly usable for identity recognition, so that it is necessary to perform recognition and authentication on a low-resolution millimeter wave face image through a high-resolution optical face image.
The face recognition method based on the millimeter wave image shown in fig. 6, wherein the millimeter wave image is obtained based on a millimeter wave security inspection imaging system, and comprises the following steps:
step S100, performing size processing on the millimeter wave image to obtain a processed image with the same size as a registered image in a face image retrieval library;
step S101, extracting face features of the processed image to establish feature extraction mapping;
step S102, comparing the feature extraction map with a preliminary feature map pre-established by the registration image to obtain a matched preliminary feature map.
In step S100, the millimeter wave image generally refers to a face part image (refer to fig. 5) in the original millimeter wave image (refer to fig. 4), and since the face part image is small in size, the face part image needs to be reconstructed by size processing to enlarge the size. The size processing may be achieved by performing a difference or super-resolution reconstruction of the millimeter wave image. The main means is to expand the smaller lower resolution size of the original millimetric wave image to the same size as the registered image. The size processing means include, but are not limited to, difference processing and super resolution reconstruction processing.
If the original millimeter wave image obtained by the millimeter wave security inspection imaging system is a whole body image, the whole body image needs to be intercepted before entering step S100 to obtain the millimeter wave image (i.e. the image of the face part) before processing in step S100. The original millimeter wave image is the image of the face part, and the original millimeter wave image is not required to be intercepted before entering the S100 processing.
In step S101, there may be various ways of extracting the facial features of the processed image, specifically, reference may be made to an image feature extraction method.
The face features can be extracted specifically through the following steps based on the convolutional neural network:
firstly, extracting face features with different scales from a processed face image to be recognized;
and then fusing the face features with different scales.
The convolutional neural network can be a classical google net (a deep neural network built by google) and deep id (a deep neural network built by university of hong kong chinese for face recognition) network or an improvement on the above. A particular construction of a particular convolutional neural network may also include a network structure constructed of convolutional layers, pooling layers, connection layers, and full connection layers. The convolution layer of the convolution neural network mainly convolves the face image and extracts the face characteristics after convolution; the pooling layer pools the characteristic images, reduces the size of the characteristic images and reduces the parameters of the network; the connecting layer plays a role of continuous connection; the full connection layer is used for mapping data from the feature space to the classifier space and plays a role of a classifier.
Training models according to convolutional neural network based algorithms may be trained according to historical samples to obtain pre-trained training models, as shown in connection with fig. 7. Fig. 7 illustrates training a Convolutional Neural Network (CNN) model by using a historical millimeter wave image sample and a registered image, and a relatively reliable training model and a relevant feature library are formed after training, so that the training model can be used for performing direct feature extraction of an acquired original millimeter wave image for identity recognition. In the technical scheme of the invention, the training model can be utilized to extract the face characteristics of the millimeter wave image processed in real time so as to establish characteristic extraction mapping. In general, the feature extraction mapping is a feature vector formed based on the recognized face features, and how to construct the feature-to-feature vector mapping is based on the training model.
According to step S102, the comparison of the feature extraction map with the preliminary feature map pre-established by the registration image may be based on the following algorithm:
let U and V be the feature extraction map and the preliminary feature map pre-established by the registered image, respectively, then the objective function may be expressed as:
min{D(U(I M )，V(I Gi ))}，i＝1，2，3......N (3)
wherein D represents a distance function for calculating feature identities of the feature extraction map and the preliminary feature map, I M Representing the millimeter wave image to be identified at this time, I Gi Representing a registered image I in the face image retrieval library G I takes a natural number from 1 to N, where N is a natural number greater than or equal to 1.
In another scheme that may be parallel processing or may be separately applicable, step S102 further includes a process that includes the following steps:
extracting face features of the historical millimeter wave images by utilizing the historical millimeter wave images and the training model in advance so as to establish historical feature extraction mapping of the historical face features;
establishing a corresponding feature library of the historical feature extraction mapping and the preliminary feature mapping;
and comparing the current feature extraction mapping with feature vectors in the corresponding feature library, and taking the face feature with the minimum distance between the current feature extraction mapping and the feature vectors as the current matching feature.
The step of comparing is to train a large number of history matched millimeter wave face images and optical images to obtain a feature library containing the corresponding millimeter wave face biological features and optical face biological features. When the method is applied, corresponding interpolation sampling or super-resolution reconstruction processing is carried out on the millimeter wave face image tested at the time so as to achieve the consistency of the size and the optical image registered in the search library, then the feature extraction is carried out on the millimeter wave face image by using a model established by a trained convolutional neural network algorithm, finally the feature extraction is carried out by comparing the feature extraction with feature vectors in a corresponding feature library according to a certain rule, and the recognition can be achieved by selecting the minimum distance of the feature vectors.
The processing is also based on the algorithm (3), but when the comparison is performed using the relevant corresponding feature library, V in the function includes not only the preliminary feature map but also the preliminary feature mapA historical feature extraction map in the corresponding feature library that matches the preliminary feature map, in this case I Gi And also represents the historical millimeter wave image corresponding to the ith registered image in the face image retrieval library. In this process, the comparison object of the present feature extraction map may be one feature vector or two feature vectors of the preliminary feature map and the historical feature extraction map that are correspondingly matched with the preliminary feature map.
In another example of the technical scheme of the present invention, as shown in fig. 8, a face recognition method based on millimeter wave images, wherein the millimeter wave images are obtained based on a millimeter wave security inspection imaging system, and the method comprises steps 100 to 102 of the method shown in fig. 6, and further comprises the following steps:
step S103, corresponding registered images or face features are obtained based on the matched preliminary feature mapping to conduct face recognition.
According to step S103, if the matching result obtained in S102 is positive, the present feature extraction map has a preliminary feature map matched with the present feature extraction map (even if the present feature extraction map is actually matched with the present feature extraction map, the history feature extraction map has a preliminary feature map corresponding to the present feature extraction map, and the preliminary feature map and the present feature extraction map are also positive-matched). When the face recognition is carried out, the face features and the registered images corresponding to the original millimeter wave images transmitted at this time can be obtained according to the matched preliminary feature mapping, and the face features or the registered images can be utilized to obtain the technical application of the face recognition.
Based on the face recognition method based on the millimeter wave image, the technical scheme of the invention also provides a face recognition system 1 based on the millimeter wave image as shown in fig. 9, and a millimeter wave security inspection imaging system based on the millimeter wave image, which comprises the following steps: a size processing module 100, a feature extraction module 101 and a comparison module 102, which can be connected with the millimeter wave security inspection imaging system.
The size processing module 100 may receive the original millimeter wave image transmitted by the millimeter wave security inspection imaging system, and the size processing module 100, the feature extraction module 101 and the comparison module 102 sequentially execute the processing steps of steps S100 to S102.
The face recognition system 2 based on millimeter wave images as shown in fig. 10 includes a training module 203 in addition to a size processing module 200, a feature extraction module 201, and a comparison module 202. The training module 203 is adapted to train a model in advance based on a convolutional neural network algorithm, and the feature extraction module 201 is adapted to extract facial features of the processed image according to the training model trained by the training module 203 to establish a feature extraction map. Reference is made to the above for the remaining technical means.
The millimeter wave image-based face recognition system 3 shown in fig. 11 includes, in addition to a size processing module 300, a feature extraction module 301, and a comparison module 302, also includes: the history feature extraction module 303, the history feature extraction module 303 being adapted to:
extracting face features of the historical millimeter wave images by utilizing the historical millimeter wave images and the training model in advance so as to establish historical feature extraction mapping of the historical face features;
and establishing a corresponding feature library of the historical feature extraction mapping and the preliminary feature mapping.
The comparison module 302 is further adapted to compare the current feature extraction map with feature vectors in the corresponding feature library, and use a face feature with a minimum distance between the current feature extraction map and the feature vectors as a current matching feature.
The system 2 may also be used in combination with the historical feature extraction module 303, and the training module used by the historical feature extraction module 303 may be a training model obtained by training by the training module 203.
The above systems 1 to 3 may also be combined with an identification module 400 for performing the functions of step S103 described above, respectively. Fig. 12 shows an example of a face recognition system 4 based on millimeter wave images, comprising an identification module 400 in combination with the system 1 to form the system 4.
In the system of the technical scheme of the invention, the related database or the feature library can be part of the system or can be provided by external equipment of the system. Similarly, the training and outputting of the training model may be obtained by a training module as in the system 2, or may be a training model directly output by an external device of the system.
The modules of the system of the technical scheme of the invention correspondingly execute the step flow of the method of the technical scheme of the invention, the modules can perform sub-modularization or inter-module function integration on the processing functions of the modules according to the compiling requirement of actual step execution, and the system has different inter-module structures according to different modularization and falls into the protection scope of the technical scheme of the invention.
The foregoing describes specific embodiments of the present invention. It is to be understood that the invention is not limited to the particular embodiments described above, and that various changes and modifications may be made by one skilled in the art within the scope of the claims without affecting the spirit of the invention.
Claims (1)
1. The face recognition method based on the millimeter wave image is characterized by comprising the following steps of:
performing size processing on the millimeter wave image to obtain a processed image with the same size as the registered image in the face image retrieval library;
extracting face features of the processed image to establish feature extraction mapping;
comparing the feature extraction map with a pre-established preliminary feature map of the registration image to obtain a matched preliminary feature map;
the performing size processing on the millimeter wave image includes:
performing difference value or super-resolution reconstruction on the millimeter wave image to realize the size processing;
the extracting the face features of the processed image to establish a feature extraction map includes:
training a model based on a convolutional neural network algorithm in advance;
extracting face features of the processed image according to the training model to establish feature extraction mapping;
the comparing the feature extraction map with the preliminary feature map pre-established by the registration image includes:
extracting face features of the historical millimeter wave images by utilizing the historical millimeter wave images and the training model in advance so as to establish historical feature extraction mapping of the historical face features;
establishing a corresponding feature library of the historical feature extraction mapping and the preliminary feature mapping;
comparing the current feature extraction mapping with feature vectors in the corresponding feature library, and taking the face feature with the minimum distance between the current feature extraction mapping and the feature vectors as the current matching feature;
the comparison of the feature extraction map with the preliminary feature map pre-established for the registered image is based on the following algorithm:
let U and V be the feature extraction map and the preliminary feature map pre-established by the registered image, respectively, then the objective function may be expressed as:
；
wherein D represents a distance function for calculating feature extraction mapping and feature recognition of preliminary feature mapping, IM represents a millimeter wave image to be recognized at this time, I Gi Representing a registered image I in the face image retrieval library G I takes a natural number from 1 to N, where N is a natural number greater than or equal to 1;
when the relevant corresponding feature library is used for comparison, V in the function comprises not only the preliminary feature map but also the historical feature extraction map corresponding to and matched with the preliminary feature map in the corresponding feature library, and I Gi The comparison object of the feature extraction map in the process can be a feature direction of a history feature extraction map which is matched with the prepared feature mapComparing the quantity or the two characteristic vectors at the same time;
the face recognition method further comprises the following steps:
and obtaining the corresponding registered image or the facial features based on the matched preliminary feature mapping so as to carry out face recognition.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2019100904776 | 2019-01-30 | ||
| CN201910090477.6A CN109886157A (en) | 2019-01-30 | 2019-01-30 | A kind of face identification method and system based on millimeter-wave image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111597867A CN111597867A (en) | 2020-08-28 |
| CN111597867B true CN111597867B (en) | 2024-01-12 |
Family
ID=66927360
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910090477.6A Pending CN109886157A (en) | 2019-01-30 | 2019-01-30 | A kind of face identification method and system based on millimeter-wave image |
| CN201911116583.3A Active CN111597867B (en) | 2019-01-30 | 2019-11-15 | Face recognition method and system based on millimeter wave image |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910090477.6A Pending CN109886157A (en) | 2019-01-30 | 2019-01-30 | A kind of face identification method and system based on millimeter-wave image |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN109886157A (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110807403B (en) * | 2019-10-29 | 2022-12-02 | 中新智擎科技有限公司 | User identity identification method and device and electronic equipment |
| CN110779149A (en) * | 2019-11-14 | 2020-02-11 | 宁波奥克斯电气股份有限公司 | Air conditioner control method and device and air conditioner |
| CN111523478B (en) * | 2020-04-24 | 2023-04-28 | 中山大学 | Pedestrian image detection method acting on target detection system |
| CN111931762B (en) * | 2020-09-25 | 2021-07-30 | 广州佰锐网络科技有限公司 | AI-based image recognition solution method, device and readable storage medium |
| CN116434289A (en) * | 2021-12-30 | 2023-07-14 | 同方威视技术股份有限公司 | Millimeter wave or terahertz wave identity verification method and device, millimeter wave or terahertz wave security inspection equipment and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101710386A (en) * | 2009-12-25 | 2010-05-19 | 西安交通大学 | Super-resolution face recognition method based on relevant characteristic and non-liner mapping |
| CN106469315A (en) * | 2016-09-05 | 2017-03-01 | 南京理工大学 | Based on the multi-mode complex probe target identification method improving One Class SVM algorithm |
| CN107423690A (en) * | 2017-06-26 | 2017-12-01 | 广东工业大学 | A kind of face identification method and device |
| CN107609459A (en) * | 2016-12-15 | 2018-01-19 | 平安科技（深圳）有限公司 | A kind of face identification method and device based on deep learning |
| EP3425421A1 (en) * | 2017-07-07 | 2019-01-09 | Infineon Technologies AG | System and method for identifying a biological target using radar sensors |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7873182B2 (en) * | 2007-08-08 | 2011-01-18 | Brijot Imaging Systems, Inc. | Multiple camera imaging method and system for detecting concealed objects |
| CN104574304A (en) * | 2014-12-25 | 2015-04-29 | 深圳市一体太赫兹科技有限公司 | Millimeter wave image reconstruction method and system |
| CN104537619A (en) * | 2014-12-25 | 2015-04-22 | 深圳市一体太赫兹科技有限公司 | Millimeter wave image recovery method and system |
| CN106326834B (en) * | 2016-07-29 | 2019-12-10 | 华讯方舟科技有限公司 | method and device for automatically identifying sex of human body |
| CN106529602B (en) * | 2016-11-21 | 2019-08-13 | 中国科学院上海微系统与信息技术研究所 | A kind of millimeter-wave image automatic target recognition method and device |
-
2019
- 2019-01-30 CN CN201910090477.6A patent/CN109886157A/en active Pending
- 2019-11-15 CN CN201911116583.3A patent/CN111597867B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101710386A (en) * | 2009-12-25 | 2010-05-19 | 西安交通大学 | Super-resolution face recognition method based on relevant characteristic and non-liner mapping |
| CN106469315A (en) * | 2016-09-05 | 2017-03-01 | 南京理工大学 | Based on the multi-mode complex probe target identification method improving One Class SVM algorithm |
| CN107609459A (en) * | 2016-12-15 | 2018-01-19 | 平安科技（深圳）有限公司 | A kind of face identification method and device based on deep learning |
| CN107423690A (en) * | 2017-06-26 | 2017-12-01 | 广东工业大学 | A kind of face identification method and device |
| EP3425421A1 (en) * | 2017-07-07 | 2019-01-09 | Infineon Technologies AG | System and method for identifying a biological target using radar sensors |
Non-Patent Citations (2)
| Title |
|---|
| B.G.Alefs等.Thorax biometrics from millimeter-wave images.Pattern Recognition Letters 31.2010,摘要、第2.1节和第3节以及图4. * |
| 基于小波域正则化的毫米波图像重构;汪先平;;漳州职业技术学院学报(01);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109886157A (en) | 2019-06-14 |
| CN111597867A (en) | 2020-08-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111597867B (en) | Face recognition method and system based on millimeter wave image | |
| Cui et al. | Image data augmentation for SAR sensor via generative adversarial nets | |
| Seyfioglu et al. | DNN transfer learning from diversified micro-Doppler for motion classification | |
| CN106156744B (en) | SAR target detection method based on CFAR detection and deep learning | |
| He et al. | A forward approach to establish parametric scattering center models for known complex radar targets applied to SAR ATR | |
| Gurbuz et al. | Cross-frequency training with adversarial learning for radar micro-Doppler signature classification (Rising Researcher) | |
| Feng et al. | Robust and efficient algorithms for separating latent overlapped fingerprints | |
| CN108509910A (en) | Deep learning gesture identification method based on fmcw radar signal | |
| Zhou et al. | CANet: An unsupervised deep convolutional neural network for efficient cluster-analysis-based multibaseline InSAR phase unwrapping | |
| CN105809693A (en) | SAR image registration method based on deep neural networks | |
| Soltanpour et al. | Multimodal 2D–3D face recognition using local descriptors: pyramidal shape map and structural context | |
| CN109753887B (en) | SAR image target identification method based on enhanced kernel sparse representation | |
| CN110516525A (en) | SAR image target recognition method based on GAN and SVM | |
| Shi et al. | Robust gait recognition based on deep cnns with camera and radar sensor fusion | |
| Ding et al. | Human target localization using Doppler through-wall radar based on micro-Doppler frequency estimation | |
| Lin et al. | Using CNN to classify hyperspectral data based on spatial-spectral information | |
| Wang et al. | Buried target detection method for ground penetrating radar based on deep learning | |
| Bhanu et al. | Recognizing target variants and articulations in synthetic aperture radar images | |
| CN116311067A (en) | Target comprehensive identification method, device and equipment based on high-dimensional characteristic map | |
| Bensaïd et al. | Partial matching of nonrigid shapes by learning piecewise smooth functions | |
| CN114418913B (en) | ISAR and infrared image pixel level fusion method based on wavelet transformation | |
| NAWWAR et al. | Improvement of confusion matrix for hand vein recognition based on deep-learning multi-classifier decisions | |
| Gong et al. | Human activity recognition with FMCW radar using few-shot learning | |
| Zhai et al. | MF‐SarNet: Effective CNN with data augmentation for SAR automatic target recognition | |
| Zhang et al. | Based on Siamese network with self-attention model for gait recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |