Face recognition-based built-in ultra-narrow identification door system
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING SATISFACTION LINGHANG CURTAIN WALL DECORATION ENG CO LTD
- Filing Date
- 2026-04-07
- Publication Date
- 2026-07-03
Smart Images
Figure CN122336818A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of smart home technology, and in particular to a built-in ultra-narrow recognition door system based on facial recognition. Background Technology
[0002] The field of smart home technology encompasses a comprehensive application system centered on device interconnection, identity recognition, and access control in residential or public living spaces. This technology primarily involves the collaborative operation of building entrance management equipment, personnel identity verification equipment, information collection equipment, and access control equipment. It acquires facial image information of personnel through camera acquisition devices, confirms personnel identity through image feature extraction and identity comparison, and completes access control by combining door structure and access control devices. Its overall technical system typically includes access control equipment structural design, personnel identity collection devices, access control mechanisms, data comparison equipment, and network connection equipment, and is widely used in residential community entrances, building corridors, community entrances and exits, and home entrances for personnel access identification and management.
[0003] Among them, the built-in ultra-narrow recognition gate system based on facial recognition refers to an entrance and exit equipment structure that sets up a facial information acquisition device inside the gate structure and combines it with an access control structure for personnel identification and access management. The technical aspects involved mainly include the design of the ultra-narrow gate structure, the installation structure of the facial image acquisition device, the method of acquiring personnel facial images, the image feature information comparison process, and the gate opening control method. By installing a camera acquisition device inside the gate to acquire personnel facial image data, the feature points of the acquired facial images are extracted and compared with facial feature information in the identity database to confirm identity, and the gate is opened or closed by an electronically controlled drive mechanism in the gate, thus constituting a recognition gate system that includes a gate structure, an image acquisition device, an identity information comparison device, and a gate drive mechanism.
[0004] During operation, identification typically relies on a single image. Changes in lighting, posture, or partial facial occlusion in the entrance environment can alter the grayscale distribution of the image. The recognition process lacks a coordinated verification condition between environmental brightness and facial structural information. In scenarios with uneven lighting or rapid passage of people, grayscale changes in the facial edge area are difficult to maintain stability, resulting in discrete contour structure information. Furthermore, the recognition process lacks a combined constraint mechanism for facial contour structure and proportional relationships. Identification judgment mainly relies on single-layer image features, resulting in insufficient stability of structural information. Consequently, in scenarios with continuous passage at community entrances or building passages, fluctuations in recognition judgment are likely to occur, affecting the continuity of entrance and exit management and the reliability of passage judgment. Summary of the Invention
[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a built-in ultra-narrow recognition gate system based on face recognition.
[0006] To achieve the above objectives, the present invention adopts the following technical solution: a built-in ultra-narrow recognition gate system based on face recognition, the system comprising: The proximity detection module acquires the working information of the built-in ultra-narrow recognition door entrance infrared proximity sensor based on face recognition, the door column camera module lens, and the door channel lighting panel. It judges the proximity signal in the entrance area and acquires the image from the camera module lens, reads the brightness information of the lighting panel and compares the brightness distribution of the image, locates the facial area and retains the corresponding image content, forms a grayscale image arrangement of the facial area, and generates a grayscale image matrix of the face of the person passing through. The contour extraction module reads the grayscale distribution based on the grayscale image matrix of the passer's face, determines the obvious grayscale transition position and records the boundary coordinates, constructs the facial boundary direction according to the coordinate order, identifies the contour turning positions corresponding to the jaw edge, cheekbone edge and forehead edge, arranges the turning change order of each edge, and generates a facial contour curvature arrangement sequence. The gradient filtering module extracts the changes in the jawline, cheekbone, and forehead edges based on the facial contour curvature arrangement sequence, compares the order and continuity of curvature changes in each region, filters identity records with consistent contour change order, and merges and organizes the filtering results of the three regions to generate a contour curvature gradient identity list. The proportion construction module reads the images captured by the built-in ultra-narrow recognition gate pillar camera module based on face recognition, according to the identity list of contour curvature gradients. It identifies the center position of the left eye, the center position of the right eye, the tip of the nose, the chin, the left corner of the mouth, and the right corner of the mouth. It establishes the eye spacing relationship, the vertical relationship between the nose and chin, and the horizontal relationship between the corners of the mouth. It compares the facial proportion relationship in the identity records and filters records with consistent arrangement to generate a facial proportion matching list.
[0007] As a further aspect of the present invention, the grayscale image matrix of the passerby's face includes grayscale level distribution, grayscale density features, grayscale contrast structure, regional grayscale clustering, and pixel grayscale dispersion; the facial contour curvature arrangement sequence includes contour curvature amplitude, curvature change rhythm, curvature continuous segments, curvature transition density, and curvature morphological features; the contour curvature gradient identity list includes gradient feature identifier, gradient stability level, gradient combination features, gradient distribution pattern, and gradient correspondence index; and the facial proportion matching list includes proportion feature vector, proportion consistency index, proportion distribution pattern, proportion deviation parameter, and proportion matching index.
[0008] As a further embodiment of the present invention, the proximity detection module includes a proximity signal acquisition submodule, an illumination matching processing submodule, and a face matrix generation submodule; The proximity signal acquisition submodule acquires the infrared proximity sensor signal at the entrance of the ultra-narrow recognition door based on face recognition, the frame sequence captured by the camera module lens of the door pillar, and the working brightness of the door channel lighting panel. It detects the change range of the infrared proximity sensor signal and records the entrance trigger time marker. It performs corresponding processing according to the time sequence of the frame sequence captured by the camera module lens and the entrance trigger time marker, reads the working brightness of the lighting panel and marks the corresponding trigger time period frame image set, and generates the entrance trigger image sequence. The lighting matching processing submodule reads the grayscale distribution area of the image corresponding to the camera module lens according to the entrance trigger image sequence, collects the working brightness of the door channel lighting board and identifies the grayscale distribution area of the image, performs position correspondence recognition processing according to the grayscale distribution area of the image and the working brightness of the lighting board, determines the relationship between the grayscale change area and the preset illumination reference interval and locates the boundary position of the facial area, and generates the facial area coordinate interval. The facial matrix generation submodule reads the grayscale distribution content of the image corresponding to the camera module lens based on the facial region coordinate range, extracts the corresponding image grayscale distribution content according to the facial region coordinate range, performs grayscale arrangement processing according to the image row and column positions, performs combination and sorting according to the pixel row and column order based on the grayscale arrangement content, records the row and column grayscale distribution relationship, and generates a grayscale image matrix of the passerby's face.
[0009] As a further embodiment of the present invention, the contour extraction module includes a grayscale transition recognition submodule, a boundary orientation construction submodule, and a contour curvature arrangement submodule; The grayscale transition recognition submodule reads the pixel grayscale distribution sequence based on the grayscale image matrix of the passer's face, detects the grayscale change range of adjacent pixels in the pixel grayscale distribution sequence and records the grayscale change position, determines the correspondence between the grayscale change position and the preset grayscale transition recognition threshold and extracts the grayscale transition position coordinates, records the grayscale transition position coordinates in the row and column distribution range of the face grayscale image matrix and organizes them into a boundary positioning coordinate set to generate the face boundary coordinate range; The boundary direction construction submodule, based on the facial boundary coordinate range, reads the coordinate arrangement order, calls the coordinate arrangement order to perform continuous coordinate connection processing and records the boundary connection direction segments, determines the relationship between the boundary connection direction segments and the facial grayscale image matrix contour distribution, and identifies the mandibular edge segment, cheekbone edge segment, and forehead edge segment. It records the position of each edge segment in the coordinate arrangement order and organizes them into an edge direction set to generate a facial boundary direction sequence. The contour curvature arrangement submodule reads the corresponding coordinate positions of the mandibular edge segment, the cheekbone edge segment, and the forehead edge segment according to the facial boundary direction sequence. It calls the coordinate positions of each segment to identify contour turning points and records the coordinate arrangement order of the turning points. It detects the change relationship of the coordinate arrangement order of the turning points and organizes the change order of each edge turning point. Based on the change order of the turning points, it establishes the contour curvature arrangement relationship and generates a facial contour curvature arrangement sequence.
[0010] As a further embodiment of the present invention, the gradient filtering module includes an edge change extraction submodule, a sequential relationship filtering submodule, and an identity list generation submodule; The edge change extraction submodule reads the contour curvature distribution position and detects adjacent curvature change position segments according to the facial contour curvature arrangement sequence. It extracts the corresponding mandibular edge coordinate set, cheekbone edge coordinate set, and forehead edge coordinate set for the curvature change position segments. It records the arrangement order of each edge coordinate set in the facial contour curvature arrangement sequence and organizes them into an edge change record set to generate the contour edge change interval. The sequential relationship filtering submodule reads the corresponding arrangement positions of the jaw edge change segment, cheekbone edge change segment, and forehead edge change segment based on the contour edge change interval, detects the continuous relationship of the arrangement order of each edge change segment and records the connection position of adjacent segments, judges the correspondence between the continuous relationship of the arrangement order and the preset curvature order recognition threshold, and filters identity records with consistent arrangement order to obtain a set of records with consistent curvature order. The identity list generation submodule reads the corresponding identity identifier sequences of the mandibular edge change segment, zygomatic edge change segment, and forehead edge change segment according to the curvature order consistent record set, calls each identity identifier sequence to perform cross-merging processing and organizes the identity arrangement position relationship, detects the identity arrangement position relationship after merging and records the contour change arrangement relationship corresponding to each identity identifier, and generates a contour curvature gradient identity list.
[0011] As a further aspect of the present invention, the ratio construction module includes a feature point recognition submodule, a ratio relationship establishment submodule, and a ratio record filtering submodule; The feature point recognition submodule reads the image captured by the camera module lens of the face recognition built-in ultra-narrow recognition gate pillar based on the contour curvature gradient identity list, detects the gray-scale distribution boundary position of the image and marks the contour segment of the facial region, extracts the coordinates of the center position of the left eye, the center position of the right eye, the tip of the nose, the chin, the left corner of the mouth, and the right corner of the mouth and records them in the image coordinate arrangement sequence, organizes the coordinates of each position in the row and column distribution range of the image, and generates a set of key facial coordinates; The proportional relationship establishment submodule, based on the set of facial key coordinates, reads the coordinates of the center position of the left eye, the center position of the right eye, the tip of the nose, the chin, the left corner of the mouth, and the right corner of the mouth. It detects the spacing between the center positions of the left and right eyes and records the eye interval segment. It detects the vertical alignment of the coordinates of the tip of the nose and the chin and records the vertical segment below the nose. It detects the horizontal alignment of the coordinates of the left and right corners of the mouth and records the horizontal segment of the corners of the mouth. It generates a facial proportional relationship sequence. The proportion record filtering submodule reads the arrangement positions of the eye interval segment, the vertical segment under the nose, and the horizontal segment at the corner of the mouth according to the facial proportion relationship sequence, detects the arrangement order relationship of each segment and records the proportion arrangement position sequence, determines the arrangement relationship between the proportion arrangement position sequence and the corresponding identity record arrangement relationship in the contour curvature gradient identity list and filters identity records with the same arrangement order, organizes the filtered identity identifier arrangement sequence and records the corresponding proportion relationship position, and generates a facial proportion matching list.
[0012] As a further aspect of the present invention, the system further includes: The door control execution module reads the working information of the built-in ultra-narrow recognition door drive motor control board, door slide rail assembly, and door opening and closing sensor based on the facial proportion matching list, compares the correspondence between the identity record and the authorization record in the facial proportion matching list, adjusts the output of the drive motor control board to open the door, reads the feedback from the door opening and closing sensor and verifies the door action process, and generates the access recognition result of the built-in ultra-narrow recognition door based on facial recognition. The pass-through results of the built-in ultra-narrow recognition gate based on face recognition include identity confirmation mark, pass-through status mark, gate control response mark, recognition credibility level, and pass-through record number.
[0013] As a further embodiment of the present invention, the gate control execution module includes a device status reading submodule, an authorization relationship determination submodule, and a passage result generation submodule; The device status reading submodule reads the working signals of the face recognition-based built-in ultra-narrow recognition door drive motor control board, the position status of the door slide rail assembly, and the feedback signals of the door opening and closing sensors according to the face ratio matching list. It detects the arrangement order of the output port signals of the drive motor control board and records the sliding section of the door slide rail assembly. It collects the feedback signals of the door opening and closing sensors and marks the opening and closing status position. It organizes the corresponding arrangement relationship of the drive motor control board signals, slide rail position status, and sensor feedback signals to generate a door action status sequence. The authorization relationship determination submodule reads the facial proportion matching list identity identifier sequence based on the door action state sequence, collects and records the authorization record sequence stored by the recognition door control unit and records the identity identifier arrangement order, detects the correspondence between the facial proportion matching list identity identifier sequence and the authorization record sequence and performs identity identifier comparison processing, determines the correspondence between the identity identifier arrangement order and the authorization record order and filters authorized identity records to obtain an authorized identity record set; The passage result generation submodule reads the control signal interface of the drive motor control board and records the output port arrangement status according to the authorized identity record set, detects the correspondence between the control signal interface of the drive motor control board and the sliding section of the door slide rail assembly and executes the door opening command sending process, collects the feedback signal of the door opening and closing sensor and detects the sequence of door movement position changes, records the correspondence between the feedback signal of the door opening and closing sensor and the control signal of the drive motor control board, and generates the passage recognition result of the ultra-narrow recognition door based on face recognition.
[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the approach signal to the entrance and the brightness information of the channel lighting are combined to form the basis for environmental perception. The grayscale information of the facial region forms a stable grayscale matrix structure under the condition of brightness and darkness distribution verification. The grayscale change trajectory is further transformed into the curvature arrangement features of the jaw contour, cheekbone contour and forehead contour. The contour turning sequence and continuity relationship constitute structural gradient information. Identity structure screening is completed through the consistency of curvature in multiple regions, so that the facial contour structure has stable distinguishability. Subsequently, proportional structural features are constructed based on the eye interval relationship, the vertical relationship of the nose and chin and the horizontal relationship of the corners of the mouth, and combined with the contour structure information to form a judgment basis. This allows the identity confirmation process to have dual constraints of structural curvature features and proportional relationship features. The passage recognition process still maintains structural stability and proportional consistency in complex environments, thereby improving the reliability of entrance recognition and enhancing the stability of passage control judgment. Attached Figure Description
[0015] Figure 1 This is a system flowchart of the present invention; Figure 2 This is a flowchart of the system modules of the present invention. Detailed Implementation
[0016] The technical solution of the present invention will now be described with reference to the accompanying drawings.
[0017] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.
[0018] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.
[0019] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.
[0020] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.
[0021] Please see Figure 1 This invention provides a technical solution: a built-in ultra-narrow recognition gate system based on face recognition, the system comprising: The proximity detection module acquires the working information of the built-in ultra-narrow recognition door entrance infrared proximity sensor based on face recognition, the door column camera module lens, and the door channel lighting panel. It judges the proximity signal in the entrance area and acquires the image from the camera module lens, reads the brightness information of the lighting panel and compares the brightness distribution of the image, locates the facial area and retains the corresponding image content, forms a grayscale image arrangement of the facial area, and generates a grayscale image matrix of the face of the person passing through. The contour extraction module reads the grayscale distribution based on the grayscale image matrix of the faces of passersby, determines the obvious grayscale transition positions and records the boundary coordinates, constructs the facial boundary direction according to the coordinate order, identifies the contour turning positions corresponding to the jaw edge, cheekbone edge and forehead edge, arranges the turning change order of each edge, and generates a facial contour curvature arrangement sequence. The gradient filtering module extracts the changes in the jawline, cheekbone, and forehead edges based on the sequence of facial contour curvature. It compares the order and continuity of curvature changes in each region, filters out identity records with consistent contour change order, and merges and organizes the filtering results of the three regions to generate a contour curvature gradient identity list. The proportion construction module reads images captured by the built-in ultra-narrow recognition gate pillar camera module based on face recognition, according to the identity list of contour curvature gradients. It identifies the center position of the left eye, the center position of the right eye, the tip of the nose, the chin, the left corner of the mouth, and the right corner of the mouth. It establishes the eye spacing relationship, the vertical relationship between the nose and chin, and the horizontal relationship between the corners of the mouth. It compares the facial proportion relationship in the identity records and filters records with consistent arrangement to generate a facial proportion matching list. The gate control execution module reads the working information of the built-in ultra-narrow recognition gate drive motor control board, gate slide rail assembly, and gate opening and closing sensor based on the facial proportion matching list. It compares the correspondence between the identity record and the authorization record in the facial proportion matching list, adjusts the output of the drive motor control board to open the door, reads the feedback from the gate opening and closing sensor and verifies the gate action process, and generates the access recognition result of the built-in ultra-narrow recognition gate based on facial recognition.
[0022] The grayscale image matrix of the passerby's face includes grayscale level distribution, grayscale density features, grayscale contrast structure, regional grayscale clustering, and pixel grayscale dispersion. The facial contour curvature arrangement sequence includes contour curvature amplitude, curvature change rhythm, curvature continuous segments, curvature transition density, and curvature morphological features. The contour curvature gradient identity list includes gradient feature identifier, gradient stability level, gradient combination features, gradient distribution pattern, and gradient correspondence index. The facial proportion matching list includes proportion feature vector, proportion consistency index, proportion distribution pattern, proportion deviation parameter, and proportion matching index. The pass recognition result based on the built-in ultra-narrow recognition gate of face recognition includes identity confirmation identifier, pass status marker, gate control response identifier, recognition credibility level, and pass record number.
[0023] Please see Figure 2 The proximity detection module includes a proximity signal acquisition submodule, an illumination matching processing submodule, and a face matrix generation submodule; The proximity signal acquisition submodule acquires the infrared proximity sensor signal at the entrance of the ultra-narrow recognition door based on face recognition, the frame sequence captured by the camera module lens of the door pillar, and the working brightness of the door channel lighting panel. It detects the change range of the infrared proximity sensor signal and records the entrance trigger time marker. It performs corresponding processing according to the time sequence of the frame sequence captured by the camera module lens and the entrance trigger time marker, reads the working brightness of the lighting panel and marks the corresponding trigger time period frame image set, and generates the entrance trigger image sequence. Based on the signals from the infrared proximity sensor built into the face recognition ultra-narrow recognition door entrance, the frame sequences captured by the camera module lens on the door pillar, and the working brightness of the door passage lighting panel, a passive infrared sensor (PIR) integrated on the edge of the door frame is used to output a TTL level transition signal through a signal conditioning circuit. The level transition edge data output by the infrared proximity sensor is retrieved, and the infrared trigger level threshold is set. The input trigger start point is defined as 3.3V. When the sensor output voltage rises from 0V to 3.3V, it is recorded as the input trigger start point. With the current clock counter value at 152640, image frame data from the camera module within this clock counter interval is retrieved synchronously. A continuous set of images with frame numbers F100 to F150 is extracted, and the current sampling value corresponding to the current PWM duty cycle value of 85% fed back from the lighting board is obtained. For 1.2A, the timestamp in the header data element of each frame in the frame sequence is compared with... Perform arithmetic subtraction, selecting image frames where the absolute value of the difference is less than or equal to 16ms. For example, frame F102 has a timestamp of 152655. It is determined that it is within the matching interval, and a sorting index is performed on the marked frame image set to adjust the illumination parameters. The private data segments of each frame from F100 to F150 are written in the form of numerical tags to form an entry trigger image sequence with ambient brightness attributes.
[0024] The lighting matching processing submodule reads the grayscale distribution area of the image corresponding to the camera module lens according to the entrance trigger image sequence, collects the working brightness of the door channel lighting board and identifies the grayscale distribution area of the image, performs position correspondence recognition processing according to the grayscale distribution area of the image and the working brightness of the lighting board, determines the relationship between the grayscale change area and the preset illumination reference interval and locates the boundary position of the facial area, and generates the facial area coordinate interval. Based on the entry trigger image sequence, retrieve the grayscale value of the pixel at coordinates (x, y) in frame F105. The grayscale value is obtained by performing YUV to grayscale mapping through the ISP processor inside the camera module, and the average grayscale value is obtained by traversing all pixels in the image. The value is 145. Simultaneously, the working brightness parameter L of the lighting board is read as 800 nits. A two-dimensional pixel coordinate system is established and divided into 16x16 sub-block regions. The average grayscale of each sub-block is calculated and compared with... Perform a difference calculation; if the difference is greater than 15, mark it as a high-brightness area; if the difference is less than -15, mark it as a shadow area. Then, set the physical center coordinates of the lighting panel... Affine transformation mapping is performed with the image coordinate system to determine the projected coordinates of the center point of the illumination principal axis on the image as (320, 240), and a preset illumination reference interval is set. Given [120, 180], compare the current sub-block's average grayscale value of 155 with... Based on the inclusion relationship, perform grayscale gradient direction detection and calculate the sum of squares of grayscale differences between adjacent pixels. When the S value changes by more than 80 gray levels within 50 consecutive pixels in the horizontal direction, it is determined to be the facial edge contour line. The starting coordinates of the upper left corner of the face are determined to be (120, 80), and the ending coordinates of the lower right corner are determined to be (280, 320). The vertex values of this closed rectangular area are recorded to form the facial area coordinate range.
[0025] The face matrix generation submodule reads the grayscale distribution of the image corresponding to the camera module lens based on the coordinate range of the face region, extracts the corresponding image grayscale distribution based on the coordinate range of the face region, performs grayscale arrangement processing according to the row and column positions of the image, performs combination and sorting according to the pixel row and column order based on the grayscale arrangement content, records the row and column grayscale distribution relationship, and generates a grayscale image matrix of the face of the person passing by. Based on the facial region coordinate range, the grayscale distribution of the image corresponding to the camera module lens is retrieved. Pixel storage units with row addresses from 120 to 280 and column addresses from 80 to 320 within the coordinate range are located. The original pixel stream is retrieved from the image buffer using DMA (Direct Memory Access) technology. The 8-bit deep grayscale value corresponding to each storage unit is extracted row by row. For example, the pixel sequence extracted for row 125 is [130, 132, 135, ..., 128]. The extracted grayscale values are then mapped to a two-dimensional array structure in dynamic memory. Perform row and column position rearrangement operations, setting the matrix row index i to correspond to the image's vertical coordinate displacement and the column index j to correspond to the image's horizontal coordinate displacement, using the calculation formula. Determine the logical offset of each pixel's grayscale value within a contiguous memory block, where To extract a region with a width of 160 pixels, where i ranges from 0 to 240 and j ranges from 0 to 160, the grayscale data of each pixel is processed according to... The grayscale values are sequentially filled into the corresponding storage arrays, and the average value of each row is recorded. and standard deviation To characterize the distribution features of facial textures, such as the mean of row 10 being 142 and the standard deviation being 12, the logical encapsulation of the spatial distribution relationship of all facial pixels is finally completed, generating a grayscale image matrix of the faces of passersby.
[0026] The contour extraction module includes a grayscale transition recognition submodule, a boundary orientation construction submodule, and a contour curvature arrangement submodule; The grayscale transition recognition submodule reads the pixel grayscale distribution sequence based on the grayscale image matrix of the face of the person passing through, detects the grayscale change range of adjacent pixels in the pixel grayscale distribution sequence and records the grayscale change position, determines the correspondence between the grayscale change position and the preset grayscale transition recognition threshold and extracts the grayscale transition position coordinates, records the grayscale transition position coordinates in the row and column distribution range of the face grayscale image matrix and organizes them into a boundary positioning coordinate set, generating the face boundary coordinate range; Based on the grayscale image matrix of the faces of the passersby, the grayscale values of each pixel stored in the two-dimensional array are retrieved. Perform grayscale difference calculation on adjacent pixels in the horizontal direction i. Perform grayscale difference calculation on adjacent pixels in the vertical j-direction. Set a preset grayscale transition recognition threshold. The threshold is 45, which is set based on the typical contrast between the edge of a face and the background wall in a home entrance environment. If the contrast at a certain pixel location is... 52 or If the value is 48, the current coordinates are determined to meet the transition feature. The coordinates of the point are recorded as (150, 210) and stored in a temporary buffer. The pixel sequence from row 80 to row 320 in the image matrix is traversed. For each row of detected transition points, a continuity check is performed. If the distance between adjacent transition points is less than 1.5 pixels according to the Euclidean distance calculation, the point is included in the boundary positioning coordinate set. For example, the transition point coordinates (151, 211) are detected in row 151. Calculate distance The system determines that the point is valid and updates the set content, counts the row and column distribution range of all coordinate points in the set, determines the minimum bounding box coordinate range of the boundary in the matrix, and generates the facial boundary coordinate range.
[0027] The boundary orientation construction submodule, based on the facial boundary coordinate range, reads the coordinate arrangement order, calls the coordinate arrangement order to perform continuous coordinate connection processing and records the boundary connection orientation segments, determines the relationship between the boundary connection orientation segments and the facial grayscale image matrix contour distribution, and identifies the mandibular edge segment, cheekbone edge segment, and forehead edge segment. It records the position of each edge segment in the coordinate arrangement order and organizes them into an edge orientation set to generate a facial boundary orientation sequence. Based on the facial boundary coordinate range, the coordinate arrangement order is read, and discrete point columns are extracted from the boundary positioning coordinate set. Chain coding is then performed to logically connect the coordinate points according to their spatial proximity. The displacement vector between the current point and the predicted point is defined as follows. The angle changes of each vector segment are recorded, and the boundary line is divided into several straight line segments and arc segments. Facial anatomical statistical distribution parameters are retrieved, and the vertical coordinate interval of the mandibular edge is set to [240, 320], the vertical coordinate interval of the zygomatic edge is set to [160, 240], and the vertical coordinate interval of the forehead edge is set to [80, 160]. The distribution positions of the center points of the boundary line segments on the vertical axis are compared. When the average vertical coordinate of a continuous segment is 285, it is determined to belong to the mandibular edge segment. When a segment exhibits an outward arc-shaped expansion and has an average ordinate of 205, it is identified as a zygomatic edge segment. If the average ordinate of a segment is 110 and exhibits a smooth lateral direction, it is identified as a forehead edge segment. The starting index of the mandibular edge segment in the sequence is recorded as 0 and the ending index as 120, the starting index of the zygomatic edge segment is recorded as 121 and the ending index as 240, and the starting index of the forehead edge segment is recorded as 241 and the ending index as 360. These are then organized to form an edge direction set, generating a facial boundary direction sequence.
[0028] The contour curvature arrangement submodule reads the corresponding coordinate positions of the mandibular edge segment, zygomatic edge segment, and forehead edge segment according to the facial boundary direction sequence. It calls the coordinate positions of each segment to identify contour turning points and records the arrangement order of the turning point coordinates. It detects the change relationship of the arrangement order of the turning point coordinates and sorts out the change order of each edge turning point. Based on the change order of the turning point, it establishes the contour curvature arrangement relationship and generates the facial contour curvature arrangement sequence. Based on the facial boundary sequence, the coordinate positions of the mandibular edge segment, zygomatic edge segment, and forehead edge segment are read. A three-point local curvature calculation is performed, selecting three adjacent points in the sequence. , , The reciprocal of the radius of the circumcircle of the triangle formed by these three points is calculated as the local curvature value C, and the benchmark value for determining the inflection point is set. The baseline value is 0.8, which is set based on the statistical mean of the curvature of adult facial bones. If the calculated curvature value C at a certain position is 1.2, the position is determined to be a contour turning point and its coordinates are recorded. Three main turning points are identified in the mandibular edge area and their coordinates are recorded as [(160,310),(220,315),(280,310)], two turning points are identified in the cheekbone area and their coordinates are recorded as [(130,220),(310,220)], and four turning points are identified in the forehead area and their coordinates are recorded as [(140,90),(180,85),(260,85),(300,90)]. The coordinate order of the turning points is retrieved, and the slope change rate between adjacent turning points is calculated. ,like A value greater than 0.15 is considered a significant transition. The transition features within each segment are arranged in anatomical order from bottom to top. A composite data stream containing curvature values, transition angles, and segment labels is constructed to generate a facial contour curvature arrangement sequence.
[0029] The gradient filtering module includes an edge change extraction submodule, a sequence relationship filtering submodule, and an identity list generation submodule; The edge change extraction submodule reads the distribution position of the contour curvature according to the facial contour curvature arrangement sequence and detects adjacent curvature change position segments. It extracts the corresponding mandibular edge coordinate set, cheekbone edge coordinate set, and forehead edge coordinate set for the curvature change position segments. It records the arrangement order of each edge coordinate set in the facial contour curvature arrangement sequence and organizes them into an edge change record set to generate the contour edge change interval. Based on the facial contour curvature arrangement sequence, a composite data stream containing curvature values and turning angles is retrieved. The pixel indexes where curvature values abruptly change are located. Continuous coordinate segments with an absolute value greater than 0.05 for the first derivative of curvature are extracted. The sets of mandibular edge coordinates (vertical coordinates between 240 and 320), zygomatic edge coordinates (vertical coordinates between 160 and 240), and forehead edge coordinates (vertical coordinates between 80 and 160) are identified. Spatial topological sorting is performed on each edge coordinate set, and sorting weight coefficients are set. The weighted position parameter is obtained by multiplying the mandibular coordinate point (160, 310) with its original index value in the sequence. If the relative offset of a coordinate point in the sequence is 45 and the corresponding curvature value is 1.2, it is recorded as the core node in the edge change record set. The displacement of each key node in the mandible, cheekbone and forehead is traversed in turn. The absolute start displacement and end displacement of each segment in the facial contour curvature arrangement sequence are recorded to generate the contour edge change interval.
[0030] The sequence relationship filtering submodule reads the corresponding arrangement positions of the jaw edge change segment, cheekbone edge change segment, and forehead edge change segment based on the contour edge change interval, detects the continuous relationship of the arrangement order of each edge change segment and records the connection position of adjacent segments, judges the correspondence between the continuous relationship of the arrangement order and the preset curvature order recognition threshold, and filters identity records with consistent arrangement order to obtain a set of records with consistent curvature order. Based on the contour edge variation range, the mapping lengths of the mandibular edge, cheekbone edge, and forehead edge on the coordinate axes are read to obtain the length of the mandibular variation segment. The length of the cheekbone variation segment is 150 pixels. The length of the forehead variation segment is 120 pixels. 180 pixels; Calculate the overlap rate of adjacent segment connection locations. 100%, if the number of overlapping pixels at the junction of the jawline and cheekbone. 12 and total number of pixels The value is 270, and the calculated overlap rate is 4.4%. A preset curvature order recognition threshold is set. The threshold is 15%, set with reference to the anatomical continuity tolerance at the connection of human facial edges. Since 4.4% is less than 15%, the arrangement order of each segment is considered to have logical continuity. The known identity template sequence stored in the database is retrieved, and the Manhattan distance between the current edge change order and the template sequence is calculated. If the calculated total distance If the value is less than the set threshold of 50, it is determined that the currently collected feature matches the template identity ID 001. The identity number that meets the consistency check and its corresponding sequence index are associated and stored to obtain a curvature order consistent record set.
[0031] The identity list generation submodule reads the corresponding identity identifier sequences of the jaw edge change segment, cheekbone edge change segment, and forehead edge change segment according to the curvature order consistent record set, calls each identity identifier sequence to perform cross-merging processing and sorts out the identity arrangement position relationship, detects the identity arrangement position relationship after merging and records the contour change arrangement relationship corresponding to each identity identifier, and generates a contour curvature gradient identity list. Based on the consistent curvature order record set, the identity identifier sequence containing IDs 001, 005, and 012 is read. The contour change feature vectors corresponding to each identity identifier are retrieved, and cross-merging of the three-party data is performed. The mean mandibular curvature (1.15), mean zygomatic curvature (0.85), and mean forehead curvature (0.65) of ID 001 are extracted. An identity arrangement position relationship matrix is established. For each group of identity identifiers, contour change gradient consistency is judged, and a gradient deviation coefficient is set. The coefficient is 0.08. This coefficient is set with reference to the average perturbation deviation of facial contour extraction accuracy caused by different ambient lighting in a smart home environment, and is used to calculate the currently detected gradient vector. With template vector Euclidean distance: Comparison of D and The size relationship is such that 0.0866 is slightly greater than 0.08 but within the system's preset fault tolerance upper limit range of 0.10. It is judged as a high-confidence candidate record and retained in the sequence to be compared. If the calculated distance D is greater than 0.10, the identity removal action is performed and the current recognition cache is cleared. The feature distribution topology map of each candidate identity in the spatial coordinate system is recorded and integrated to form a structured information list with identity index, curvature gradient distribution and edge coordinate mapping relationship, and a contour curvature gradient identity list is generated.
[0032] The proportion building module includes a feature point recognition submodule, a proportion relationship establishment submodule, and a proportion record filtering submodule; The feature point recognition submodule reads the image captured by the camera module built into the face recognition ultra-narrow recognition gate based on the identity list of contour curvature gradient, detects the gray-scale distribution boundary position of the image and marks the contour segment of the facial region, extracts the coordinates of the center position of the left eye, the center position of the right eye, the tip of the nose, the chin, the left corner of the mouth, and the right corner of the mouth and records them in the image coordinate arrangement sequence, sorts out the distribution range of each position coordinate in the row and column of the image, and generates a set of key facial coordinates; Based on the identity list of contour curvature gradients, retrieve the infrared trigger time marker of the camera module lens. The original image data acquired in the corresponding frame sequence is subjected to full-image pixel grayscale stretching to obtain an image matrix with enhanced brightness and contrast. The search window for the facial contour segment is set to 120 to 280 pixels horizontally and 80 to 320 pixels vertically. Local binarization is performed to extract feature regions. The coordinates of the left eye center (165, 120), right eye center (235, 120), nose tip (200, 185), chin (200, 290), left corner of mouth (175, 230), and right corner of mouth (225, 230) are identified. The absolute row and column address index values of each coordinate point in the image storage space are recorded. The extreme values of the feature points in the horizontal x and vertical y directions are statistically analyzed. The coordinates of the left eye (165, 120) and the right eye (235, 120) are stored at the beginning of the coordinate arrangement sequence. The coordinates of the mouth and nose are arranged in ascending order vertically to generate a set of facial key coordinates.
[0033] The proportional relationship establishment submodule, based on the set of key facial coordinates, reads the coordinates of the center position of the left eye, the center position of the right eye, the tip of the nose, the chin, the left corner of the mouth, and the right corner of the mouth. It detects the spacing between the center positions of the left and right eyes and records the eye interval segment, detects the vertical alignment of the coordinates of the tip of the nose and the chin and records the vertical segment below the nose, and detects the horizontal alignment of the coordinates of the left and right corners of the mouth and records the horizontal segment of the corners of the mouth, generating a facial proportional relationship sequence. Based on the set of facial key coordinates, the coordinates of the left eye center (165, 120) and the right eye center (235, 120) are retrieved and Euclidean distance is calculated. The eye interval segment value is obtained as 70. The vertical coordinate difference calculation is performed between the nose tip coordinates (200, 185) and the chin coordinates (200, 290). The vertical segment below the nose is obtained as 105. The coordinates of the left corner of the mouth (175, 230) and the right corner of the mouth (225, 230) are retrieved and the horizontal coordinate difference is calculated. The horizontal segment value of the corner of the mouth is set to 50. A sequence of ratios between the values of each segment is established, and the ratio of the distance between the eyes to the height below the nose is calculated. Calculate the ratio of mouth width to eye distance. The execution logic is encapsulated according to the anatomical structure hierarchy of the eyes, nose, and mouth, and the floating-point representation of each ratio parameter in memory is recorded to generate a sequence of facial proportions.
[0034] The proportion record filtering submodule reads the arrangement positions of the eye interval segment, the vertical segment under the nose, and the horizontal segment of the corner of the mouth according to the facial proportion relationship sequence. It detects the arrangement order relationship of each segment and records the proportion arrangement position sequence. It judges the arrangement relationship between the proportion arrangement position sequence and the corresponding identity record arrangement in the contour curvature gradient identity list and filters identity records with the same arrangement order. It organizes the filtered identity identifier arrangement sequence and records the corresponding proportion relationship position to generate a facial proportion matching list. Based on the facial proportion sequence, the positions of the eye interval segment, the vertical segment below the nose, and the horizontal segment at the corner of the mouth are read. Feature values such as 0.667 and 0.714 are retrieved. The preset proportion benchmark value with ID 001 is retrieved from the contour curvature gradient identity list, and a threshold for determining consistent proportion arrangement is set. The threshold is 0.05, which is set with reference to the physiological fluctuation range of feature point displacements of household members under different facial expressions, and the current ratio is applied. Ratio to benchmark Absolute value operation of difference Since 0.017 is less than 0.05, the eye-nose ratio is deemed to be correct. Execute the current ratio Ratio to benchmark Absolute value operation of difference Since 0.014 is less than 0.05, the mouth corner ratio is considered to be consistent. A logical AND operation is performed on all feature ratios for verification. If all ratios are within the consistency judgment range, the identity identifier ID 001 is extracted and its physical storage location in the memory index is marked. The dynamic feature distribution map and time series attributes corresponding to the identity identifier are recorded, and a facial ratio matching list is generated.
[0035] The gate control execution module includes a device status reading submodule, an authorization relationship determination submodule, and a passage result generation submodule; The device status reading submodule reads the working signals of the drive motor control board of the face recognition-based ultra-narrow recognition door, the position status of the door slide rail assembly, and the feedback signals of the door opening and closing sensors, based on the face ratio matching list. It detects the arrangement order of the output port signals of the drive motor control board and records the sliding section of the door slide rail assembly. It collects the feedback signals of the door opening and closing sensors and marks the opening and closing status position. It organizes the corresponding arrangement relationship of the drive motor control board signals, slide rail position status, and sensor feedback signals to generate a door action status sequence. Based on the facial proportion matching list, the system reads the working signals of the face recognition-based built-in ultra-narrow recognition door drive motor control board, the position status of the door slide rail assembly, and the feedback signals of the door opening and closing sensors. It then retrieves the pulse width modulation (PWM) signal output from the drive motor control board's I / O port, reads the current duty cycle value as 0% to determine that the motor is in a stationary locked state, and retrieves the absolute displacement encoder value of the door slide rail assembly. The displacement is 0mm. This displacement is obtained by calculating the pulse count value fed back by the high-precision photoelectric rotary encoder through the ARM core processor. By reading the low-level signal 0V output by the door opening and closing sensor, it is determined that the door is currently in a fully closed position. The door travel reference range is set to 0 to 800mm. The signal level of the enable terminal of the drive motor control board is detected to be high level 3.3V. The starting segment of the zero point coordinate of the door slide rail assembly on the guide rail is recorded. The value of the lock tongue entering the groove fed back by the sensor is obtained as 50N. The level signal, displacement value and pressure parameter are logically connected to establish a feature data mapping table including motor standby, slide rail home position and sensor lock state. The position order of each signal in the memory address is recorded to generate the door action state sequence.
[0036] The authorization relationship determination submodule reads the identity identifier sequence of the face ratio matching list based on the door action state sequence, collects and recognizes the authorization record sequence stored by the door control unit and records the identity identifier arrangement order, detects the correspondence between the identity identifier sequence of the face ratio matching list and the authorization record sequence and performs identity identifier comparison processing, determines the correspondence between the identity identifier arrangement order and the authorization record order and filters the authorized identity records to obtain the authorized identity record set; Based on the door action state sequence, read the facial proportion matching list identity identifier sequence, retrieve the pre-stored authorization record sequence in the EEPROM memory of the recognition door control unit, extract the authorized identity ID list [001,002,005], read the currently recognized identity identifier ID 001 in the facial proportion matching list, and perform a logical comparison operation between the current identity identifier and the authorization record list; Calculate the comparison difference ,when If a match is successfully determined, and the difference between the matches is not zero, the system automatically resets the current frame recognition counter and sends an unauthorized alarm log to the main controller, and sets the authorization weight coefficient. The coefficient is set to 1.0, which is based on the highest permission level in the smart home security level. For visitors with ID 001, permission level verification is performed. The current clock signal is retrieved to read the current authorized access period. The successfully matched identity identifier and its corresponding biometric sampling data are hashed and associated. The index offset in the authorization list is recorded. The door opening delay parameter for the corresponding ID is extracted to be 5 seconds. The authorized identity information that meets the conditions is stored in the pending execution buffer to form a valid access record with a unique verification code, thus obtaining the authorized identity record set.
[0037] The passage result generation submodule reads the control signal interface of the drive motor control board and records the output port arrangement status according to the authorized identity record set, detects the correspondence between the control signal interface of the drive motor control board and the sliding section of the door slide rail assembly and executes the door opening command sending process, collects the feedback signal of the door opening and closing sensor and detects the sequence of door movement position changes, records the correspondence between the feedback signal of the door opening and closing sensor and the control signal of the drive motor control board, and generates the passage recognition result based on the face recognition built-in ultra-narrow recognition door; Based on the authorized identity record set, read the control signal interface of the drive motor control board and record the output port arrangement status. Retrieve the STEP step pulse pin and the DIR direction control pin, set the DIR pin level to a high level of 3.3V to determine the door opening direction, and calculate the total number of pulses required for door opening. The total travel of the door body The displacement is 800mm per pulse. Given a depth of 0.1mm, N=8000 is calculated. The control interface is invoked to send a sequence of 8000 square waves at a frequency of 2kHz. The frequency change of the Hall level feedback from the door opening / closing sensor from 0V to 3.3V is monitored. The real-time coordinate distribution sequence of the door slide rail assembly during its movement from 0mm to 800mm is recorded. The absolute value of the difference between the number of pulses sent in the command and the number of pulses fed back by the sensor is compared to see if it is less than the set out-of-step threshold. If the actual feedback pulse is 7985, calculate 50. If the value is less than 50, the action is considered complete. The motor stop signal, the fully open position of the door, the authorized identity ID, and the passage record are then encapsulated in a structured manner to generate the passage recognition result of the ultra-narrow recognition door with built-in face recognition.
[0038] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A built-in type very narrow recognition door system based on face recognition, characterized by, The system includes: The proximity detection module acquires the working information of the built-in ultra-narrow recognition door entrance infrared proximity sensor based on face recognition, the door column camera module lens, and the door channel lighting panel. It judges the proximity signal in the entrance area and acquires the image from the camera module lens, reads the brightness information of the lighting panel and compares the brightness distribution of the image, locates the facial area and retains the corresponding image content, forms a grayscale image arrangement of the facial area, and generates a grayscale image matrix of the face of the person passing through. The contour extraction module reads the grayscale distribution based on the grayscale image matrix of the passer's face, determines the obvious grayscale transition position and records the boundary coordinates, constructs the facial boundary direction according to the coordinate order, identifies the contour turning positions corresponding to the jaw edge, cheekbone edge and forehead edge, arranges the turning change order of each edge, and generates a facial contour curvature arrangement sequence. The gradient filtering module extracts the changes in the jawline, cheekbone, and forehead edges based on the facial contour curvature arrangement sequence, compares the order and continuity of curvature changes in each region, filters identity records with consistent contour change order, and merges and organizes the filtering results of the three regions to generate a contour curvature gradient identity list. The proportion construction module reads the images captured by the built-in ultra-narrow recognition gate pillar camera module based on face recognition, according to the identity list of contour curvature gradients. It identifies the center position of the left eye, the center position of the right eye, the tip of the nose, the chin, the left corner of the mouth, and the right corner of the mouth. It establishes the eye spacing relationship, the vertical relationship between the nose and chin, and the horizontal relationship between the corners of the mouth. It compares the facial proportion relationship in the identity records and filters records with consistent arrangement to generate a facial proportion matching list.
2. The built-in ultra-narrow recognition gate system based on face recognition according to claim 1, characterized in that: The grayscale image matrix of the passerby's face includes grayscale level distribution, grayscale density features, grayscale contrast structure, regional grayscale clustering, and pixel grayscale dispersion. The facial contour curvature arrangement sequence includes contour curvature amplitude, curvature change rhythm, curvature continuous segments, curvature transition density, and curvature morphological features. The contour curvature gradient identity list includes gradient feature identifier, gradient stability level, gradient combination features, gradient distribution pattern, and gradient correspondence index. The facial proportion matching list includes proportion feature vector, proportion consistency index, proportion distribution pattern, proportion deviation parameter, and proportion matching index.
3. The built-in ultra-narrow recognition gate system based on face recognition according to claim 1, characterized in that: The proximity detection module includes a proximity signal acquisition submodule, an illumination matching processing submodule, and a face matrix generation submodule; The proximity signal acquisition submodule acquires the infrared proximity sensor signal at the entrance of the ultra-narrow recognition door based on face recognition, the frame sequence captured by the camera module lens of the door pillar, and the working brightness of the door channel lighting panel. It detects the change range of the infrared proximity sensor signal and records the entrance trigger time marker. It performs corresponding processing according to the time sequence of the frame sequence captured by the camera module lens and the entrance trigger time marker, reads the working brightness of the lighting panel and marks the corresponding trigger time period frame image set, and generates the entrance trigger image sequence. The lighting matching processing submodule reads the grayscale distribution area of the image corresponding to the camera module lens according to the entrance trigger image sequence, collects the working brightness of the door channel lighting board and identifies the grayscale distribution area of the image, performs position correspondence recognition processing according to the grayscale distribution area of the image and the working brightness of the lighting board, determines the relationship between the grayscale change area and the preset illumination reference interval and locates the boundary position of the facial area, and generates the facial area coordinate interval. The facial matrix generation submodule reads the grayscale distribution content of the image corresponding to the camera module lens based on the facial region coordinate range, extracts the corresponding image grayscale distribution content according to the facial region coordinate range, performs grayscale arrangement processing according to the image row and column positions, performs combination and sorting according to the pixel row and column order based on the grayscale arrangement content, records the row and column grayscale distribution relationship, and generates a grayscale image matrix of the passerby's face.
4. The built-in ultra-narrow recognition gate system based on face recognition according to claim 1, characterized in that: The contour extraction module includes a grayscale transition recognition submodule, a boundary orientation construction submodule, and a contour curvature arrangement submodule. The grayscale transition recognition submodule reads the pixel grayscale distribution sequence based on the grayscale image matrix of the passer's face, detects the grayscale change range of adjacent pixels in the pixel grayscale distribution sequence and records the grayscale change position, determines the correspondence between the grayscale change position and the preset grayscale transition recognition threshold and extracts the grayscale transition position coordinates, records the grayscale transition position coordinates in the row and column distribution range of the face grayscale image matrix and organizes them into a boundary positioning coordinate set to generate the face boundary coordinate range; The boundary direction construction submodule, based on the facial boundary coordinate range, reads the coordinate arrangement order, calls the coordinate arrangement order to perform continuous coordinate connection processing and records the boundary connection direction segments, determines the relationship between the boundary connection direction segments and the facial grayscale image matrix contour distribution, and identifies the mandibular edge segment, cheekbone edge segment, and forehead edge segment. It records the position of each edge segment in the coordinate arrangement order and organizes them into an edge direction set to generate a facial boundary direction sequence. The contour curvature arrangement submodule reads the corresponding coordinate positions of the mandibular edge segment, the cheekbone edge segment, and the forehead edge segment according to the facial boundary direction sequence. It calls the coordinate positions of each segment to identify contour turning points and records the coordinate arrangement order of the turning points. It detects the change relationship of the coordinate arrangement order of the turning points and organizes the change order of each edge turning point. Based on the change order of the turning points, it establishes the contour curvature arrangement relationship and generates a facial contour curvature arrangement sequence.
5. The built-in ultra-narrow recognition gate system based on face recognition according to claim 1, characterized in that: The gradient filtering module includes an edge change extraction submodule, a sequential relationship filtering submodule, and an identity list generation submodule; The edge change extraction submodule reads the contour curvature distribution position and detects adjacent curvature change position segments according to the facial contour curvature arrangement sequence. It extracts the corresponding mandibular edge coordinate set, cheekbone edge coordinate set, and forehead edge coordinate set for the curvature change position segments. It records the arrangement order of each edge coordinate set in the facial contour curvature arrangement sequence and organizes them into an edge change record set to generate the contour edge change interval. The sequential relationship filtering submodule reads the corresponding arrangement positions of the jaw edge change segment, cheekbone edge change segment, and forehead edge change segment based on the contour edge change interval, detects the continuous relationship of the arrangement order of each edge change segment and records the connection position of adjacent segments, judges the correspondence between the continuous relationship of the arrangement order and the preset curvature order recognition threshold, and filters identity records with consistent arrangement order to obtain a set of records with consistent curvature order. The identity list generation submodule reads the corresponding identity identifier sequences of the mandibular edge change segment, zygomatic edge change segment, and forehead edge change segment according to the curvature order consistent record set, calls each identity identifier sequence to perform cross-merging processing and organizes the identity arrangement position relationship, detects the identity arrangement position relationship after merging and records the contour change arrangement relationship corresponding to each identity identifier, and generates a contour curvature gradient identity list.
6. The built-in ultra-narrow recognition gate system based on face recognition according to claim 1, characterized in that: The ratio construction module includes a feature point recognition submodule, a ratio relationship establishment submodule, and a ratio record filtering submodule; The feature point recognition submodule reads the image captured by the camera module lens of the face recognition built-in ultra-narrow recognition gate pillar based on the contour curvature gradient identity list, detects the gray-scale distribution boundary position of the image and marks the contour segment of the facial region, extracts the coordinates of the center position of the left eye, the center position of the right eye, the tip of the nose, the chin, the left corner of the mouth, and the right corner of the mouth and records them in the image coordinate arrangement sequence, organizes the coordinates of each position in the row and column distribution range of the image, and generates a set of key facial coordinates; The proportional relationship establishment submodule, based on the set of facial key coordinates, reads the coordinates of the center position of the left eye, the center position of the right eye, the tip of the nose, the chin, the left corner of the mouth, and the right corner of the mouth. It detects the spacing between the center positions of the left and right eyes and records the eye interval segment. It detects the vertical alignment of the coordinates of the tip of the nose and the chin and records the vertical segment below the nose. It detects the horizontal alignment of the coordinates of the left and right corners of the mouth and records the horizontal segment of the corners of the mouth. It generates a facial proportional relationship sequence. The proportion record filtering submodule reads the arrangement positions of the eye interval segment, the vertical segment under the nose, and the horizontal segment at the corner of the mouth according to the facial proportion relationship sequence, detects the arrangement order relationship of each segment and records the proportion arrangement position sequence, determines the arrangement relationship between the proportion arrangement position sequence and the corresponding identity record arrangement relationship in the contour curvature gradient identity list and filters identity records with the same arrangement order, organizes the filtered identity identifier arrangement sequence and records the corresponding proportion relationship position, and generates a facial proportion matching list.
7. The built-in ultra-narrow recognition gate system based on face recognition according to claim 1, characterized in that: The system also includes: The door control execution module reads the working information of the built-in ultra-narrow recognition door drive motor control board, door slide rail assembly, and door opening and closing sensor based on the facial proportion matching list, compares the correspondence between the identity record and the authorization record in the facial proportion matching list, adjusts the output of the drive motor control board to open the door, reads the feedback from the door opening and closing sensor and verifies the door action process, and generates the access recognition result of the built-in ultra-narrow recognition door based on facial recognition. The pass-through results of the built-in ultra-narrow recognition gate based on face recognition include identity confirmation mark, pass-through status mark, gate control response mark, recognition credibility level, and pass-through record number.
8. The built-in ultra-narrow recognition gate system based on face recognition according to claim 7, characterized in that: The gate control execution module includes a device status reading submodule, an authorization relationship determination submodule, and a passage result generation submodule; The device status reading submodule reads the working signals of the face recognition-based built-in ultra-narrow recognition door drive motor control board, the position status of the door slide rail assembly, and the feedback signals of the door opening and closing sensors according to the face ratio matching list. It detects the arrangement order of the output port signals of the drive motor control board and records the sliding section of the door slide rail assembly. It collects the feedback signals of the door opening and closing sensors and marks the opening and closing status position. It organizes the corresponding arrangement relationship of the drive motor control board signals, slide rail position status, and sensor feedback signals to generate a door action status sequence. The authorization relationship determination submodule reads the facial proportion matching list identity identifier sequence based on the door action state sequence, collects and records the authorization record sequence stored by the recognition door control unit and records the identity identifier arrangement order, detects the correspondence between the facial proportion matching list identity identifier sequence and the authorization record sequence and performs identity identifier comparison processing, determines the correspondence between the identity identifier arrangement order and the authorization record order and filters authorized identity records to obtain an authorized identity record set; The passage result generation submodule reads the control signal interface of the drive motor control board and records the output port arrangement status according to the authorized identity record set, detects the correspondence between the control signal interface of the drive motor control board and the sliding section of the door slide rail assembly and executes the door opening command sending process, collects the feedback signal of the door opening and closing sensor and detects the sequence of door movement position changes, records the correspondence between the feedback signal of the door opening and closing sensor and the control signal of the drive motor control board, and generates the passage recognition result of the ultra-narrow recognition door based on face recognition.