Big data-based face biological recognition method

Abstract

The invention discloses a face biological recognition method based on big data, and relates to the technical field of data processing. Face image data of a person is collected through a camera installed indoors, and voice data of the person is collected through microphone equipment. According to the face biological recognition method based on the big data, recognition operation is carried out on image data, matching is carried out according to dynamic change of a face and set data of a historical database, recognition operation is carried out on voice data in parallel, and matching is carried out according to expression change of the voice data and the set data of the historical database; and then assignment fusion of the image data and the voice data is realized to determine the state of the person, the abnormal person condition is positioned by voice, and the state of the person is determined by combining the decibel of the voice and the abnormal condition of the facial features of the person, so that the evaluation value of the learning state of the person is better evaluated, the condition can be more quickly found and timely processed, and the user experience is improved. And the learning quality in the education process is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and specifically provides a face biometric recognition method based on big data. Background Art

[0002] In indoor scenarios, accurately obtaining the face status of personnel plays a crucial role in improving learning effects. With the rapid development of big data and artificial intelligence technologies, existing face recognition methods still have problems such as slow recognition speed and low accuracy when dealing with large-scale face data.

[0003] Refer to the patent with the title: An Image Recognition Method Based on Big Data (Patent Publication No.: CN115630180A, Patent Publication Date: January 20, 2023), which includes an image acquisition module. The output end of the image acquisition module is connected to an image preprocessing module and a GPS positioning module. The output end of the image preprocessing module is connected to a feature extraction module. The output end of the feature extraction module is connected to a main processor module. The output end of the main processor module is connected to a classification mode selection module, which can manually or automatically classify the target image according to the user. According to the feature values of the target image, the target image is matched and compared with the images in the image database or the images indexed by big data, and then the user can manually or automatically classify the type of the target image, thereby improving the classification accuracy of the target image, reducing the classification workload of the target image, and making the classification efficiency of the target image higher.

[0004] Based on the description of the above document, in the existing face biometric recognition operations, especially in the teaching process, in order to better improve the teaching quality, the specific learning situation of personnel during class cannot be continuously monitored. Especially when there are situations such as whispering and dozing off, the subtle facial changes cannot be detected in time. These problems will cause the corresponding guiding strategies to fail to respond to the operations of personnel. Therefore, the present invention provides a face biometric recognition method based on big data. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a face biometric recognition method based on big data. In the existing face biometric recognition operations, especially in the teaching process, in order to better improve the teaching quality, the specific learning situation of personnel during class cannot be continuously monitored. Especially when there are situations such as whispering and dozing off, the subtle facial changes cannot be detected in time. These problems will cause the corresponding guiding strategies to fail to respond to the operations of personnel.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A face biometric recognition method based on big data, specifically including the following steps:

[0007] A1. Collect the facial image data of personnel through a camera, and collect the voice data of personnel using a microphone device. Then, transmit the collected data to a processing system through wireless communication technology;

[0008] A2. Use the processing system to analyze and process the image data and voice data, and realize state recognition by combining them. The specific processing operations are as follows:

[0009] a21. Realize the screening operation of the image data and voice data, and extract the required data for subsequent processing;

[0010] a22. Perform recognition operations on the image data and voice data. Realize recognition operations on the voice data. Match the expression changes of the voice data with the set data in the historical database to determine the sound source and identify the corresponding personnel. And match the dynamic changes of the face with the set data in the historical database, and then realize the assignment and fusion of the image data and voice data to determine the personnel state;

[0011] a23. Transmit the determined personnel state to the display terminal and display it in a combination of text and images;

[0012] A3. Implement adaptive strategy adjustment operations according to the results.

[0013] Preferably, in step a21, the screening operation of the image data and voice data is realized;

[0014] B1. Receive the image data, and realize the recognition of the characteristic part in the image data, mark the seat characteristics, retain the image with the characteristics of the front side of the seat, and splice the image data to form an overall seat map;

[0015] B2. Then, trace back to the acquisition camera according to the corresponding overall seat map data, and retain the subsequent image data collected by the acquisition camera;

[0016] B3. Receive the voice data, realize the noise reduction processing of the voice data, and truncate the voice data to retain only the voice segments with sound.

[0017] Preferably, the operation of splicing to form an overall seat map in step B1 is as follows:

[0018] b11. Extract the image containing the seat characteristics, and determine the size of the adjacent seat characteristics near the center of the image;

[0019] b12. Connect the midpoints of the left and right sides of the image, and determine the intersection points of the connection line and the marked adjacent seat characteristics as a1, a2, a3, and a4 from left to right. And when the distance between a1 and a2 is equal to the distance between a3 and a4, the current image data is retained;

[0020] b13. Then, perform the image stitching operation according to the order and direction of image acquisition. First, determine the number of seat features with equal distances between adjacent image data, and use the seat features with equal distances at the end of the first image data as a reference. Then, determine the seat features corresponding to the reference seat features in the second image data, and perform an alignment and covering operation at the corresponding seat features. The common part where the second image data covers the first image data is made transparent until the stitching of the remaining images is completed.

[0021] Preferably, the operation of matching the a22 according to the expression changes of the voice data and the set data in the historical database is as follows:

[0022] D1. Based on the personnel features to be recognized, use the mouth features of the personnel as the sound generation point, and convert the indoor space into a space coordinate system. Determine the coordinate of the generation point as (x0, y0, z0), and determine the coordinates of the positions of each microphone device as (x n 、y n 、z n );

[0023] D2. First, measure the distances between each microphone device and the generation point, and the microphone device with the smallest separation distance is used for the recognition of the current personnel's language data;

[0024] And the distance calculation formula between the microphone device and the generation point is:

[0025]

[0026] H is the distance between the microphone device and the generation point, x n 、y n 、z n are the vertical distances from the position of the microphone device to the coordinate axes of the space coordinate system respectively;

[0027] D3. And determine and mark the position of the corresponding personnel in combination with the direction source of the sound generation, the decibel level of the sound, and the distance between the microphone device and the generation point.

[0028] Preferably, the operation of matching the a22 according to the dynamic changes of the face and the set data in the historical database is as follows:

[0029] C1. Based on the determined and marked position of the corresponding personnel, find the subsequent image data of the corresponding seat map for feature extraction operation. Only the image data with portrait features is retained in this extraction;

[0030] C2. Search for the positioning points on the image data with portrait features, and determine the deviation distance between the cross cursor of the camera and the positioning points. If the deviation distance meets the set deviation threshold, the image data is retained;

[0031] C3. Extract the setting data about the dynamic changes of portrait features based on the historical database, perform pixel conversion on the image data retained in the C2 operation, convert it into the same pixels as the setting data, and determine the status result based on the situation after the retained image data matches the setting data.

[0032] Preferably, the operation of finding the positioning point in C2 is:

[0033] c201, the upper line segment is formed by connecting the symmetrical eye corners in the portrait features, and the lower line segment is formed by connecting the two sides of the mouth corners in the portrait features;

[0034] c202. Connect the midpoints of the upper line segment and the lower line segment to form a vertical line segment. The midpoint of the vertical line segment is the positioning point of the portrait feature.

[0035] Preferably, the comparison operation of the deviation distance and the deviation threshold in C2 is:

[0036] c221. Establish a coordinate axis based on the image with the positioning point as the center point, establish the Y axis in the direction of the vertical line segment, and establish the X axis starting from the center point and at a right angle to the Y axis;

[0037] c222. At this time, the coordinates of the positioning point are (0, 0), and the coordinates of the cross cursor of the camera based on the XY coordinate axis are (x, y), and the corresponding deviation distance is calculated based on the coordinate values;

[0038] c223. Extract the deviation threshold in the historical data and mark it as L, and compare the deviation distance and the deviation threshold.

[0039] Preferably, the formula for calculating the deviation distance in c222 is:

[0040]

[0041] S is the distance from the camera's crosshairs to the positioning point, i.e., the deviation distance. x and y correspond to the vertical distances from the Y axis and X axis, respectively.

[0042] The comparison results of the deviation distance and the deviation threshold are:

[0043] If S≤L, the deviation distance meets the set deviation threshold, and the image data is retained;

[0044] If S>L, the deviation distance exceeds the set deviation threshold, and the image data is removed. When a certain portrait feature cannot be collected, the camera's collection position is adjusted so that the cross cursor corresponds to the positioning point to be collected, thereby retaining the image data.

[0045] Preferably, the operation of determining the status result according to the matching situation between the retained image data and the set data in C3 is as follows:

[0046] c31. After matching the retained image data with the set data, determine the status of the mouth feature actions, head feature postures, and eye feature status of the person in the image data;

[0047] c32. Based on the mouth feature in the continuous image data, compare the distance between the midpoint of the lower edge of the upper lip feature and the midpoint of the upper edge of the lower lip feature. If the distance changes continuously, the status is abnormal;

[0048] Based on the head feature posture in the continuous image data, compare the direction of the vertical line segment with the vertical central axis in the image data. If the deviation angle between the two changes continuously, the status is abnormal;

[0049] Based on the eye feature of the person in the continuous image data, compare the distance between the midpoint of the lower edge of the eyelid feature and the midpoint of the upper edge of the lower eyelid feature. If the distance is continuously 0, the status is abnormal;

[0050] c33. Obtain the current status result of the person according to the specific situation after feature matching.

[0051] Preferably, the operation of realizing the assignment fusion of image data and voice data to determine the status of the person in a22 is as follows:

[0052] Perform assignment fusion on the facial feature status and voice expression status results corresponding to each time period, and the expression is:

[0053] F = w×(p / R)+v×(q / T);

[0054] F is the evaluation value after the assignment fusion of a single person based on image data and voice data, w is the weight ratio based on image data, v is the weight ratio based on voice data, p is the number of abnormal images of the facial feature during the image data acquisition time period, q is the number of recognized times during the voice data acquisition time period, R is the number of acquisitions during the facial feature acquisition time period, and T is the number of acquisitions during the voice data acquisition time period;

[0055] Set the over-standard threshold of the evaluation value as K. Therefore, when the evaluation value F < K, the current status of the person meets the requirements, and when the evaluation value F ≥ K, the current status of the person does not meet the requirements and needs to be adjusted adaptively.

[0056] The present invention provides a face biometric recognition method based on big data. Compared with the prior art, it has the following beneficial effects:

[0057] 1. The face biometric recognition method based on big data performs recognition operations on image data, matches the dynamic changes of the face with the set data in the historical database, and concurrently performs recognition operations on voice data, matches the expression changes of the voice data with the set data in the historical database, and then realizes the assignment and fusion of image data and voice data to determine the personnel status, and realizes the situation of abnormal personnel in voice positioning. At the same time, it combines the decibel of the voice and the abnormal situation of the facial features of the personnel to determine the status of the personnel, so as to better evaluate the evaluation value of the personnel's learning status, facilitate the educator to discover the situation faster and handle it in time, and improve the learning quality in the education process.

[0058] 2. The face biometric recognition method based on big data performs feature extraction operations on the subsequent image data according to the corresponding seat map. This extraction only retains the image data with portrait features. Locate points on the image data based on portrait features, and determine the deviation distance between the cross cursor of the camera and the positioning point. If the deviation distance meets the set deviation threshold, the image data is retained, and three-level processing of the image data is realized. First, retain the image with the characteristics of the front side of the seat, then retain the image data with portrait features, and finally retain the image data where the camera cross cursor and the portrait feature positioning point meet the requirements. This not only reduces the processing of the image data volume, but also ensures that the data for judging the personnel status is more accurate.

[0059] 3. The face biometric recognition method based on big data assigns and fuses the facial feature status and voice expression status results for each time period, and obtains the evaluation value of a single person after the assignment and fusion of image data and voice data, so as to effectively judge whether there are abnormal situations during the learning process of the personnel. At the same time, it improves the error tolerance rate, so that the strategy adjustment will not be immediately generated during the discovery of similar situations, and it is also convenient to more accurately judge the personnel status. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is the operation flow chart of the face biometric recognition method of the present invention;

[0061] Figure 2 is the screening operation flow chart of the image data and voice data of the present invention;

[0062] Figure 3 is the matching operation flow chart of the facial dynamic changes of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0063] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0064] Please refer to Figures 1-3 , the present invention provides a technical solution:

[0065] A face biometric recognition method based on big data specifically includes the following steps:

[0066] A1. Collect the facial image data of personnel through a camera installed indoors, and collect the voice data of personnel by using a microphone device, and then transmit the collected data to the processing system through wireless communication technology;

[0067] A2. Use the processing system to analyze and process the image data and voice data, and realize state recognition in combination. The specific processing operations are as follows:

[0068] a21. Realize the screening operation of the image data and voice data, and extract the required data for subsequent processing;

[0069] a22. Perform recognition operations on the image data and voice data respectively. Match the expression changes of the voice data with the set data in the historical database to determine the corresponding personnel of the sound source, and match the dynamic changes of the face with the set data in the historical database, and then realize the assignment and fusion of the image data and voice data to determine the personnel status;

[0070] a23. Transmit the determined personnel status to the display terminal and display it in a combination of text and images;

[0071] A3. Implement an adaptive strategy adjustment operation according to the result.

[0072] Among them, by performing recognition operations on the image data, matching according to the dynamic changes of the face with the set data in the historical database, simultaneously performing recognition operations on the voice data, matching the expression changes of the voice data with the set data in the historical database, and then realizing the assignment and fusion of the image data and voice data to determine the personnel status, and realizing the situation of abnormal voice positioning of personnel. At the same time, combining the decibel of the voice and the abnormal situation of the personnel's facial features to determine the personnel's status, so as to better evaluate the evaluation value of the personnel's learning status, facilitate faster discovery of the situation and timely processing, and improve the learning quality in the education process.

[0073] In the embodiment of the present invention, in a21, the screening operation of the image data and voice data is realized;

[0074] B1. Receive image data, identify the feature parts in the image data, mark the seat features, retain the images with the front-side features of the seat, and splice the image data to form an overall seat map;

[0075] B2. Then trace back to the acquisition camera according to the corresponding overall seat map data, and retain the subsequent image data collected by the acquisition camera;

[0076] B3. Receive voice data, perform noise reduction processing on the voice data, and truncate the voice data to retain only the voice segments with sound.

[0077] The noise reduction processing of the voice data is an existing mature technology, and the truncation processing of the voice data is based on judging whether there is an expression of text content in the voice data. After there is an expression of text content, the previous voice segment is cut to remove the voice without text content and retain the voice segment with text content and voice.

[0078] In the embodiment of the present invention, the operation of splicing to form an overall seat map in B1 is as follows:

[0079] b11. Extract the images containing seat features, and determine the sizes of adjacent seat features near the center of the image;

[0080] b12. Connect the midpoints of the left and right sides of the image, and determine that the intersection points of the connection line and the marked adjacent seat features are a1, a2, a3, and a4 from left to right. If the distance between a1 and a2 is equal to the distance between a3 and a4, the current image data is retained;

[0081] b13. Then perform the splicing operation of the images according to the order and direction of image acquisition. First, determine the number of seat features with equal distances between adjacent image data, and take the seat features with equal distances at the end of the first image data as the reference. Then, determine the seat features corresponding to the reference seat features in the second image data, and perform the alignment and covering operation at the corresponding seat features. The common part of the second image data covering the first image data is made transparent until the splicing of the retained images is completed.

[0082] During the splicing process, based on the common seat features determined by the first image data and the second image data, determine the boundary of a common seat feature. Then, use the boundary as the dividing line between the first image data and the second image data, and make the common part of the second image data covering the first image data transparent at the dividing line.

[0083] In the embodiment of the present invention, the operation of a22 matching the expression change of the voice data with the set data in the historical database is:

[0084] D1. Based on the characteristics of the person to be identified, and taking the characteristics of the person's mouth as the sound generation point, the indoor space is converted into a spatial coordinate system, the coordinates of the generation point are determined as (x0, y0, z0), and the coordinates of the position of each microphone device are determined as (x n ,y n 、z n );

[0085] D2. First, the distance between each microphone device and the generating point is measured, and the microphone device with the smallest distance is used for the recognition of the current person's language data;

[0086] And the distance calculation formula between the microphone device and the generating point is:

[0087]

[0088] H is the distance between the microphone device and the generating point, x n ,y n 、z n are respectively the vertical distances from the microphone device position to the coordinate axis of the space coordinate system;

[0089] D3, and determine and mark the position of the corresponding person based on the direction of the sound, the decibel level of the sound, and the distance between the microphone device and the sound generation point.

[0090] Among them, the indoor space is converted into a spatial coordinate system with the bottom surface of the room as the reference, and the intersection with the other two walls as the origin, and then the X-axis and Y-axis of the spatial coordinate system are formed from the origin and the intersection boundary of the wall and the bottom surface of the room, and the Z-axis of the spatial coordinate system is formed with the current wall intersection boundary.

[0091] In the embodiment of the present invention, the operation of matching the dynamic changes of the face with the historical database setting data in a22 is:

[0092] C1. Find the subsequent image data of the corresponding seat map based on the position of the corresponding person determined and marked, and perform feature extraction operation. Only image data with portrait features is retained in this extraction;

[0093] C2. Finding a positioning point on the image data based on the portrait feature, and determining the deviation distance between the cross cursor of the camera and the positioning point. If the deviation distance meets the set deviation threshold, the image data is retained;

[0094] C3. Extract the setting data about the dynamic changes of portrait features based on the historical database, perform pixel conversion on the image data retained in the C2 operation, convert it into the same pixels as the setting data, and determine the status result based on the situation after the retained image data matches the setting data.

[0095] By performing a feature extraction operation based on subsequent image data of the corresponding seat map, only the image data with portrait features is retained in this extraction. Locate points are searched for based on the image data with portrait features, and the deviation distance between the crosshair of the camera and the locate points is determined. If the deviation distance meets the set deviation threshold, the retention of the image data is achieved, and three-level processing of the image data is realized. First, retain the image with the features of the front side of the seat, then retain the image data with portrait features, and finally retain the image data where the camera crosshair and the portrait feature locate points meet the requirements. This not only reduces the processing of the amount of image data but also ensures that the data for judging the personnel status is more accurate.

[0096] In the embodiment of the present invention, the operation of searching for locate points in C2 is as follows:

[0097] c201. Form an upper line segment by connecting the symmetric outer corners of the eyes in the portrait features, and form a lower line segment by connecting the corners of the two sides of the mouth in the portrait features;

[0098] c202. Connect the midpoints of the upper line segment and the lower line segment to form a vertical line segment, and the midpoint of the vertical line segment is the locate point of the portrait features.

[0099] In the embodiment of the present invention, the operation of comparing the deviation distance with the deviation threshold in C2 is as follows:

[0100] c221. Establish a coordinate axis on the image with the locate point as the center point, establish the Y-axis in the direction of the vertical line segment, and establish the X-axis starting from the center point and perpendicular to the Y-axis;

[0101] c222. At this time, the coordinate of the locate point is (0, 0), and the coordinate of the crosshair of the camera on the X - Y coordinate axis is (x, y), and the corresponding deviation distance is calculated based on the coordinate values;

[0102] c223. Extract the deviation threshold marked as L from the historical data and realize the comparison between the deviation distance and the deviation threshold.

[0103] In the embodiment of the present invention, the formula for calculating the deviation distance in c222 is:

[0104]

[0105] S is the distance from the crosshair of the camera to the locate point, that is, the deviation distance, and x and y respectively correspond to the vertical distances from the Y-axis and the X-axis;

[0106] The comparison result between the deviation distance and the deviation threshold is:

[0107] If S ≤ L, the deviation distance meets the set deviation threshold, and the retention of the image data is achieved;

[0108] If S > L, the deviation distance exceeds the set deviation threshold, and the image data is removed. When the portrait features at a certain location cannot be collected, after adjusting the collection position of the camera, the cross cursor corresponds to the positioning point to be collected, and the collection and retention of the image data are realized.

[0109] In the embodiment of the present invention, the operation of determining the status result according to the matching situation between the retained image data and the set data in C3 is as follows:

[0110] c31. After matching the retained image data with the set data, determine the status of the mouth feature movement, the status of the head feature posture, and the status of the eye feature of the person in the image data;

[0111] c32. Based on the mouth feature in the continuous image data, compare the distance between the midpoint of the lower edge of the upper lip feature and the midpoint of the upper edge of the lower lip feature. If the distance changes continuously, the status is abnormal;

[0112] Based on the head feature posture in the continuous image data, compare the direction of the vertical line segment with the vertical central axis in the image data. If the deviation angle between the two changes continuously, the status is abnormal;

[0113] Based on the person's eye feature in the continuous image data, compare the distance between the midpoint of the lower edge of the eyelid feature and the midpoint of the upper edge of the lower eyelid feature. If the distance is continuously 0, the status is abnormal;

[0114] c33. Obtain the current status result of the person according to the specific situation after feature matching.

[0115] In the embodiment of the present invention, the operation of realizing the assignment and fusion of image data and voice data to determine the person's status in a22 is as follows:

[0116] Perform assignment and fusion on the facial feature status and voice expression status results corresponding to each time period, and the expression is:

[0117] F = w×(p / R)+v×(q / T);

[0118] F is the evaluation value of a single person after assignment and fusion based on image data and voice data, w is the weight ratio based on image data, v is the weight ratio based on voice data, p is the number of abnormal images of the facial feature during the image data collection time period, q is the number of times recognized during the voice data collection time period, R is the number of collections during the facial feature collection time period, and T is the number of collections during the voice data collection time period;

[0119] Set the over - standard threshold of the evaluation value as K. Therefore, when the evaluation value F < K, the current state of the person meets the requirements. When the evaluation value F ≥ K, the current state of the person does not meet the requirements and needs to be adjusted adaptively.

[0120] By assigning and fusing the facial feature states and voice expression state results corresponding to each time period, and obtaining the evaluation value of a single person after the assignment and fusion based on image data and voice data, it can effectively judge whether there are abnormal situations during the learning process of the person. At the same time, it improves the error tolerance rate, so that strategy adjustment will not be immediately generated during the discovery of similar situations, and it is also convenient to more accurately judge the state of the person.

[0121] At the same time, the content not described in detail in this specification belongs to the prior art well - known to those skilled in the art.

[0122] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non - exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0123] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A face biometric recognition method based on big data, characterized in that: Specifically, it includes the following steps: A1. Collect the facial image data of personnel through a camera, and collect the voice data of personnel using a microphone device. Then, transmit the collected data to the processing system through wireless communication technology. A2. Use the processing system to analyze and process the image data and voice data, and realize state recognition by combining them. The specific processing operations are as follows: a21. Realize the screening operation of the image data and voice data, and extract the required data for subsequent processing. a22. Perform recognition operations on the image data and voice data. Realize recognition operations on the voice data, match the expression changes of the voice data with the data set in the historical database to determine the sound source and identify the corresponding personnel, and match the dynamic changes of the face with the data set in the historical database. Then, realize the assignment and fusion of the image data and voice data to determine the personnel state. a23. Transmit the determined personnel state to the display terminal and display it in a combined manner of text and image. A3. Implement the adaptive strategy adjustment operation according to the result.

2. The face biometric recognition method based on big data according to claim 1, wherein: In the above a21, realize the screening operation of the image data and voice data. B1. Receive the image data, and realize the recognition of the characteristic part in the image data, mark the seat characteristics, retain the image with the characteristics of the front side of the seat, and splice the image data to form an overall seat map. B2. Then, trace back to the acquisition camera according to the data of the corresponding overall seat map, and retain the subsequent image data collected by the acquisition camera. B3. Receive the voice data, realize the denoising process of the voice data, and truncate the voice data to only retain the voice segments with sound.

3. The face biometric recognition method based on big data according to claim 2, characterized in that: The operation of splicing to form an overall seat map in the above B1 is as follows: b11. Extract the image containing the seat characteristics, and determine the size of the adjacent seat characteristics near the center of the image. b12. Connect the midpoints of the left and right sides of the image, and determine that the intersection points of the connection line and the marked adjacent seat characteristics are a1, a2, a3, and a4 from left to right. When the distance between a1 and a2 is equal to the distance between a3 and a4, the current image data is retained. b13. Then, realize the splicing operation of the images according to the order and direction of image acquisition. First, determine the number of seat characteristics with equal distances between adjacent image data, and use the seat characteristics with equal distances at the end of the first image data as the benchmark. Then, determine the seat characteristics corresponding to the benchmark seat characteristics in the second image data, and perform alignment and covering operations at the corresponding seat characteristics. The common part where the second image data covers the first image data is made transparent until the splicing of the retained images is completed.

4. A face biometric recognition method based on big data according to claim 1, characterized in that: The operation of matching the expression changes of the voice data with the data set in the historical database in the above a22 is as follows: D1. Based on the personal characteristics to be recognized, taking the mouth characteristics of the person as the sound generation point, and transforming the indoor space into a space coordinate system, determining the coordinate of the generation point as (x0, y0, z0), and determining the coordinates of the positions of each microphone device as (x n , y n , z n ); D2. First, measure the distances of each microphone device from the sound generation point, and the microphone device with the smallest distance is used for the recognition of the language data of the current personnel. And the distance calculation formula of the microphone device from the sound generation point is: H is the distance between the microphone device and the generation point, x n , y n , z n are respectively the vertical distances from the position of the microphone device to the coordinate axes of the spatial coordinate system; D3. Combine the direction source of the sound generation, the decibel level of the sound, and the distance of the microphone device from the sound generation point to determine and mark the position of the corresponding personnel.

5. A face biometric recognition method based on big data according to claim 1, characterized in that: The operation of matching the dynamic changes of the face with the set data in the historical database in a22 is as follows: C1. Based on determining and marking the position of the corresponding person, find the subsequent image data of the corresponding seat map for feature extraction operation. Only the image data with portrait features is retained in this extraction; C2. Search for positioning points based on the image data with portrait features, and determine the deviation distance between the crosshair of the camera and the positioning points. If the deviation distance meets the set deviation threshold, the retention of the image data is achieved; C3. Extract the set data regarding the dynamic changes of portrait features from the historical database, perform pixel conversion on the image data retained in the C2 operation to the same pixels as the set data, and determine the status result according to the matching situation between the retained image data and the set data.

6. The face biometric recognition method based on big data according to claim 5, wherein: The operation of searching for positioning points in C2 is as follows: c201. Form an upper line segment by connecting the symmetric eye corners in the portrait features, and form a lower line segment by connecting the two sides of the mouth corners in the portrait features; c202. Connect the midpoints of the upper line segment and the lower line segment to form a vertical line segment. The midpoint of the vertical line segment is the positioning point of the portrait features.

7. A face biometric recognition method based on big data according to claim 5, characterized in that: The comparison operation of the deviation distance and the deviation threshold in C2 is as follows: c221. Based on the image, establish a coordinate axis with the positioning point as the center point, establish the Y-axis in the direction of the vertical line segment, and establish the X-axis starting from the center point and perpendicular to the Y-axis; c222. At this time, the coordinates of the positioning point are (0, 0), and the coordinates of the crosshair of the camera on the X-Y coordinate axis are (x, y), and calculate the corresponding deviation distance based on the coordinate values; c223. Extract the deviation threshold in the historical data and label it as L, and perform the comparison between the deviation distance and the deviation threshold.

8. A face biometric recognition method based on big data according to claim 7, characterized in that: The formula for calculating the deviation distance in c222 is as follows: S is the distance from the crosshair of the camera to the positioning point, that is, the deviation distance, and x and y respectively correspond to the vertical distances from the Y-axis and the X-axis; The comparison result of the deviation distance and the deviation threshold is as follows: If S ≤ L, the deviation distance meets the set deviation threshold, and the retention of the image data is achieved; If S > L, the deviation distance exceeds the set deviation threshold, and the removal of the image data is achieved. When a certain portrait feature fails to be collected, after adjusting the collection position of the camera, the crosshair is made to correspond to the positioning point to be collected, and the collection and retention of the image data are achieved.

9. A face biometric recognition method based on big data according to claim 5, characterized in that: The operation of determining the status result according to the matching situation between the retained image data and the set data in C3 is as follows: c31. After matching the retained image data with the set data, determine the status of the mouth feature actions, the status of the head feature postures, and the status of the eye features of the person in the image data; c32. Based on the mouth features in the continuous image data, compare the distance between the midpoint of the lower edge of the upper lip feature and the midpoint of the upper edge of the lower lip feature. If the distance continuously changes, the status is abnormal; Based on the head feature postures in the continuous image data, compare the direction of the vertical line segment with the vertical central axis in the image data. If the deviation angle between the two continuously changes, the status is abnormal; Based on the eye features of a person in continuous image data, compare the distance between the midpoint of the lower edge of the eyelid feature and the midpoint of the upper edge of the lower eyelid feature. If the distance remains 0 continuously, the state is abnormal. c33. Obtain the current state result of the person according to the specific situation after feature matching.

10. A face biometric recognition method based on big data according to claim 1, characterized in that: The operation of realizing the assignment fusion of image data and voice data to determine the person's state in a22 is as follows: Perform assignment fusion on the facial feature state and voice expression state results corresponding to each time period, and the expression is: F = w×(p / R) + v×(q / T); F is the evaluation value after assignment fusion of a single person based on image data and voice data, w is the weight ratio based on image data, v is the weight ratio based on voice data, p is the number of abnormal images of the facial feature during the image data acquisition time period, q is the number of times recognized during the voice data acquisition time period, R is the number of acquisitions during the facial feature acquisition time period, and T is the number of acquisitions during the voice data acquisition time period. Set the excess threshold of the evaluation value as K. Therefore, when the evaluation value F < K, the current state of the person meets the requirements, and when the evaluation value F ≥ K, the current state of the person does not meet the requirements and needs to be adjusted adaptively.

Images