An identity recognition method and device, and a computer readable storage medium

Abstract

Embodiments of the present application disclose an identity recognition method and device, and a computer readable storage medium, wherein the method comprises: collecting a desk scene desk picture image in a classroom; identifying the desk scene image based on a desk desktop recognition model to determine at least one desk; determining a target student activity area based on the at least one desk; and determining the identity information of a target student corresponding to the target student activity area based on a mapping relationship between a preset student activity area and student identity information. In the field of image recognition, the above method only needs to determine the target student activity area, and then uses the mapping relationship between the student activity area and the student identity information to determine the identity information of the target student. In the case that the student face image cannot be collected and the face image of the student in the back row is blurred, the student identity can still be normally recognized, and the recognition accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of image recognition, and more particularly to an identity recognition method and apparatus, and a computer-readable storage medium. Background Technology

[0002] In modern smart classrooms, analyzing students' facial expressions, concentration, and classroom behavior using big data and AI technologies can effectively promote educational development. However, the most important and fundamental step in analyzing each individual's expressions and behavior is verifying their identity.

[0003] Currently, facial recognition is the most widely used method of identity authentication for humans, with advantages such as simple, convenient, and covert data collection.

[0004] However, facial recognition is greatly affected by the external environment. In classroom settings, students may be looking down at books and their faces cannot be captured, and the captured images of students in the back rows may be blurry, resulting in low recognition accuracy. Summary of the Invention

[0005] This invention provides an identity recognition method and apparatus, as well as a computer-readable storage medium, which can effectively recognize student identities in classroom settings and improve recognition accuracy.

[0006] The technical solution of this invention is implemented as follows:

[0007] This invention provides an identity recognition method, the method comprising:

[0008] Capture images of desks in a classroom;

[0009] Based on the desk desktop recognition model, the desk scene image is recognized to identify at least one desk;

[0010] Based on the at least one desk, determine the target student activity area;

[0011] Based on the preset mapping relationship between student activity areas and student identity information, the identity information of the target student corresponding to the target student activity area is determined.

[0012] In the above scheme, before identifying at least one desk based on the desk desktop recognition model in the desk scene image, the method further includes:

[0013] Obtain classroom desk scene samples;

[0014] An initial desk desktop recognition model is used to extract features from the desk scene samples to obtain sample feature images; and the sample feature images are then used for classification training to obtain training results.

[0015] The initial desk desktop recognition model is trained based on the training results. If the error between the training results and the true values ​​of the samples does not meet the preset threshold, the desk desktop recognition model is trained again based on the training results until the error between the training results and the true values ​​of the samples meets the preset threshold, thus obtaining the desk desktop recognition model.

[0016] In the above scheme, the step of using an initial desk desktop recognition model to extract features from the desk scene samples to obtain sample feature images includes:

[0017] Using the initial desk desktop recognition model, the desk scene samples are divided into N levels to obtain N-level desk scene sub-samples; where N is a positive integer greater than 1.

[0018] The i-th level desk scene sub-sample in the N-level desk scene sub-sample is sampled at K different resolutions to obtain the i-th level sample feature image, thereby obtaining N sample feature images; where i is greater than or equal to 1 and less than or equal to N; K is a positive integer greater than 1;

[0019] The N sample feature images are fused to obtain the sample feature image.

[0020] In the above scheme, the step of sampling the i-th level desk scene sub-sample from the N-th level desk scene sub-samples at K different resolutions to obtain the i-th level sample feature image includes:

[0021] The i-th level desk scene sub-sample in the N-level desk scene sub-sample is sampled at the j-th layer to obtain the j-th layer sampled feature image;

[0022] Obtain the other j-th layer sampled feature images corresponding to the other level desk scene sub-samples in the N-level desk scene sub-samples, excluding the i-th level desk scene sub-sample;

[0023] After fusing the j-th layer sampled feature image with the other j-th layer sampled feature images, sampling of the (j+1)-th layer continues until sampling of the K-th layer is reached, resulting in the k-th layer sampled feature image; the k-th layer sampled feature image is used as the i-th level sampled feature image.

[0024] In the above scheme, the step of identifying at least one desk based on the desk desktop recognition model to recognize the desk scene image includes:

[0025] Based on the desk desktop recognition model, the desk scene image is recognized to determine the location information of at least one desk, thereby identifying the at least one desk.

[0026] In the above scheme, determining the target student activity area based on the at least one desk includes:

[0027] Based on the location information of one of the at least one desks, the boundary is extended along the first direction to the adjacent desks to determine a student activity area corresponding to the one desk, thereby determining at least one student activity area;

[0028] The area containing objects of a preset type in the at least one student activity area is determined as the target student activity area.

[0029] In the above scheme, before determining the identity information of the target student corresponding to the target student activity area based on the preset mapping relationship between student activity areas and student identity information, the method further includes:

[0030] Capture panoramic images of desks in the classroom;

[0031] Based on the desk desktop recognition model, the panoramic image of the desk is recognized to determine the panoramic desk;

[0032] Based on the panoramic desks, the overall student activity area is determined;

[0033] Collect and identify student facial information for each student activity area within the global student activity area;

[0034] The student facial information within the student activity area is matched with the facial information in the identity feature database to determine the student identity information within each student activity area;

[0035] The student identity information within each student activity area is mapped to the corresponding student activity area to determine the preset mapping relationship between student activity areas and student identity information.

[0036] In the above scheme, the step of collecting and identifying student facial information in each student activity area of ​​the global student activity area includes:

[0037] A zoom acquisition device was used to acquire student face images for each student activity area within the global student activity area;

[0038] The student's facial image was subjected to facial recognition to obtain the recognition result;

[0039] If the recognition result indicates that the recognition has failed, the zoom acquisition device is refocused and the acquisition and recognition are continued until the student face information of each student activity area is obtained.

[0040] If the recognition result indicates successful recognition, then the student's facial information for each student activity area is obtained.

[0041] In the above scheme, after mapping the student identity information within each student activity area to the corresponding student activity area to determine the preset mapping relationship between student activity areas and student identity information, the method further includes:

[0042] Receive student modification requests;

[0043] In response to the student's modification request, the mapping relationship between the preset student activity area and student identity information is updated.

[0044] In the above scheme, after mapping the student identity information within each student activity area to the corresponding student activity area to determine the preset mapping relationship between student activity areas and student identity information, the method further includes:

[0045] Receive region partitioning requests;

[0046] In response to the area division request, the desk area is re-divided and identified, and the mapping relationship between the preset student activity area and student identity information is updated.

[0047] This invention provides an identity recognition device, which includes a data acquisition unit and a determination unit; wherein,

[0048] The acquisition unit is used to acquire images of desks in the classroom.

[0049] The determining unit is configured to identify at least one desk in the desk scene image based on a desk desktop recognition model; determine a target student activity area based on the at least one desk; and determine the identity information of the target student corresponding to the target student activity area based on a preset mapping relationship between student activity areas and student identity information. This invention provides an identity recognition device, which includes:

[0050] Memory, used to store executable data instructions;

[0051] A processor is configured to execute executable instructions stored in the memory, and when the executable instructions are executed, the processor executes the identity recognition method.

[0052] This invention provides a computer-readable storage medium storing executable instructions, which, when executed by one or more processors, enable the processors to perform the identity verification method.

[0053] This invention provides an identity recognition method and apparatus, and a computer-readable storage medium. The method involves acquiring images of desks in a classroom setting; identifying at least one desk based on a desk desktop recognition model; determining a target student activity area based on the at least one desk; and determining the identity information of the target student corresponding to the target student activity area based on a preset mapping relationship between the student activity area and student identity information. Using this method, it is only necessary to determine the target student activity area, and then use the mapping relationship between the student activity area and student identity information to determine the target student's identity. During this process, it is not necessary to acquire the student's facial information. Therefore, even when student facial images cannot be acquired or the facial images of students in the back rows are blurry, student identity can still be correctly identified, improving the accuracy of recognition. Attached Figure Description

[0054] Figure 1 An optional flowchart of the identity recognition method provided in an embodiment of the present invention. Figure 1 ;

[0055] Figure 2 A classroom desk scene diagram illustrating the identity recognition method provided in this embodiment of the invention;

[0056] Figure 3 This is a diagram illustrating the recognition effect of a desk scene image in the identity recognition method provided in an embodiment of the present invention.

[0057] Figure 4 An optional flowchart of the identity recognition method provided in an embodiment of the present invention. Figure 2 ;

[0058] Figure 5 An optional flowchart of the identity recognition method provided in an embodiment of the present invention. Figure 3 ;

[0059] Figure 6 An optional flowchart of the identity recognition method provided in an embodiment of the present invention. Figure 4 ;

[0060] Figure 7 An optional flowchart of the identity recognition method provided in an embodiment of the present invention. Figure 5 ;

[0061] Figure 8 An optional flowchart of the identity recognition method provided in an embodiment of the present invention. Figure 6 ;

[0062] Figure 9 An optional flowchart of the identity recognition method provided in an embodiment of the present invention. Figure 7 ;

[0063] Figure 10 An optional flowchart of the identity recognition method provided in an embodiment of the present invention. Figure 8 ;

[0064] Figure 11 This is a diagram illustrating the architecture of a desk desktop recognition model for the identity recognition method provided in this embodiment of the invention.

[0065] Figure 12 This is a diagram illustrating the effect of a desk desktop recognition method provided in an embodiment of the present invention.

[0066] Figure 13 A schematic diagram of the structure of the identity recognition device provided in the embodiments of the present invention. Figure 1 ;

[0067] Figure 14 A schematic diagram of the structure of the identity recognition device provided in the embodiments of the present invention. Figure 2 . Detailed Implementation

[0068] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0069] Figure 1 This is an optional flowchart of the identity recognition method provided in this embodiment of the invention. Figure 1 , will combine Figure 1 The steps shown are explained.

[0070] S101. Collect images of the classroom desks.

[0071] The identity recognition method provided in this embodiment of the invention is proposed for various school classroom scenarios, specifically for situations where students need to understand the learning status of each student and determine their identity information when they are in class or studying on their own.

[0072] It should be noted that the schools mentioned in the embodiments of this invention can be primary and secondary schools, universities, or various training institutions on the market, etc., and this invention does not impose any limitations. In the classroom scenes of various schools, desks, chairs, a podium, a blackboard, and other classroom equipment are arranged. In the classroom scene of this application embodiment, a data acquisition device is also set at the front or back of the classroom. This device can be a zoom camera or other devices with data acquisition functions, and this invention does not impose any limitations. The data acquisition device is connected and communicates with a terminal. Thus, in the classroom scene, the terminal can use the images of the desk scene containing classroom equipment captured by the data acquisition device to identify the students in their seats and understand their learning situation.

[0073] In a school classroom setting, desks are placed for students to study. Since students' seating is fixed, the terminal can indirectly determine the student's activity area based on the fact that each student has their own unique desk.

[0074] In this embodiment of the invention, the terminal can capture images of a classroom scene of desks by activating the acquisition function of the acquisition device. The desk scene image includes at least one desk and at least one student, and the terminal can identify the target student from among at least one student.

[0075] For example, Figure 2 Images of classroom desks were collected from... Figure 2 It can be seen that there are many desks, arranged in 8 columns and 6 rows. Figure 2 There are also students A, B, C, D, E, and F. Student A is in the 3rd column, 2nd row; student B is in the 4th column, 1st row; student C is in the 4th column, 3rd row; student D is in the 1st column, 5th row; student E is in the 6th column, 3rd row; and student F is in the 5th column, 5th row. There is a bookshelf 1 at the back left of the classroom, next to which is cabinet 1; there is a bookshelf 2 at the back right of the classroom, next to which is cabinet 2. Both bookshelves contain extracurricular reading materials for students to read during their free time. The two cabinets contain some prizes and class honor certificates. It can be seen that students A, E, and F are lying on their desks, while students B, C, and D are reading.

[0076] S102. Based on the desk desktop recognition model, identify the desk scene image and determine at least one desk.

[0077] The desk desktop recognition model in the identity recognition method provided in this embodiment of the invention is based on a neural network framework and mainly consists of two parts: a CNN network and a DNN network. The CNN network contains convolutional layers used for feature extraction; the DNN network contains fully connected layers used for classification and recognition.

[0078] In this embodiment of the invention, the terminal can identify the desk scene image through the desk desktop recognition model, identify the location information of at least one desk, and obtain a desk based on the location information of one desk. Therefore, at least one desk can be obtained based on the location information of at least one desk.

[0079] For example, Figure 3 To identify desks in a scene based on a desk desktop recognition model, and to determine the recognition result image of at least one desk. From Figure 3As can be seen, the rectangular frame in desk 1 is the identified desk, as are desks 2, 3, 4, 5, and 6. The other black rectangles are also identified desks. It can also be seen that student A is located at desk 1, student B at desk 2, student C at desk 3, student D at desk 4, student E at desk 5, and student F at desk 6. Figure 3 Taking the rectangular frame of desk 1 as an example, the coordinates of the four vertices labeled A, B, C, and D are the coordinates of the desk surface.

[0080] S103. Determine the target student activity area based on at least one desk.

[0081] In the identity recognition method provided in this embodiment of the invention, students are either listening to lectures in their seats or reading and doing homework at their desks. The area formed by the desks and chairs constitutes the student activity area. Therefore, the student activity area can be determined based on the desks. The target student activity area is the student activity area where the student to be identified is located.

[0082] In some embodiments of the present invention, step S103 can be implemented by steps S1031 and S1032, as follows:

[0083] S1031. Based on the location information of one of the desks in at least one desk, extend the boundary along the first direction to the adjacent desks to determine a student activity area corresponding to one desk, thereby determining at least one student activity area.

[0084] In this embodiment of the invention, the terminal can use the location information of one of the desks to extend the desk surface close to the student's boundary line along a first direction until it reaches the boundary line of the adjacent desk. The desk surface and the extended area together constitute a student activity area, thereby determining at least one student activity area.

[0085] For example, in a classroom setting, the desks in the classroom have rectangular tabletops, but this embodiment of the invention is not limited to this. The rectangular tabletop is divided into four boundaries: top, bottom, left, and right. The boundary closest to the student is the bottom boundary. The boundary extends along a first direction until it reaches the tabletop of the adjacent desk. A tabletop and its extended area together constitute a student activity area, thereby obtaining all student activity areas in the classroom.

[0086] It should be noted that the first direction is the opposite direction from when the student is facing the blackboard, such as... Figure 3 The direction indicated by arrow 1 in the image.

[0087] S1032. The area containing at least one student activity area of ​​a preset type is determined as the target student activity area. In this embodiment of the invention, the terminal can discover the area containing the preset type of object through at least one student activity area; this area is the target student activity area.

[0088] For example, in a classroom setting, the area containing objects of a predefined type is identified among all student activity areas; this area is the target student activity area.

[0089] It should be noted that, in this embodiment of the invention, the object of the preset type is the target student.

[0090] Understandably, the terminal identifies at least one student activity area through at least one desk, and then identifies the target student activity area within that at least one activity area, thus improving the efficiency of identifying the target student activity area.

[0091] S104. Based on the preset mapping relationship between student activity areas and student identity information, determine the identity information of the target student corresponding to the target student activity area.

[0092] In the identity recognition method provided in this embodiment of the invention, the mapping relationship between preset student activity areas and student identity information is pre-set. The terminal determines all student activity areas based on the positions of all desks, then determines the student identity in each student activity area, and finally maps and saves the identity information of all students corresponding to all student activity areas, forming the preset mapping relationship between student activity areas and student identity information.

[0093] In this embodiment of the invention, the terminal can find the student activity area number corresponding to the target student activity area through a pre-set mapping relationship between student activity areas and student identity information, and then find the corresponding target student identity information based on the student activity area number, thereby determining the identity information of the target student corresponding to the target student activity area.

[0094] For example, the target student activity area corresponds to area 1 in the student activity area. In the preset mapping relationship between student activity areas and student identity information, the student identity corresponding to area 1 in the student activity area is Zhang San. Then, the identity information of the student in the target student activity area is Zhang San.

[0095] It should be noted that, based on the pre-defined mapping relationship between student activity areas and student identity information, the identity information of the target student corresponding to the target student activity area can be determined quickly and accurately.

[0096] Understandably, the terminal determines the identity of the target student by mapping the target student's activity area and the student's identity information. It does not need to obtain the student's facial information. Even when it is impossible to collect the student's face image or the face image of the student in the back row is blurry, it can still identify the student's identity normally, thus improving the accuracy of recognition.

[0097] In some embodiments of the present invention, see Figure 4 , Figure 4 This is an optional flowchart of the identity recognition method provided in this embodiment of the invention. Figure 2 ,based on Figure 1 The process includes steps S105-S107 before step S102, as follows:

[0098] S105. Obtain a sample of a classroom desk scene.

[0099] In the identity recognition method provided in this embodiment of the invention, the desk scene samples refer to desk scene images collected from different classrooms. These different classrooms can be different classes within the same school, or classes from multiple schools; this embodiment of the invention does not impose any limitations. Therefore, the desk scene samples are diverse, facilitating subsequent model training.

[0100] In this embodiment of the invention, the terminal can collect sample images of desk scenes in different classrooms by activating the collection function of the collection device, and the collected sample images of desk scenes constitute a desk scene sample.

[0101] For example, the shape of the desk desktop in the desk scene sample can be rectangular, square, circular, trapezoidal, etc., and the embodiments of the present invention are not limited thereto.

[0102] S106. Using the initial desk desktop recognition model, feature extraction is performed on the desk scene samples to obtain sample feature images; and the sample feature images are then used for classification training to obtain training results.

[0103] The initial desk desktop recognition model in the identity recognition method provided in this embodiment of the invention is based on a neural network framework, mainly composed of CNN convolutional layers and DNN fully connected layers, and is a newly built model. The sample feature image refers to the image obtained by sampling the desk scene samples. The training result refers to the classification result obtained by classifying the sample feature images, specifically including the position and coordinates of the desk in the desk scene samples.

[0104] In this embodiment of the invention, the terminal can extract features from a desk scene sample using an initial desk desktop recognition model to obtain a sample feature image, and then continue to classify and recognize the sample feature image to obtain the training result.

[0105] In some embodiments of the present invention, see Figure 5 , Figure 5 This is an optional flowchart of the identity recognition method provided in this embodiment of the invention. Figure 3 ,based on Figure 4 Step S106 can be achieved through steps S1061 to S1063, combining... Figure 5 The steps shown are explained.

[0106] S1061. Using the initial desk desktop recognition model, the desk scene samples are divided into N levels to obtain N-level desk scene sub-samples; where N is a positive integer greater than 1.

[0107] In this embodiment of the invention, the terminal can use an initial desk desktop recognition model to divide the collected desk scene samples into N levels, resulting in N-level desk scene sub-samples. Specifically, the division is based on the different resolutions of the desk scene samples to perform N-level divisions on the input desk scene samples, thus obtaining N-level desk scene sub-samples.

[0108] For example, the desk scene samples are divided into N levels, where N is set to 3. Based on the initial desk desktop recognition model, the input desk scene samples are divided into 3 levels according to their different resolutions, resulting in 3 levels of desk scene sub-samples: first-resolution desk scene sub-samples, second-resolution desk scene sub-samples, and third-resolution desk scene sub-samples.

[0109] It should be noted that, based on the initial desk desktop recognition model, the input desk scene samples are divided into N levels according to the different resolutions of the desk scene samples, resulting in N levels of desk scene sub-samples. This can effectively solve the problem of having a good recognition accuracy for the collected desk scene samples under different resolution conditions.

[0110] S1062. Sample the i-th level desk scene sub-sample in the N-level desk scene sub-samples at different resolutions through K layers to obtain the i-th level sample feature image, thereby obtaining N sample feature images; where i is greater than or equal to 1 and less than or equal to N; K is a positive integer greater than 1.

[0111] In this embodiment of the invention, the terminal can sample the i-th level desk scene sub-sample in the N-level desk scene sub-sample at different resolutions in K layers to obtain the i-th level sample feature image, thereby obtaining N sample feature images.

[0112] In some embodiments of the present invention, step S1062 can be implemented by S10621, S10622 and S10623, as follows:

[0113] S10621. Sample the i-th level desk scene sub-sample in the N-level desk scene sub-samples at the j-th layer to obtain the j-th layer sampled feature image.

[0114] In this embodiment of the invention, the terminal can sample the i-th level desk scene sub-sample from the N-level desk scene sub-sample to obtain the j-th level sampled feature image.

[0115] For example, N is set to 3, i to 1, j to 2, and k to 10. The first-level desk scene sub-sample in the 3-level desk scene sub-sample is sampled in the second layer to obtain the second-layer sampled feature image.

[0116] S10622. Take the desk scene sub-samples of the N-level classroom scene (excluding the i-th level desk scene sub-sample) and sample the feature images of the other level desk scene sub-samples at the j-th layer respectively to obtain other j-th layer sampled feature images.

[0117] In this embodiment of the invention, the terminal can sample the other level desk scene sub-samples (excluding the i-th level desk scene sub-sample) from the N-level desk scene sub-samples, and obtain other j-th level sampled feature images.

[0118] For example, N is set to 3, i to 1, j to 2, and k to 10. The second-level desk scene sub-sample in the third-level desk scene sub-sample is sampled to the second layer to obtain the second-level second-layer sampled feature image; the third-level desk scene sub-sample in the third-level desk scene sub-sample is sampled to the second layer to obtain the second-level third-layer sampled feature image.

[0119] S10623. After fusing the j-th layer sampled feature image with other j-th layer sampled feature images, continue sampling the (j+1)-th layer until sampling the K-th layer, to obtain the k-th layer sampled feature image; use the k-th layer sampled feature image as the i-th level sampled feature image.

[0120] In this embodiment of the invention, the terminal can fuse the feature image of the j-th layer with other j-th layer sampled feature images, and continue to sample the fused feature image of the (j+1)-th layer until the sampling of the K-th layer is reached, to obtain the k-th layer sampled feature image; the k-th layer sampled feature image is used as the i-th level sampled feature image.

[0121] For example, N is set to 3, i to 1, j to 2, and k to 10. The second-level sampled feature image of level 1, the second-level sampled feature image of level 2, and the second-level sampled feature image of level 3 are fused together, and the sampling of the third level of level 1 continues until the sampling of the tenth level of level 1 is reached, to obtain the tenth-level sampled feature image of level 1; the tenth-level sampled feature image of level 1 is used as the first-level sample feature image.

[0122] Understandably, sampling the i-th level desk scene sub-sample from the N-level desk scene sub-samples at different resolutions using K layers yields the i-th level sample feature image, which improves the accuracy of desk scene image recognition.

[0123] S1063. Fuse the N sample feature images to obtain the sample feature image.

[0124] In this embodiment of the invention, the terminal can fuse N sample feature images to obtain a sample feature image.

[0125] It should be noted that the feature images of the multiple sub-samples are fused to obtain the feature image of the final desk scene sample, which facilitates the subsequent training and recognition of the feature image.

[0126] S107. Train the initial desk desktop recognition model based on the training results. If the error between the training results and the true values ​​of the samples does not meet the preset threshold, continue to train the desk desktop recognition model based on the training results until the error between the training results and the true values ​​of the samples meets the preset threshold, and obtain the desk desktop recognition model.

[0127] In the identity recognition method provided in this embodiment of the invention, the true value of the sample refers to the manually labeled desktop markers and their coordinates on the collected images of the desk scene. The preset threshold is also a pre-set value.

[0128] In this embodiment of the invention, the terminal can continue training based on the training results obtained from the initial desktop recognition model until the error between the obtained training result and the true sample value is less than a preset threshold. At this point, training ends, and a desktop recognition model is obtained. Specific operations include: comparing the obtained training result with the true sample value; if the error is greater than the preset threshold, automatically modifying the network parameters and continuing training; if the error between the training result and the true sample value is less than the preset threshold, stopping training, saving the current network model, which is the obtained desktop recognition model.

[0129] Understandably, the initial desk desktop recognition model is trained to obtain the desk desktop recognition model. The terminal can use the desk desktop recognition model to identify the position and coordinates of the desk in the classroom scene image, thereby improving the accuracy of desk recognition.

[0130] In other embodiments of the present invention, see Figure 6 , Figure 6 This is an optional flowchart of the identity recognition method provided in this embodiment of the invention. Figure 4 ,based on Figure 1 Before step S104, steps S108-S1013 are also included, as follows:

[0131] S108. Collect panoramic images of the desks in the classroom.

[0132] In the identity recognition method provided in this embodiment of the invention, the panoramic image of the desks refers to the image that does not contain students and teachers, but only all the desks in the classroom.

[0133] In this embodiment of the invention, the terminal can acquire panoramic images of the desks in the classroom by activating the acquisition function of the acquisition device. The acquisition device acquires images of the desk positions in the unattended classroom by adjusting the angle, thus obtaining a panoramic image of the desks in the classroom.

[0134] S109. Based on the desk desktop recognition model, identify the panoramic image of the desk to determine the panoramic desk.

[0135] In the identity recognition method provided in this embodiment of the invention, the panoramic desk is an image in which all desks in a panoramic image of desks are identified and labeled.

[0136] In this embodiment of the invention, the terminal can identify the panoramic image of a desk using a desk desktop recognition model to obtain a panoramic view of the desks. The panoramic image of the desks is input into the desk desktop recognition model for identification to obtain the positions of all desks in the image. Based on the positions of all desks in the image, the panoramic view of the desks is obtained.

[0137] S1010: Based on panoramic desks, determine the overall student activity area.

[0138] In the identity recognition method provided in this embodiment of the invention, the global student activity area refers to the activity area of ​​all students.

[0139] In this embodiment of the invention, the terminal can determine the activity areas of all students, i.e., the global student activity area, through the panoramic desk. In a classroom scenario, based on the position information of one desk in the panoramic desk, the boundary is extended along a first direction to adjacent desks to obtain a student activity area corresponding to one desk, and thus all student activity areas are obtained, i.e., the global activity area.

[0140] S1011. Collect and identify the facial information of students in each student activity area of ​​the global student activity area.

[0141] In this embodiment of the invention, the terminal can collect the face information of students in each student activity area of ​​the global student activity area by enabling the collection function of the collection device, and then recognize the collected face information.

[0142] In some embodiments of the present invention, see Figure 7 , Figure 7 This is an optional flowchart of the identity recognition method provided in this embodiment of the invention. Figure 5 ,based on Figure 6 Step S1011 can be achieved through steps S10111 to S10114, combining... Figure 7 The steps shown are explained.

[0143] S10111. Use a zoom acquisition device to acquire student face images of each student activity area in the global student activity area.

[0144] In this embodiment of the invention, the terminal can activate the zoom acquisition device to capture the facial information of students in each student activity area within the global student activity area, thereby obtaining facial images of students in each student activity area within the global student activity area. The zoom acquisition device sequentially aligns with each student activity area within the global student activity area, capturing the facial image of students in each student activity area.

[0145] S10112. Perform face recognition on the student's face image and obtain the recognition result.

[0146] In this embodiment of the invention, the terminal can perform face recognition on the collected student face images to obtain the face recognition result. Based on the face recognition model, the collected student face images are recognized to obtain the recognition result. The recognition result includes two categories: one is that the collected student face image is successfully recognized, meaning the collected student face image is clear; the other is that the collected student face image cannot be recognized, meaning the collected student face image is unclear.

[0147] S10113. If the recognition result indicates that the recognition has failed, the zoom acquisition device is refocused and the acquisition and recognition are continued until the student face information of each student activity area is obtained.

[0148] In this embodiment of the invention, if the recognition result is unclear, indicating recognition failure, the zoom acquisition device is refocused and the student face information of each student activity area is acquired and recognized until the student face information of each student activity area is obtained. If the recognition result of the student face image acquired in a student activity area fails, the zoom acquisition device is refocused, and the student face image in the student activity area is re-acquired and recognized until the recognition result is successful, thus obtaining the student face information. The above operation is repeated for each student activity area until the student face information of each student activity area is obtained.

[0149] It should be noted that if the recognition result indicates recognition failure, the zoom acquisition device is refocused and the student facial information of each student activity area is collected and then recognized until the student facial information of each student activity area is obtained. This process realizes the acquisition of student facial information.

[0150] S10114. If the recognition result indicates successful recognition, then the student facial information for each student activity area is obtained.

[0151] In this embodiment of the invention, if the recognition result is clear, it indicates successful recognition, and the collected student face image is the student face information, thus obtaining the student face information for each student activity area. When the recognition result of the student face image collected in the student activity area is successful, the collected student face image is the obtained student face information, and the corresponding student face information for each student activity area is obtained.

[0152] Understandably, collecting student facial information from each student activity area in the global student activity area and recognizing the collected facial information can obtain student facial information, thereby improving the accuracy and efficiency of identity recognition.

[0153] S1012. Match the student facial information within the student activity area with the facial information in the identity feature database to determine the student identity information within each student activity area.

[0154] In this embodiment of the invention, the terminal can compare the facial information of students within a student activity area with facial information in an identity feature database to determine the identity information of each student within that area. Based on the facial information in the identity feature database, the terminal finds facial information in the database that matches the facial information of a student within the student activity area, and then determines the corresponding identity information based on the found facial information, thereby determining the identity information of each student within the student activity area.

[0155] For example, when students enroll, their facial and identity information is collected, and each student's facial information is matched one-to-one with their identity information to establish an identity feature database. The facial information of students within a given activity area is then matched with the facial information in the identity feature database to determine the identity information of each student within that area. For instance, if the facial information of a student within a certain activity area is matched to "Li Si" in the identity feature database, then the student in that activity area is identified as "Li Si".

[0156] It should be noted that matching the facial information of students within the student activity area with the facial information in the identity feature database to determine the student identity information within each student activity area lays the groundwork for establishing the mapping relationship between the preset student activity areas and student identity information.

[0157] S1013. Map the student identity information in each student activity area to the corresponding student activity area to determine the preset mapping relationship between student activity areas and student identity information.

[0158] In this embodiment of the invention, the terminal can obtain a preset mapping relationship between student activity areas and student identity information by mapping the student identity information in each student activity area to the corresponding student activity area. Based on the obtained student identity information in each student activity area and the corresponding student activity area, the two are matched one-to-one, with one student activity area corresponding to one student identity information. This matching process is repeated for all student activity areas and student identity information to obtain the preset mapping relationship between student activity areas and student identity information.

[0159] For example, student activity areas are numbered, such as: Area 1, Area 2, Area 3, etc. The student's identity information in Area 1 is Zhang San, the student's identity information in Area 2 is Li Si, the student's identity information in Area 3 is Wang Wu, etc. The established preset mapping relationship between student activity areas and student identity information is: Area 1 is Zhang San, Area 2 is Li Si, Area 3 is Wang Wu, etc. This embodiment of the invention is not limited.

[0160] Understandably, establishing a mapping relationship between preset student activity areas and student identity information allows for quick retrieval of the identity information of the person to be identified within the mapping relationship during identity recognition, thereby improving the accuracy of identity recognition.

[0161] In some embodiments of the present invention, see Figure 8 , Figure 8 This is an optional flowchart of the identity recognition method provided in this embodiment of the invention. Figure 6 ,based on Figure 6 Following step S1013, steps S1014-S1015 are also included, as follows:

[0162] S1014. Receive student modification requests.

[0163] In this embodiment of the invention, when a student's seating information changes, the terminal receives a request from the student to modify the information.

[0164] For example, if the teacher swaps the positions of students Zhang San and Li Si, a student modification request will be sent to the terminal.

[0165] S1015. In response to a student's modification request, update the mapping relationship between the preset student activity area and student identity information.

[0166] In this embodiment of the invention, the terminal can receive a student modification request, check the request information, and update the preset mapping relationship between student activity areas and student identity information. The terminal receives the student modification request when student seating information changes. One method is for the terminal to modify the mapping relationship between student activity areas and student identity information accordingly based on the student seating change information, thereby updating the preset mapping relationship. Another method is for the terminal to re-collect and identify student facial information in each student activity area within the global student activity area, thereby determining the student identity information in each student activity area and updating the preset mapping relationship between student activity areas and student identity information.

[0167] For example, when a teacher swaps the seats of students Zhang San and Li Si, a student modification request is sent to the terminal. Originally, in the preset mapping relationship between student activity areas and student identity information, area 1 was Zhang San and area 2 was Li Si. Since Zhang San and Li Si have switched seats, the preset mapping relationship is modified so that area 1 is Li Si and area 2 is Zhang San, while the rest of the mapping remains unchanged, thus updating the preset mapping relationship between student activity areas and student identity information.

[0168] For example, when a teacher changes the seats of a large number of students, a student modification request is sent to the terminal. The teacher then re-collects and identifies the student facial information in each student activity area of ​​the global student activity area, thereby determining the student identity information in the student activity area and updating the preset mapping relationship between student activity areas and student identity information.

[0169] Understandably, the terminal responds to student modification requests by updating the mapping relationship between the preset student activity area and student identity information, which can adapt to changing environments and improve the applicability of the identification method.

[0170] In some embodiments of the present invention, see Figure 9 , Figure 9This is an optional flowchart of the identity recognition method provided in this embodiment of the invention. Figure 7 ,based on Figure 6 Following step S1013, steps S1016-S1017 are also included, as follows:

[0171] S1016, Receive area division request.

[0172] In this embodiment of the invention, when the arrangement of desks in the classroom changes, the terminal receives a region division request.

[0173] For example, the desks in the classroom are changed from three groups of four rows to four groups of three rows, and a region division request is sent to the terminal.

[0174] S1017. In response to the area division request, the desk area is re-divided and identified to update the mapping relationship between the preset student activity area and student identity information.

[0175] In this embodiment of the invention, the terminal can receive a region division request, re-identify and divide the desk areas, and update the mapping relationship between the preset student activity areas and student identity information. When the position of the desks in the classroom changes, the terminal will receive a region division request and then re-execute S108-S1013 to update the mapping relationship between the preset student activity areas and student identity information.

[0176] For example, when the desks in the classroom change from three groups of four rows to four groups of three rows, a region division request is sent to the terminal, and then S108-S1013 are re-executed to update the mapping relationship between the preset student activity area and student identity information.

[0177] Understandably, in response to the area division request, the terminal re-divides and identifies the desk areas to update the mapping relationship between the preset student activity areas and student identity information. This can adapt to changing environments and improve the applicability of the identification method.

[0178] The following will describe an exemplary application of the embodiments of this application in a real-world application scenario.

[0179] An optional flowchart of an exemplary identity recognition method provided in this application embodiment. Figure 8 ,like Figure 10 As shown.

[0180] For example, in a typical classroom setting, the terminal can identify a student in a seat by collecting images of the classroom desks from the acquisition device, thus determining the student's identity information.

[0181] The process is as follows: 1. Acquire images of desks in the classroom. 2. Based on a desk desktop recognition model, identify the desk scene images to obtain the positions and coordinates of all desks. 3. Based on the positions of all desks, extend the area to obtain all student activity areas. 4. Locate the target student activity area among all student activity areas. 5. Based on the preset mapping relationship between student activity areas and student identity information, determine the identity information of the target student corresponding to the target student activity area.

[0182] Based on the above brief description, this application provides an identity recognition method. In this application, the identity recognition method may include the following steps:

[0183] 1. Collect images of classroom desks.

[0184] The terminal can capture images of classroom desks, including students, by enabling the acquisition function of the acquisition device.

[0185] 2. Based on the desk desktop recognition model, the desk scene image is recognized to obtain the position and coordinates of all desks.

[0186] The architecture diagram of the desk desktop recognition model is as follows: Figure 11 As shown, the process includes: input images of a classroom desk scene, a CNN network, a DNN network, and a training target. The desk desktop recognition model is used to identify the desk scene images and obtain the positions and coordinates of all desks. This includes: collecting and labeling classroom desk scene images to create a dataset; inputting the dataset into the desk desktop recognition model, using a CNN network to extract features and obtain feature images; and using a DNN network to train the DNN network for classification and recognition to obtain the training target.

[0187] For example, the desk scene sample is a desk scene image, which is divided into N levels, where N is set to 3. The CNN network internal architecture has three resolution subnetworks (high, medium, and low) in parallel to extract features. The high, medium, and low resolution layers are respectively... Figure 9 H(k), M(k), and L(k) are used in the model. The image X representing the desk scene is divided into three sub-samples, X1, X2, and X3, which are input into the CNN network to represent high-resolution, medium-resolution, and low-resolution images (equivalent to the first, second, and third resolution images, respectively). Sampling is performed on each resolution image, and the output is a sampling result Y. Each layer produces three sampling results Y1, Y2, and Y3, where the resolution and width are the same as the input. Each output is a set of input mappings. Y k This represents the feature map after fusing the results of each sampling. The function a(X) i (k) From resolution i to k for xi It consists of upsampling or downsampling, where k = 64. Downsampling is performed using a 3×3 convolution. For example, downsampling with a stride of 2×2 is performed on a 3×3 convolution. For upsampling, the simplest nearest neighbor sampling is used, calibrating the number of channels from a 1×1 convolution. If i = k, then a(X i ,k) is just a recognition connection: a(X i ,k)=X i The system outputs the feature image of the desk scene image X. The terminal can perform classification training on the feature image of the desk scene image X to obtain the desktop surfaces and coordinates of all desks in the scene image. This effectively solves the problem of achieving good recognition accuracy even with pixel images of varying resolutions.

[0188] For example, the training objective is to obtain the desktop surfaces and coordinates of all desks in a classroom scene image. Figure 12 As shown in the figure, the black box represents the identified desk desktop, and points A, B, C, and D are the coordinates of the desktop.

[0189] 3. Based on the location and coordinates of all desks, obtain all student activity areas through area extension.

[0190] For example, in a real-world scenario, students are seated in chairs and need to extend their desks, for instance... Figure 12 The diagram shows four points A, B, C, and D on a desk's surface. Extending line AB upwards (along the first direction) until it touches line CD on another desk's surface. The desk surfaces and the extended area together form a student activity area. This method is used to obtain all student activity areas for all desks. The width of the maximum desk area is Lmax, the pixel distance from line AC in the diagram. Setting the extension length to not exceed Lmax solves the problem of potentially infinite extension for the last row of seats.

[0191] 4. Locate the target student activity area among all student activity areas.

[0192] The terminal can determine the student activity area number of the target student by the student's location.

[0193] For example, student A corresponds to region 1, such as Figure 12 As shown, the area of ​​region 1 includes the region jointly formed by rectangle ABCD and rectangle ABC1D2.

[0194] 5. Based on the pre-defined mapping relationship between student activity areas and student identity information, determine the identity information of the target student corresponding to the target student activity area.

[0195] The mapping relationship between pre-defined student activity areas and student identity information is established as follows:

[0196] 5.1 Collect panoramic images of the desks in the classroom.

[0197] 5.2 Based on the desk desktop recognition model, the panoramic image of the desks is identified to determine all desks;

[0198] For example, the camera captures panoramic images of the desk positions in an unmanned classroom, without changing the desk positions. The panoramic view of the desk positions within the classroom is used to divide the classroom into areas, labeled 1, 2, 3, ..., M (where M is the total number of desks). Alternatively, a desk recognition model can be used to identify the desk positions in the classroom, dividing the activity range of each desk area into areas labeled 1, 2, 3, ..., M (where M is the total number of desks). The resulting area division by the model is shown below. Figure 12 As shown.

[0199] 5.3 Determine the overall student activity area based on panoramic desks.

[0200] 5.4 Collect and identify the facial information of students in each student activity area of ​​the global student activity area, match it with the facial information in the identity feature database, determine the identity information of students in all student activity areas, and record the identity information of the students corresponding to the student activity areas.

[0201] For example, in a classroom setting, the camera zooms in to individually recognize the faces of students in M ​​student activity areas until a student's identity is identified. Then, it switches to the next student activity area and records the identity information of the student corresponding to that area. For example, area 1 corresponds to Zhang San.

[0202] 5.5 After obtaining the student identity information corresponding to the student activity area, the terminal will generate a mapping table, as shown in Table 1.

[0203] Table 1. Mapping Relationship Between Student Activity Areas and Student Identity Information

[0204] Student activity area Student status Area 1 Zhang San Area 2 Li Si …… ……

[0205] 5.6 If a student's location changes, the system can either send a modification request directly to the terminal to update the information or automatically re-match the data to form a new mapping table. Alternatively, if a desk's location changes, the system can send a region division request to the terminal to re-divide the regions and form a new mapping table.

[0206] For example, student A corresponds to region 1. According to the mapping relationship table between student activity areas and student identity information, region 1 is Zhang San, thus determining that student A's identity is Zhang San.

[0207] Understandably, in long-term, multi-person scenarios with fixed desks and dynamic student positions in a classroom, identification methods based on clothing or other items, as well as knowledge-based methods, are not well-suited for this environment. Students typically wear uniforms, making high-accuracy identification difficult. This method addresses this issue. Furthermore, it accurately identifies students even when no face is detected or the face is blurry, while significantly reducing the computation time for face recognition in multi-person scenarios.

[0208] Based on the positioning method of the above embodiments, this application also provides an identity recognition device, such as... Figure 13 As shown, Figure 13 A schematic diagram of the structure of an identity recognition device provided in this application embodiment. Figure 1 ,like Figure 13 As shown, the identity recognition device 13 includes: a data acquisition unit 1301 and a determination unit 1302; wherein,

[0209] The acquisition unit 1301 is used to acquire images of desks in the classroom.

[0210] The determining unit 1302 is used to identify the desk scene image based on the desk desktop recognition model to determine at least one desk; based on the at least one desk, determine the target student activity area; and based on the preset mapping relationship between the student activity area and student identity information, determine the identity information of the target student corresponding to the target student activity area.

[0211] In some embodiments of this application, the identity recognition device 13 further includes an acquisition unit 1303; wherein,

[0212] The acquisition unit 1303 is used to acquire a desk scene sample;

[0213] The acquisition unit 1303 is further configured to use an initial desk desktop recognition model to extract features from the desk scene sample to obtain a sample feature image;

[0214] The acquisition unit 1303 is also used to perform classification training on the sample feature image to obtain training results;

[0215] The acquisition unit 1303 is further configured to train the initial desk desktop recognition model based on the training results. If the error between the training results and the true values ​​of the samples does not meet a preset threshold, the desk desktop recognition model is further trained using the training results until the error between the training results and the true values ​​of the samples meets the preset threshold, thereby obtaining the desk desktop recognition model.

[0216] In an embodiment of this application, the acquisition unit 1303 is further configured to use the initial desk desktop recognition model to divide the desk scene sample into N levels to obtain N-level desk scene sub-samples; wherein, N is a positive integer greater than 1;

[0217] The acquisition unit 1303 is further configured to sample the i-th level desk scene sub-sample in the N-level desk scene sub-samples at different resolutions by K layers to obtain the i-th level sample feature image, thereby obtaining N sample feature images; where i is greater than or equal to 1 and less than or equal to N; K is a positive integer greater than 1;

[0218] The acquisition unit 1303 is further configured to fuse the N sample feature images to obtain the sample feature image.

[0219] In some embodiments of this application, the acquisition unit 1303 is further configured to sample the i-th level desk scene sub-sample in the N-th level desk scene sub-sample at the j-th layer to obtain the j-th layer sampled feature image;

[0220] The acquisition unit 1303 is also used to acquire other j-th layer sampling feature images corresponding to other level desk scene sub-samples other than the i-th level desk scene sub-sample in the N-level desk scene sub-sample;

[0221] The acquisition unit 1303 is further configured to fuse the j-th layer sampled feature image with the other j-th layer sampled feature images, and then continue to sample the (j+1)-th layer until the sampling of the K-th layer is reached, so as to obtain the k-th layer sampled feature image.

[0222] The acquisition unit 1303 is further configured to use the k-th layer sampled feature image as the i-th level sample feature map.

[0223] In some embodiments of this application, the determining unit 1302 is further configured to identify the desk scene image based on the desk desktop recognition model, determine the position information of the at least one desk, and thus determine the at least one desk.

[0224] In some embodiments of this application, the determining unit 1302 is further configured to determine a student activity area corresponding to the one desk by extending the boundary along a first direction to adjacent desks based on the position information of one of the at least one desks, thereby determining at least one student activity area;

[0225] The determining unit 1302 is further configured to determine the area in the at least one student activity area containing an object of a preset type as the target student activity area.

[0226] In some embodiments of this application, the acquisition unit 1301 is also used to acquire panoramic images of desks in the classroom and to acquire and identify student face information in each student activity area of ​​the global student activity area.

[0227] In some embodiments of this application, the determining unit 1302 is further configured to identify the panoramic image of the desk based on the desk desktop recognition model, and determine the panoramic desk.

[0228] The determining unit 1302 is also used to determine the global student activity area based on the panoramic desk;

[0229] The determining unit 1302 is further configured to match the student facial information in the student activity area with the facial information in the identity feature database to determine the student identity information in each student activity area;

[0230] The determining unit 1302 is further configured to map the student identity information within each student activity area to the corresponding student activity area, thereby determining the mapping relationship between the preset student activity area and the student identity information.

[0231] In some embodiments of this application, the acquisition unit 1301 is further configured to acquire student face images of each student activity area in the global student activity area using a zoom acquisition device.

[0232] In some embodiments of this application, the acquisition unit 1303 is further configured to perform face recognition on the student's face image to obtain a recognition result;

[0233] The acquisition unit 1303 is also used to adjust the focus of the zoom acquisition device and continue to acquire and identify if the recognition result indicates that the recognition has failed, until the student face information of each student activity area is obtained.

[0234] The acquisition unit 1303 is further configured to obtain the student face information of each student activity area if the recognition result indicates successful recognition.

[0235] In some embodiments of this application, the identity recognition device 13 further includes a receiving unit 1304 and an updating unit 1305; wherein,

[0236] The receiving unit 1304 is used to receive student modification requests;

[0237] The update unit 1305 is used to update the mapping relationship between the preset student activity area and student identity information.

[0238] In some embodiments of this application, the receiving unit 1304 is further configured to receive a region division request;

[0239] In an embodiment of this application, the updating unit 1305 is further configured to re-divide and identify the desk area to update the mapping relationship between the preset student activity area and student identity information.

[0240] Understandably, with the aforementioned device implementation, the terminal only needs to determine the target student's activity area, and then use the mapping relationship between the student's activity area and student identity information to determine the target student's identity. During this process, it is not necessary to obtain the student's facial information. Therefore, even when student facial images cannot be captured or the facial images of students in the back rows are blurry, student identity can still be correctly identified, improving the accuracy of recognition.

[0241] Based on the identity recognition method described in the above embodiments, this application also provides an identity recognition device, such as... Figure 14 As shown, Figure 14 A schematic diagram of the structure of an identity recognition device provided in this application embodiment. Figure 2 The device includes a processor 1401 and a memory 1402; the memory 1402 stores one or more programs executable by the processor, and when one or more programs are executed, the processor 1401 executes any of the identity recognition methods described in the previous embodiments.

[0242] Based on the identity recognition method in the above embodiments, this application also provides a computer-readable storage medium that stores one or more programs. The one or more programs can be executed by one or more processors, and when the programs are executed by the processors, they implement the identity recognition method as described in the embodiments of this disclosure.

[0243] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0244] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0245] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0246] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0247] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention.

Claims

1. An identity recognition method, characterized in that, The method includes: Capture images of desks in a classroom; Based on the desk desktop recognition model, the desk scene image is recognized to identify at least one desk; Based on the at least one desk, determine the target student activity area; Based on the pre-defined mapping relationship between student activity areas and student identity information, the identity information of the target student corresponding to the target student activity area is determined; The step of identifying at least one desk based on a desk desktop recognition model includes: Based on the desk desktop recognition model, the desk scene image is recognized to determine the location information of at least one desk, thereby identifying the at least one desk; Determining the target student activity area based on the at least one desk includes: Based on the location information of one of the at least one desks, the boundary is extended along the first direction to the adjacent desks to determine a student activity area corresponding to the one desk, thereby determining at least one student activity area; The area containing objects of a preset type in the at least one student activity area is determined as the target student activity area.

2. The method according to claim 1, characterized in that, Before identifying at least one desk in the desk scene image based on the desk desktop recognition model, the method further includes: Obtain classroom desk scene samples; An initial desk desktop recognition model is used to extract features from the desk scene samples to obtain sample feature images; and the sample feature images are then used for classification training to obtain training results. The initial desk desktop recognition model is trained based on the training results. If the error between the training results and the true values ​​of the samples does not meet the preset threshold, the desk desktop recognition model is trained again using the training results until the error between the training results and the true values ​​of the samples meets the preset threshold, thus obtaining the desk desktop recognition model.

3. The method according to claim 2, characterized in that, The initial desk desktop recognition model is used to extract features from the desk scene samples to obtain sample feature images, including: Using the initial desk desktop recognition model, the desk scene samples are divided into N levels to obtain N-level desk scene sub-samples; where N is a positive integer greater than 1. The i-th level desk scene sub-sample in the N-level desk scene sub-sample is sampled at K different resolutions to obtain the i-th level sample feature image, thus obtaining N sample feature images; where i is greater than or equal to 1 and less than or equal to N; K is a positive integer greater than 1; The N sample feature images are fused to obtain the sample feature image.

4. The method according to claim 3, characterized in that, The step of sampling the i-th level desk scene sub-sample from the N-th level desk scene sub-samples at K different resolutions to obtain the i-th level sample feature image includes: The i-th level desk scene sub-sample in the N-level desk scene sub-sample is sampled at the j-th layer to obtain the j-th layer sampled feature image; Obtain the other j-th layer sampled feature images corresponding to the other level desk scene sub-samples in the N-level desk scene sub-samples, excluding the i-th level desk scene sub-sample; After fusing the j-th layer sampled feature image with the other j-th layer sampled feature images, sampling of the (j+1)-th layer continues until sampling of the K-th layer is reached, resulting in the k-th layer sampled feature image; the k-th layer sampled feature image is used as the i-th level sampled feature image.

5. The method according to claim 1, characterized in that, Before determining the identity information of the target student corresponding to the target student activity area based on the preset mapping relationship between student activity areas and student identity information, the method further includes: Capture panoramic images of desks in the classroom; Based on the desk desktop recognition model, the panoramic image of the desk is recognized to determine the panoramic desk; Based on the panoramic desks, the overall student activity area is determined; Collect and identify student facial information for each student activity area within the global student activity area; The student facial information within the student activity area is matched with the facial information in the identity feature database to determine the student identity information within each student activity area; The student identity information within each student activity area is mapped to the corresponding student activity area to determine the preset mapping relationship between student activity areas and student identity information.

6. The method according to claim 5, characterized in that, The process of collecting and identifying student facial information for each student activity area within the global student activity area includes: A zoom acquisition device was used to acquire student face images for each student activity area within the global student activity area; The student's facial image was subjected to facial recognition to obtain the recognition result; If the recognition result indicates that the recognition has failed, the zoom acquisition device is refocused and the acquisition and recognition are continued until the student face information of each student activity area is obtained. If the recognition result indicates successful recognition, then the student's facial information for each student activity area is obtained.

7. The method according to claim 5 or 6, characterized in that, After mapping the student identity information within each student activity area to the corresponding student activity area to determine the preset mapping relationship between student activity areas and student identity information, the method further includes: Receive student modification requests; In response to the student's modification request, the mapping relationship between the preset student activity area and student identity information is updated.

8. The method according to claim 5 or 6, characterized in that, After mapping the student identity information within each student activity area to the corresponding student activity area to determine the preset mapping relationship between student activity areas and student identity information, the method further includes: Receive region partitioning requests; In response to the area division request, the desk area is re-divided and identified, and the mapping relationship between the preset student activity area and student identity information is updated.

9. An identity recognition device, characterized in that, The identity recognition device includes a data acquisition unit and a determination unit; wherein... The acquisition unit is used to acquire images of desks in the classroom. The determining unit is used to identify the desk scene image based on the desk desktop recognition model to determine at least one desk; to determine the target student activity area based on the at least one desk; and to determine the identity information of the target student corresponding to the target student activity area based on the preset mapping relationship between the student activity area and the student identity information. The determining unit is further configured to: recognize the desk scene image based on the desk desktop recognition model; determine the location information of the at least one desk; thereby determining the at least one desk; based on the location information of one of the at least one desks, extend the boundary along a first direction to adjacent desks; determine a student activity area corresponding to the one desk; thereby determining at least one student activity area; and determine the area in the at least one student activity area containing objects of a preset type as the target student activity area.

10. An identity recognition device, characterized in that, include: Memory, used to store executable data instructions; A processor, when executing executable instructions stored in the memory, implements the identity recognition method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, It stores executable instructions for causing a processor to execute, thereby implementing the identity recognition method according to any one of claims 1 to 8.

Images