A single-chip offline face recognition module

Abstract

The present invention discloses a small single-chip offline face recognition module, which belongs to the field of face recognition. The small single-chip offline face recognition module includes a face image acquisition unit, a recognition detection unit, a face image processing unit, a memory unit, a control unit, a voice intercom unit and a single-chip microcomputer chip; the face image acquisition unit is used to acquire the face image of a person; the recognition detection unit is used to generate a recognition index according to the face information of the person, and when the recognition index exceeds the threshold, the face of the person is recognized, otherwise the face of the person is not recognized. When processing the face image of a person, in addition to the conventional feature vector method to generate a feature vector that uniquely describes the face of the person, the present invention can also generate a face feature coefficient according to the face features of the person. When the face features of the person are partially blocked during the face recognition process, the face feature coefficient can still be used to complete the face recognition.

Description

Technical Field

[0001] The invention relates to face recognition, and in particular to a single-chip offline face recognition module. Background Art

[0002] Face recognition is a biometric technology that identifies people based on their facial features. It uses a camera or a camera to collect images or video streams containing faces, and automatically detects and tracks faces in the images, and then performs a series of related technologies on the detected faces, which are usually also called portrait recognition and facial recognition. Due to its strong security and uniqueness, face recognition is widely used in various industries, such as enterprise management, public safety, and mobile phone manufacturing. Due to the limitations of cost and manufacturing difficulty, most of the current face recognition uses 2D face recognition technology and uses the feature vector method to recognize faces. Specifically, by extracting facial features (eyes, nose, mouth, ears, etc.), and then determining the size, position, distance and other attributes of the facial features, and then calculating their geometric feature quantities, these feature quantities form a facial feature vector that describes the face image. When performing face recognition in the future, face recognition is completed by matching the facial feature vector. Since the current face recognition is based on the unique feature vector generated by facial features, when the facial features are partially blocked, it will be impossible to recognize, resulting in the failure of face recognition. Although there is a technology that can realize face recognition by blocking part of the face, it is realized by collecting the face image after the blockage, extracting the features of the face image after the blockage, and generating the feature vector of the face after the blockage. When other features are blocked, the face cannot be recognized, resulting in the failure of face recognition. Summary of the invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a single-chip offline face recognition module.

[0004] Technical solution: A single-chip offline face recognition module, characterized in that it includes a face image acquisition unit, a recognition detection unit, a face image processing unit, a memory unit, a control unit, a voice intercom unit and a single-chip microcomputer chip;

[0005] Face image acquisition unit: used to collect face images of people;

[0006] Recognition detection unit: used to generate the recognition index according to the facial information of the person through the following formula RI ,when RI≥RI th , the current person is detected as an identified person, otherwise the current person is detected as a non-identified person, where, RI th is a preset recognition index threshold parameter;

[0007] ;

[0008] in, RI is the identification index of the currently identified person, L , W is the length and width of the face image of the person being recognized. L 0 , W 0 are the preset reference length and width respectively, t 1 , t 2 Select the time for two points of ultrasonic wave to be emitted from the transmitter and received by the receiver for each face;

[0009] Personnel image processing unit: used to extract the features of the face image collected by the face image acquisition unit, and generate facial feature vectors according to the feature vector method. At the same time, a two-dimensional coordinate system is established with the center point of the face image as the origin and each pixel as the coordinate axis unit. Then, a number of data points are calibrated on the face feature contour, and the coordinate parameters of each data point are recorded to generate the face feature coefficients:

[0010] ;

[0011] in, E is the facial feature coefficient, Δx , Δy are the differences in the horizontal and vertical coordinate values ​​between each adjacent point on the contour of the facial feature, δx , δy The facial feature vector and the facial feature coefficient are respectively the differences between the horizontal and vertical coordinate values ​​of each adjacent point under the contour of the facial feature, and the facial feature vector and the facial feature coefficient are written into the memory unit for storage; and are also used for matching the facial feature vector of the identified person with the facial feature vector stored in the memory unit when the recognition detection unit recognizes the face, and feeding back the matching result; when part of the facial features of the identified person are blocked, matching the facial feature coefficients of the unblocked person with the facial feature coefficients in the memory unit, and feeding back the matching result;

[0012] Memory unit: used to store facial feature vectors and facial feature coefficients of personnel;

[0013] Control unit: used to control the connected device according to the matching result of the negative feedback from the personnel image processing unit. When the matching result of the feedback is 1, the control device is turned on. When the matching result of the feedback is 0, the control device is turned off. It also controls the voice intercom unit.

[0014] Voice intercom unit: used to collect the voice of personnel, recognize the voice, and then play the voice of personnel, so as to realize the two-way voice intercom function;

[0015] Single-chip microcomputer chip: This chip integrates a recognition and analysis module, a feature data processing module, a comparison and matching module and a speech recognition module. The recognition and analysis module is used to help the recognition detection unit detect the recognition object. The feature data processing module is used to help the face image processing unit process the face image. The comparison and matching module is used to help the face image processing unit compare and match the facial feature vector or facial feature coefficient when recognizing the face. The speech recognition module is used to help the voice intercom module realize voice intercom.

[0016] Preferably, the recognition and detection unit includes an ultrasonic sensor module, a recognition and analysis module, and a facial information acquisition module;

[0017] Ultrasonic sensor module: used to emit ultrasonic waves and collect the time it takes for the ultrasonic waves to be emitted from the transmitter and received by the receiver at two points selected on the face;

[0018] Facial information acquisition module: used to obtain the length and width of the face of the identified person from the feature data processing module;

[0019] Identification and analysis module: It is used to generate a recognition index based on the data obtained by the ultrasonic sensor module and the facial information acquisition module, and detect whether the current person is the person who needs to be identified based on the generated recognition index.

[0020] Preferably, the face image processing unit includes a face feature extraction module, a feature data processing module, an information data writing module and a comparison and matching module;

[0021] Face feature extraction module: used to extract features of a person's face from the face image collected by the face image collection unit;

[0022] Feature data processing module: used to generate facial feature vectors of personnel according to the extracted facial features of personnel through feature vector method; and generate facial feature coefficients according to the extracted facial features of personnel;

[0023] Information data writing module: used to write the facial feature vector and facial feature coefficient processed by the feature data processing module into the memory unit for storage;

[0024] Comparison and matching unit: used to match the facial feature vector of the identified person with the facial feature vector stored in the memory unit, and feed back the matching result. When part of the facial features of the identified person are blocked, the unblocked facial feature coefficients are matched with the facial feature coefficients in the memory unit, and the matching result is fed back.

[0025] Preferably, the voice intercom unit includes a microphone module, a voice recognition module and a speaker module;

[0026] Microphone module: used to collect the voice of the person collecting the data and send the collected voice to the speech recognition module;

[0027] Speech recognition module: used to recognize and convert the voice of the person collected by the microphone module and send it to the speaker module;

[0028] Speaker module: used to convert the voice data recognized and converted by the voice recognition module into sound information and play it out.

[0029] Preferably, the memory size of the memory unit is 32 GB.

[0030] Preferably, the single-chip offline face recognition module also includes a module integrating a power interface and a communication interface, wherein the communication interface uses TCP / IP protocol for communication.

[0031] Beneficial effects:

[0032] (1) During the face recognition process, the present invention can automatically detect the person to be recognized and recognize the automatically detected person, while not recognizing the faces of other persons captured by the camera, so that the single-chip offline face recognition module has a good primary and secondary logic when recognizing faces. At the same time, it can prevent other unrecognizable persons from using the persons who can complete face recognition later to complete face recognition, thereby activating the device using the single-chip offline face recognition module, so that the single-chip offline face recognition module has high security.

[0033] (2) During the face recognition process, the present invention can automatically detect the person to be identified and determine whether the person to be identified is a real person, thereby preventing other people from using photos of people who can complete face recognition to complete face recognition authentication, thereby enabling the use of the single-chip offline face recognition small module in devices, further enhancing the security performance of the single-chip offline face recognition small module.

[0034] (3) When processing a person's facial image, the present invention can generate a facial feature coefficient based on the person's facial features in addition to the conventional feature vector method to generate a unique feature vector describing the person's face. When part of the person's facial features are blocked during face recognition, face recognition can still be completed using the facial feature coefficient, so that the single-chip offline face recognition module has a higher fault tolerance and a better face recognition effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the module principle of the present invention;

[0036] Figure 2 It is a schematic diagram of generating facial feature coefficients of the present invention;

[0037] Figure 3 It is a schematic diagram of identification detection of the present invention. DETAILED DESCRIPTION

[0038] In order to make the technical solution of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] Example:

[0040] like Figure 1 As shown, the single-chip offline face recognition module consists of a camera equipped with Sony IMX989 as a face image acquisition unit, a recognition detection unit, a face image processing unit, a memory unit with a capacity of 32GB, a control unit, a voice intercom unit, a single-chip microcomputer chip with integrated recognition analysis, feature data processing and comparison and matching functions, and a module with integrated power interface and communication interface.

[0041] When the single-chip offline face recognition module is applied to a specific device, such as a safe, door lock, access control terminal, etc., the power supply is first connected through the power interface to power on the single-chip offline face recognition module. Then, when entering the face information so that face recognition can be completed later, the face image acquisition unit first collects the face image of the person, and the collected face image of the person is handed over to the face image processing unit for processing. In the process of processing the face image of the person, the face image processing unit first extracts the features of eyes, nose, mouth, ears, etc. from the face image through the face feature extraction module, and then determines the size, position, distance and other attributes of each face feature, and calculates their geometric feature quantities using the feature vector method through the feature data processing module, and these feature quantities form a facial feature vector describing the face image (this is the current common method of face recognition, and the details will not be repeated). At the same time, the facial features of the eyes, mouth and ears are selected, and a two-dimensional coordinate system is established with the center point of the face image as the origin and each pixel as the coordinate axis unit. Several data points are marked on the contours of the eyes, mouth and ears, and the coordinate parameters of each data point are recorded, such as Figure 2 As shown, the facial feature coefficients of the eyes and mouth are calculated by the feature data processing module using the following formula:

[0042] ;

[0043] in, E is the facial feature coefficient of eyes / mouth, Δx , Δy are the differences in the horizontal and vertical coordinate values ​​between each adjacent point on the eye / mouth contour, δx , δy They are the differences in the horizontal and vertical coordinate values ​​between each adjacent point under the eye / mouth contour.

[0044] For the ear, a facial feature, since there is no concept of upper and lower contours, when calculating the facial feature coefficient of the ear, there is:

[0045] ;

[0046] in, E is the facial feature coefficient of the ear, Δx , Δy They are the differences in the horizontal and vertical coordinate values ​​between each adjacent point of the ear contour.

[0047] For the facial feature coefficient of the eye, two data points b and c are calibrated on the upper contour of the eye, two data points e and f are calibrated on the lower contour of the eye, and two data points a and d are calibrated at the outer and inner corners of the eye respectively (the two data points a and d are attributed to the upper contour), and the coordinates of the data points a, b, c, d, e, and f are (-500, -30), (-400, 10), (-300, 8), (-200, -25), (-380, -40), and (-280, -38), respectively. The facial feature coefficient of the eye can be obtained as follows:

[0048] ;

[0049] For the facial feature coefficient of the mouth, two data points h and i are calibrated on the upper contour of the mouth, two data points k and l are calibrated on the lower contour of the mouth, and two data points g and j are calibrated at the two corners of the mouth (the two data points g and j are attributed to the upper contour), and the coordinates of the data points g, h, i, j, k, and l are (-230, -720), (-100, -680), (80, -710), (230, -720), (-120, -750), and (90, -745), respectively. The facial feature coefficient of the mouth can be obtained as follows:

[0050] ;

[0051] For the facial feature coefficient of the ear, the m, n, o, and p data points are calibrated on the contour of the ear, and the coordinates of the m, o, n, and p data points are (-600, 3), (-680, -75), (-660, -133), and (-600, -176), respectively. From this, the facial feature coefficient of the ear can be obtained as follows:

[0052] ;

[0053] Then, the facial feature vector and facial feature coefficient obtained above are written into the memory unit for storage through the information data writing module.

[0054] In the later process of identifying the face of a person, the face image of the person is first collected by the face image collection unit, and the collected face image of the person is processed by the face image processing unit. In the process of processing the face image of the person by the face image processing unit, the facial information acquisition module in the recognition detection unit obtains the length and width information data of the currently captured face image of the person from the feature data processing module, such as Figure 3 As shown, the ultrasonic sensor module collects the time from the transmitter to the receiver for the ultrasonic wave at two selected points on the face. The collected information data is then sent to the recognition analysis module for processing, so that the recognition index RI of the person currently photographed can be obtained:

[0055] ;

[0056] in, L , W is the length and width of the face image of the person being recognized. L 0 , W 0 are the preset reference length and width respectively, L 0 =300mm , W 0 =150mm , t 1 , t 2 Select two points for the face to measure the time from the transmitter to the receiver. RI≥RI th , the current person is detected as an identified person, otherwise the current person is detected as a non-identified person, where, RI th =0.05 It is the preset recognition index threshold parameter.

[0057] In the process of personnel recognition, two facial images A and B are recorded, where the length and width of facial image A are 400 and 200 respectively, and the length and width of facial image B are 100 and 65 respectively. The time from the transmitter to the receiver for the selected two points of ultrasonic wave on the face of facial image A is 0.006 and 0.0058 respectively, and the time from the transmitter to the receiver for the selected two points of ultrasonic wave on the face of facial image B is 0.010 and 0.009 respectively. Therefore, the recognition index of facial image A can be obtained as follows: (400×200) / (300×150)×|170×(0.006-0.0058)|≈0.06. The recognition index of facial image B is: (100×65) / (300×150)×|170×(0.010-0.009)|≈0.02. Therefore, when recognizing a face, only facial image A is recognized, and facial image B is not recognized. In addition, when the facial photo of the person whose facial information has been collected at the beginning is placed in front of the facial image acquisition unit, the lengths of the facial images obtained are 400 and 200 respectively. Since the facial photo of the person whose facial information has been collected at the beginning is a plane, the time for the ultrasonic wave to be transmitted from the transmitter to the receiver at two points on the current face is 0.006 and 0.006 respectively. Therefore, the recognition index of the current facial image can be obtained as follows: (400×200) / (300×150)×|170×(0.006-0.006)|=0. Therefore, the face obtained by using the facial photo of the person whose facial information has been collected at the beginning is not recognized. The security performance is higher.

[0058] When processing the facial image A, the features such as eyes, nose, mouth, ears, etc. are extracted from the facial image through the facial feature extraction module, and then the attributes such as the size, position, and distance of each facial feature are determined. The geometric feature quantities are calculated by the feature vector method through the feature data processing module to obtain a feature vector that uniquely describes the face of the person, and the facial feature vector of the facial image A is compared and matched with the facial feature vector stored in the memory unit through the comparison and matching module, and the matching result is fed back to the control unit. If the match is successful, a control instruction of 1 is fed back to the control unit, and the control unit controls the device using the single-chip offline face recognition small module to turn on. Otherwise, a control instruction of 0 is fed back to the control unit to control the device using the single-chip offline face recognition small module to turn off. When the facial features of a person are partially blocked, for example, when the person wears a mask, the eyes and ears of the person are exposed. At this time, the facial feature coefficients of the eyes and ears of the current person are obtained according to the same operation method of obtaining the facial feature coefficients as above, and the facial feature coefficients of the eyes and ears of the current person are compared and matched with the facial feature coefficients in the memory unit, and the matching results are fed back to the control unit. If the match is successful, a control instruction of 1 is fed back to the control unit, and the control unit controls the device using the single-chip offline face recognition module to turn on. Otherwise, a control instruction of 0 is fed back to the control unit to control the device using the single-chip offline face recognition module to turn off. Compared with the current face recognition method, this small module can still quickly and accurately recognize the face when the face is partially blocked, and has a higher fault tolerance rate.

[0059] In addition, devices that use this single-chip offline face recognition module, such as access control terminal devices, can enable personnel to conduct voice intercoms through the voice intercom unit. The microphone module collects and records the voice of the person, which is then converted in real time through the voice recognition module and played back through the speaker, thereby realizing the voice intercom function inside and outside the access control.

[0060] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A single-chip offline face recognition module, characterized in that: It includes a face image acquisition unit, a recognition and detection unit, a face image processing unit, a memory unit, a control unit, a voice intercom unit and a single-chip microcomputer chip; Face image acquisition unit: used to collect face images of people; Recognition detection unit: used to generate the recognition index according to the facial information of the person through the following formula RI ,when RI≥RI th , the current person is detected as an identified person, otherwise the current person is detected as a non-identified person, where, RI th is a preset recognition index threshold parameter; ； in, RI is the identification index of the currently identified person, L , W is the length and width of the face image of the person being recognized. L 0 , W 0 are the preset reference length and width respectively, t 1 , t 2 Select the time for the ultrasonic wave to be emitted from the transmitter and received by the receiver at two points on the face respectively; Personnel image processing unit: used to extract the features of the face image collected by the face image acquisition unit, and generate facial feature vectors according to the feature vector method. At the same time, a two-dimensional coordinate system is established with the center point of the face image as the origin and each pixel as the coordinate axis unit. Then, a number of data points are calibrated on the face feature contour, and the coordinate parameters of each data point are recorded to generate the face feature coefficients: ； in, E is the facial feature coefficient, Δx , Δy are the differences in the horizontal and vertical coordinate values ​​between each adjacent point on the contour of the facial feature, δx , δy The facial feature vector and the facial feature coefficient are respectively the differences between the horizontal and vertical coordinate values ​​of each adjacent point under the contour of the facial feature, and the facial feature vector and the facial feature coefficient are written into the memory unit for storage; and are also used for matching the facial feature vector of the identified person with the facial feature vector stored in the memory unit when the recognition detection unit recognizes the face, and feeding back the matching result; when part of the facial features of the identified person are blocked, matching the facial feature coefficients of the unblocked person with the facial feature coefficients in the memory unit, and feeding back the matching result; Memory unit: used to store facial feature vectors and facial feature coefficients of personnel; Control unit: used to control the connected device according to the matching result of the negative feedback from the personnel image processing unit. When the matching result of the feedback is 1, the control device is turned on. When the matching result of the feedback is 0, the control device is turned off. It also controls the voice intercom unit. Voice intercom unit: used to collect the voice of personnel, recognize the voice, and then play the voice of personnel, so as to realize the two-way voice intercom function; Single-chip microcomputer chip: This chip integrates a recognition and analysis module, a feature data processing module, a comparison and matching module and a speech recognition module. The recognition and analysis module is used to help the recognition detection unit detect the recognition object. The feature data processing module is used to help the face image processing unit process the face image. The comparison and matching module is used to help the face image processing unit compare and match the facial feature vector or facial feature coefficient when recognizing the face. The speech recognition module is used to help the voice intercom module realize voice intercom.

2. The single-chip offline face recognition module according to claim 1, characterized in that: The recognition and detection unit includes an ultrasonic sensor module, a recognition and analysis module, and a facial information acquisition module; Ultrasonic sensor module: used to emit ultrasonic waves and collect the time it takes for the ultrasonic waves to be emitted from the transmitter and received by the receiver at two points selected on the face; Facial information acquisition module: used to obtain the length and width of the face of the identified person from the feature data processing module; Identification and analysis module: It is used to generate a recognition index based on the data obtained by the ultrasonic sensor module and the facial information acquisition module, and detect whether the current person is the person who needs to be identified based on the generated recognition index.

3. The single-chip offline face recognition module according to claim 1, characterized in that: The face image processing unit includes a face feature extraction module, a feature data processing module, an information data writing module and a comparison and matching module; Face feature extraction module: used to extract features of a person's face from the face image collected by the face image collection unit; Feature data processing module: used to generate facial feature vectors of personnel according to the extracted facial features of personnel through feature vector method; and generate facial feature coefficients according to the extracted facial features of personnel; Information data writing module: used to write the facial feature vector and facial feature coefficient processed by the feature data processing module into the memory unit for storage; Comparison and matching unit: used to match the facial feature vector of the identified person with the facial feature vector stored in the memory unit, and feed back the matching result. When part of the facial features of the identified person are blocked, the unblocked facial feature coefficients are matched with the facial feature coefficients in the memory unit, and the matching result is fed back.

4. The single-chip offline face recognition module according to claim 1, characterized in that: The voice intercom unit includes a microphone module, a voice recognition module and a speaker module; Microphone module: used to collect the voice of the person collecting the data and send the collected voice to the speech recognition module; Speech recognition module: used to recognize and convert the voice of the person collected by the microphone module and send it to the speaker module; Speaker module: used to convert the voice data recognized and converted by the voice recognition module into sound information and play it out.

5. The single-chip offline face recognition module according to claim 1, characterized in that: The memory size of the memory unit is 32 GB.

6. The single-chip offline face recognition module according to claim 1, characterized in that: The single-chip offline face recognition module also includes a module integrated with a power interface and a communication interface, wherein the communication interface uses the TCP / IP protocol for communication.

Images