A conference system and a conference management method

Abstract

The present disclosure provides a conference system and a conference management method, the conference system comprising: an image acquisition unit, a control unit, a reference database, an identity recognition unit and a calculation unit, the image acquisition unit acquires at least one set of images of attendees, the at least one set of images comprising multiple frames of images at a time before or after the start of a conference; the identity recognition unit extracts a feature vector of each face image in the multiple frames of images of the attendees, and compares with the reference database to obtain a comparison result, and obtains an identity recognition result of the attendees according to the comparison result; the calculation unit obtains a conference evaluation result according to the identity recognition result. The use of multiple frames of images and more than two sets of images improves the accuracy of conference automatic check-in and in-conference monitoring.

Description

Technical Field

[0001] This disclosure relates to the field of office automation technology, and in particular to a conference system and conference management method. Background Technology

[0002] With the development of network technology, video communication methods such as video conferencing, video teaching, and video telephony are becoming increasingly popular. Existing online conferencing systems include automatic attendance technology based on facial recognition. This technology uses a facial recognition module to collect and verify the facial photos of participants, and attendance is recorded after verification. For details, please refer to CN212624175U.

[0003] The aforementioned automatic meeting check-in technology based on facial recognition is relatively simple. When facing meetings in various occasions, the accuracy of meeting check-in is low. It cannot identify whether a photo is used to join the meeting, nor can it know whether the participants are continuously present, resulting in a low level of meeting management. Summary of the Invention

[0004] In view of this, the purpose of this disclosure is to propose a meeting system and meeting management method that can solve at least one of the technical problems of low accuracy of automatic meeting check-in and inability to monitor or quantitatively monitor the meeting status of participants.

[0005] To achieve the above objectives, this disclosure provides a conference system, including: an image acquisition unit, a control unit, a reference database, an identity recognition unit, and a computing unit.

[0006] The image acquisition unit acquires at least one set of images containing the attendees, and the at least one set of images includes multiple frames of images at a certain moment before or after the start of the meeting;

[0007] The control unit controls the image acquisition unit to acquire images according to preset image acquisition parameters;

[0008] The benchmark database includes the identity information of the participants and the benchmark feature vectors of their facial images;

[0009] The identity recognition unit extracts the feature vector of each attendee's face image from each frame of the multi-frame image, and compares it with the benchmark database to obtain the identity recognition result of each attendee.

[0010] The calculation unit obtains the participation evaluation result based on the identity recognition result.

[0011] Optionally, the face image reference feature vector includes face image reference feature vectors from multiple different angles.

[0012] Optionally, the benchmark database may also include body shape image benchmark feature vectors.

[0013] Optionally, the body shape image reference feature vector includes multiple body shape image reference feature vectors from different angles.

[0014] Optionally, the identity recognition unit further extracts the feature vector of the attendee's body shape image in each frame of the multi-frame images, and compares it with a benchmark database to obtain the identity recognition result of the attendee.

[0015] Optionally, the participation evaluation results include whether attendance was recorded.

[0016] Optionally, the image acquisition parameters include the shooting time of each group of images, or the time interval between shooting each group of images, or the start time, end time, or duration of the meeting.

[0017] Optionally, the participation evaluation results include the full participation rate, which is calculated as: (Total meeting duration - Cumulative absence time of the participant) / Total meeting duration.

[0018] This disclosure also provides a meeting management method, including the following steps:

[0019] S10, acquire at least one set of images containing attendees according to image acquisition parameters, wherein the at least one set of images includes multiple frames of images at a certain moment before or after the start of the meeting;

[0020] S20, extract the feature vector of each face image in the multi-frame image of the attendees;

[0021] S30, compare the feature vector obtained in step S20 with the reference data to obtain the comparison result of each frame of the face image of the attendee;

[0022] S40, confirm the identity of the attendee based on the comparison result of the facial image;

[0023] S50, obtained the evaluation results of the participants.

[0024] Optionally, step S40 further includes:

[0025] S41, determine whether the comparison results of each frame of the face image of the same attendee are the same. If they are the same, confirm the identity of the attendee based on the comparison results.

[0026] S42, if the comparison results of each frame of the face image of the same attendee are different, then the identity recognition of the attendee is unsuccessful, and the feature vector of each frame of the body image of the attendee whose identity recognition is unsuccessful is extracted.

[0027] S43, compare the feature vector of the body shape image obtained in step S42 with the reference data to obtain the comparison result of each frame of the body shape image of the attendee.

[0028] Optionally, step S44 further includes:

[0029] S45, determine whether the comparison results of each frame of the attendee's body image are the same; if they are the same, confirm the attendee's identity based on the comparison results.

[0030] S46. If the comparison results of each frame of the attendee's body image are different, the attendee's identity recognition is unsuccessful. Then, the attendee's image data is filtered out and manually identified.

[0031] Optionally, the evaluation results may include whether attendance was recorded, the overall attendance rate, or the attendance rate.

[0032] As described above, compared with the prior art, the technical solution provided in this disclosure achieves the following beneficial effects:

[0033] 1. At least one set of multi-frame images is used for attendee identification. Because the image capture time interval of multi-frame images is short and the similarity between multi-frame images is high, the impact of changes in shooting background, shooting light intensity, shooting angle, etc. caused by the continuous movement of people during the meeting on subsequent image recognition is effectively avoided, thus improving the accuracy of automatic meeting check-in and attendance monitoring. Using two or more sets of multi-frame images with a certain time interval for recognition avoids face recognition failure caused by face occlusion, head tilting, head turning, etc. at a certain shooting moment, thus improving the accuracy of personnel identification.

[0034] 2. After facial image recognition fails, body image is used for further identity recognition, which effectively improves the accuracy of personnel identification and the accuracy and efficiency of automatic meeting check-in and meeting monitoring.

[0035] 3. By using facial image feature vectors and body image feature vectors from multiple different angles as benchmark data for comparison, image recognition failures caused by image shooting angles are effectively avoided, improving the accuracy and efficiency of automatic meeting check-in and meeting monitoring.

[0036] 4. If facial or body image recognition fails, the unsuccessful image data will be filtered out for further manual recognition. This improves the efficiency of automatic meeting check-in while ensuring the accuracy of meeting check-in and attendance monitoring.

[0037] 5. In the process of facial and body image recognition, data with high frequency of participant IDs are filtered out for further recognition, which further improves the efficiency of automatic check-in and attendance monitoring.

[0038] 6. The full attendance rate is used to quantitatively evaluate the continued attendance of attendees, which improves the level of meeting management and meeting efficiency. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in this disclosure or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a schematic diagram of an embodiment of the conference system disclosed herein;

[0041] Figure 2 This is a schematic diagram of another embodiment of the conference system disclosed herein;

[0042] Figure 3 This is a schematic diagram of yet another embodiment of the conference system disclosed herein;

[0043] Figure 4 A schematic diagram of the baseline database for the conference system disclosed herein;

[0044] Figure 5 This is a schematic diagram of the identity recognition unit of the conference system disclosed herein;

[0045] Figure 6 This is a schematic diagram of an embodiment of the meeting management method disclosed herein;

[0046] Figure 7 This is a schematic diagram of another embodiment of the meeting management method disclosed herein;

[0047] Figure 8 This is a schematic diagram of yet another embodiment of the meeting management method disclosed herein. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0049] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in the embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0050] like Figure 1 As shown, this is an embodiment of a conference system disclosed herein, including: an image acquisition unit, a control unit, a reference database, an identity recognition unit, and a computing unit.

[0051] The image acquisition unit is used to acquire at least one set of images containing attendees, the at least one set of images including multiple frames of images at a certain moment before or after the start of the meeting.

[0052] Understandably, during automatic check-in or monitoring of attendance, the at least one set of images can be one, two, or more sets. The at least one set of images can be acquired within a predetermined time period before or after the meeting begins, for example, within 10 minutes before or after the meeting begins, as part of the check-in process, specifically set according to the actual situation. A given moment can be less than or equal to 2 seconds, preferably less than or equal to 1 second. The acquisition interval for each set of images is 20s to 120s, preferably 60s.

[0053] The at least one set of images consists of multiple consecutive frames at a specific moment, or multiple frames with a predetermined time interval, such as multiple frames with intervals of 10ms to 50ms. The multiple frames have a high degree of similarity. Using these multiple frames effectively avoids changes in the shooting background, lighting intensity, and shooting angle caused by continuous movement of people during the meeting, improving the success rate of image acquisition and reducing the impact of poor image acquisition quality on subsequent image recognition. This, in turn, improves the accuracy of automatic meeting check-in and attendance monitoring.

[0054] The image is a face image, which can be a face image from any angle.

[0055] Optionally, the multi-frame images further include body images, which can be body images from any angle, and the body images can be full-body images, half-body images, or headshots.

[0056] The image acquisition unit is preferably a camera, capable of capturing multiple frames of images or videos containing participants during the meeting;

[0057] Optionally, the image acquisition unit is an image or video data receiving device, capable of receiving or extracting the required image data from the image or video data captured by the camera.

[0058] Optionally, the image acquisition unit records the image capture time while acquiring the image.

[0059] Optionally, the image acquisition unit stores the acquired images or videos in the conference memory and labels them with a conference ID.

[0060] The control unit sets the image acquisition parameters of the image acquisition unit and controls the image acquisition unit to acquire images according to the set image acquisition parameters.

[0061] Optionally, the image acquisition parameters can be set through the human-machine interface of the control unit.

[0062] Optionally, the image acquisition parameters include: the number of image groups, the acquisition time and time interval for each image group, the number of frames per image group, the acquisition method for multiple frames (continuous acquisition or interval acquisition), and the time interval for each frame. In addition to the image acquisition parameters, meeting information, such as meeting ID, participants, meeting start time, end time, and meeting duration, can be set through the control unit. Meeting plans and historical data can be obtained, along with image comparison parameters, such as a first threshold, a second threshold, a frequency threshold, and the recognition interval for each image group. The recognition interval is used for monitoring attendance and calculating the cumulative absence time of participants.

[0063] Optionally, the control unit does not include memory and directly receives or processes video stream data.

[0064] Optionally, the control unit includes a memory for storing image or video data acquired by the image acquisition unit, such as... Figure 2 As shown.

[0065] The preferred method for controlling the image acquisition unit is automatic control. A timer in the control unit automatically controls the image acquisition device to acquire images or videos according to preset shooting parameters. Specifically, in one embodiment, the timer can trigger the camera to start shooting at the start of the meeting, capturing a set of images at each predetermined time point, and stopping recording at the end of the meeting.

[0066] Optionally, the image acquisition unit can be controlled by manual triggering. Specifically, after setting the shooting parameters through the human-machine interface of the control unit, the camera device can be manually clicked to start shooting.

[0067] The benchmark database stores benchmark information of the participants, such as... Figure 4 As shown; the benchmark information specifically includes feature vectors of multiple facial images of attendees from different angles, pre-recorded from the participants, as well as the corresponding attendee ID, name, gender, age, etc. For example, the benchmark database stores benchmark information for S attendees, corresponding to S records, each record including P n There are several fields. For participant S1, field P1 is the ID of participant S1, and fields P2 to P... i P1 is the feature vector of the face image of the participant S1 from i-1 different angles, P2 is the feature vector of the frontal face image of the participant S1, P3 is the feature vector of the left side face image of the participant S1, and so on.

[0068] The identity recognition unit is used to identify the actual attendees (hereinafter referred to as attendees) by performing facial recognition, such as... Figure 5 As shown, the system includes a feature extraction module and a comparison module. The feature extraction module extracts the feature vector of each face in at least one set of multi-frame face images. Optionally, the extraction method involves processing the face images using a face recognition algorithm to obtain the feature vector of each face. Specifically, assuming that the image acquisition unit obtains a set of images including N frames, and each frame includes M attendees, a total of N×M face images are obtained. Feature extraction is performed on these N×M face images sequentially to finally obtain the feature vectors of N×M faces.

[0069] The comparison module compares the feature vector of each face obtained by the feature extraction module with the benchmark features of the participants stored in the benchmark database one by one, and obtains the comparison result according to the preset first threshold.

[0070] Optionally, the identification unit includes a memory for storing the comparison results.

[0071] Specifically, the feature vector of the first frame of the face image in a certain group of images of the first attendee is compared with the feature vector of the first record (P2~P) stored in the benchmark database. iThe facial feature vectors from different angles are compared one by one. The comparison method can be subtraction or quotient, etc. The comparison result is compared with a preset first threshold. If it is greater than or equal to the first threshold, the identity of the face is considered to have been identified in this comparison. The ID of the P1 field of the current comparison record is extracted as the comparison result, denoted as D1. If it is less than the first threshold, the comparison result is not accepted, and the comparison continues with the second record (P2~P1). i The facial feature vectors from different angles are compared one by one until they are compared with S records to obtain the comparison result.

[0072] After comparing the first frame of the face image of the first attendee, the feature vector of the second frame of the face image of the first attendee in the set of images is then compared with the feature vectors of the S records (P2~P) stored in the benchmark database. i The facial feature vectors from different angles are compared one by one. The comparison result is compared with a preset first threshold. If it is greater than or equal to the first threshold, the identity of the face is considered to have been identified in this comparison, and the ID of the P1 field of the current comparison record is extracted as the comparison result, denoted as D2. If it is less than the first threshold, the comparison result is not accepted, and the comparison continues with the second record (P2~P...). i The facial feature vectors from different angles are compared one by one until they are compared with S records to obtain the comparison result.

[0073] Similarly, the feature vectors of N frames from the first attendee's image set are compared with the baseline data, and a total of D features are obtained for each image set. N One comparison result.

[0074] Similarly, the face images of the second attendee in a certain set of images are identified in N frames, and so on until the Mth attendee is identified, ultimately obtaining the M attendees and the D of each attendee. N One comparison result.

[0075] Understandably, if the image acquisition unit acquires two or more sets of images, it will continue to recognize the next set of images after the recognition of the first set of images is completed, until the recognition of all sets of images is completed.

[0076] The calculation unit obtains the meeting participation evaluation result or meeting statistics result based on the comparison results. The meeting participation evaluation result includes whether attendance was recorded, and if the same attendee's D... N If all the comparison results are the same and all are the same ID, then the identity of the attendee has been successfully identified and the attendee has been automatically confirmed to have checked in.

[0077] By confirming the attendance of those who have checked in, you can identify those who have not. If an identified attendee is not part of the current meeting, they will be ignored.

[0078] In another embodiment, the baseline database also includes pre-obtained omnidirectional body shape image feature vectors of attendees from multiple different angles, stored in the same way as the aforementioned facial image feature vectors, for example, field P i+1 ~P j Used to store feature vectors of body images from multiple different angles.

[0079] The identity recognition unit further performs body shape image recognition. When the same attendee's D N If the comparison results are not unique, it indicates that the attendee's identity was not successfully identified. The system filters out multiple attendee IDs whose IDs appear most frequently or exceed a predetermined threshold from the comparison results. For attendees whose identities were not successfully identified after face recognition, the feature extraction module extracts the feature vector from each frame of the attendee's multi-frame body image in the image set. Optionally, the extraction method involves processing the body image using a body shape recognition algorithm to obtain the feature vector. Specifically, assuming that m attendees failed to be identified after face recognition of a certain set of images, feature vectors are extracted sequentially from N frames of body image images of these m attendees, resulting in a total of m×N feature vectors for each body image.

[0080] The comparison module compares the feature vectors of m people's N frames of body shape images with the baseline features of multiple selected personnel IDs one by one, and obtains the comparison results according to the preset second threshold.

[0081] Specifically, assuming that during facial recognition of an attendee, s attendee IDs are selected, the feature vector of the first frame of the attendee's N-frame body image is compared with the P vector of the first record in the s records stored in the benchmark database. i+1 ~P j Each body feature vector from different angles is compared one by one. The comparison result is then compared with a preset second threshold. If the comparison result is greater than or equal to the second threshold, the identity of the body shape is considered to have been identified, and the ID of the P1 field of the currently compared record is extracted as the comparison result, denoted as d1. If the comparison result is less than the second threshold, the comparison result is not accepted, and the comparison continues with the P1 field of the second record. i+1 ~P j The facial feature vectors from different angles are compared one by one until they are compared with s records to obtain the comparison result.

[0082] Similarly, the feature vectors of N frames of images of the attendee are compared with s reference data points to obtain a total of d. N One comparison result.

[0083] Similarly, the body shape images of the second attendee in a certain set of images are identified in N frames, and so on until the m-th attendee, ultimately obtaining the m attendees and the d of each attendee. N One comparison result.

[0084] Understandably, if the image acquisition unit acquires two or more sets of images, it will continue to recognize the next set of images after the recognition of the first set of images is completed, until the recognition of all sets of images is completed.

[0085] The calculation unit calculates based on the d of each attendee. N The comparison results yield the evaluation results for the participants. If d N If all the comparison results are the same and all are the same ID, then the attendee's identity has been successfully identified, and it is automatically determined that the attendee has checked in.

[0086] In another embodiment, such as Figure 3 As shown, when d N If the comparison result is not unique, it means that the identity of the attendee failed to be identified. Extract the data of the failed identification, along with the relevant meeting ID, for manual comparison.

[0087] In another embodiment, the meeting attendance evaluation results also include the overall attendance rate, which enables monitoring of attendance during the meeting and quantitatively evaluates the attendance of participants. Overall attendance rate = (Total meeting duration - Cumulative absence time of the participant) / Total meeting duration.

[0088] The recognition interval for each group of images is set by the control unit to calculate the cumulative absence time of the person. If the person is recognized within a certain recognition interval, they are considered to have been present continuously during that interval. If the person is not recognized within that interval by facial image recognition, body image recognition, or manual recognition, they are considered to have been absent during that interval, and the absence time for this recognition is equal to the recognition interval. The same person is recognized throughout the entire meeting at the set recognition intervals, and the cumulative absence time is obtained by accumulating these intervals.

[0089] Optionally, the identification interval is 1 to 10 minutes, preferably 3 minutes, and can be selected according to the actual situation such as the duration of the meeting.

[0090] Specifically, assuming the recognition interval is set to 3 minutes, if a person is recognized at least once within this interval, they are considered to have been present throughout the 3-minute interval. If, within this 3-minute interval, the person is not recognized by either facial or body image recognition, the system prompts for manual verification. If the person is still not recognized after manual verification, their absence is considered to be 3 minutes. After this 3-minute interval, the system begins recognizing the person for the next 3-minute interval, continuing this process for the entire meeting. The total absence time for the person is then calculated by summing the results for the entire meeting duration.

[0091] The results of manual identification are used to label participants, including those who attended the entire meeting, were late, left early, did not attend the entire meeting, and did not attend at all.

[0092] Optionally, the recognition interval can be shortened during the start or end of the meeting to facilitate the identification of whether participants have attended the entire meeting and to automatically determine whether they are late or have left early, and to modify the manually marked parts.

[0093] The calculation unit is connected to the image acquisition unit, the control unit, and the identity recognition unit. It obtains the capture time of each set of images from the image acquisition unit and combines this with the recognition results from the identity recognition unit to determine whether a person was recognized at different time points. By accumulating these results, the unit calculates the duration for which the person was not recognized. The calculation unit obtains the meeting start time and end time from the control unit to obtain the total meeting duration. Finally, it calculates the overall attendance rate using a specific formula.

[0094] In another embodiment, the participation evaluation results include attendance rate, which is calculated as: attendance rate = actual number of attendees / total number of participants.

[0095] Another embodiment of this disclosure also provides a meeting management method, such as... Figure 6 As shown, it includes the following steps:

[0096] S10, acquire at least one set of images containing attendees according to image acquisition parameters, wherein the at least one set of images includes multiple frames of images at a certain moment before or after the start of the meeting;

[0097] Optionally, before step S10, image acquisition parameters are set, including: the shooting time for each group of images, the number of shooting groups, the shooting interval for each group of images, the recognition interval for each group of images, the number of frames captured for each group of images, the acquisition method for each group of images (continuous acquisition or interval acquisition), and the interval time for each frame. In addition to setting image acquisition parameters, meeting information can also be set, such as meeting ID, participants, meeting start time, end time, and meeting duration, to obtain meeting plans and historical meeting data; image comparison parameters, such as a first threshold, a second threshold, a frequency threshold, and the recognition interval for each group of images.

[0098] Image acquisition can begin at some point before or after the meeting starts, such as 0-10 minutes before the meeting begins as a check-in process, and continue until the meeting ends. The acquired images are stored in the meeting image storage device and associated with the meeting's ID. Optionally, a check-in process can be omitted, and the camera can be automatically triggered to start recording video at the start of the meeting, continuing until the meeting ends.

[0099] The images, as well as their acquisition methods, shooting times, and storage conditions, are the same as in other embodiments.

[0100] S20, extract the feature vector of each face image in the multi-frame image of the attendees;

[0101] S30, compare the feature vector of the face image obtained in step S20 with the reference data to obtain the comparison result of each frame of the face image of the attendee; the comparison result is the attendee ID, which is obtained according to a first threshold, that is, compare the feature vector comparison result with the preset first threshold. If it is greater than or equal to the first threshold, it is considered that the identity of the face image in this comparison has been identified, and the attendee ID corresponding to the current comparison reference data is extracted as the comparison result; the reference data includes face image feature vectors from multiple different angles;

[0102] Optionally, step S30 further includes,

[0103] S31, filter out the attendee IDs that appear more frequently than a predetermined threshold, or the IDs with the highest frequency; IDs with high frequency indicate a high probability that they are the owners of the image.

[0104] S40, confirm the identity of the attendee based on the comparison result of the facial image;

[0105] S50, obtained the evaluation results of the participants.

[0106] like Figure 7 As shown, in another embodiment of this disclosure, optionally, step S40 further includes:

[0107] S41, determine whether the comparison results of each frame of the face image of the same attendee are the same. If they are the same, the comparison result is unique. Confirm the identity of the attendee based on the comparison result and use the person ID corresponding to the comparison result as the attendee's identity ID.

[0108] S42, if the comparison results of each frame of the face image of the same attendee are different and the comparison results are not unique, then the identity recognition of the attendee is unsuccessful. Extract the feature vector of each frame of the body image of the attendee whose identity recognition is unsuccessful.

[0109] S43, compare the feature vector of the body shape image obtained in step S42 with the reference data to obtain the comparison result of each frame of the attendee's body shape image; the comparison result is the attendee ID, which is obtained according to the second threshold, that is, compare the feature vector comparison result with the preset second threshold. If it is greater than or equal to the second threshold, it is considered that the identity of the frame of body shape image has been identified in this comparison, and the attendee ID corresponding to the current comparison reference data is extracted as the comparison result; the reference data includes body shape image feature vectors from multiple different angles;

[0110] S44. The identity of the attendee is confirmed based on the comparison result of the body image.

[0111] Optionally, in step S43, the reference data is the reference data corresponding to the ID selected in step S31; by using the reference data of the person ID selected in face image recognition as the comparison data for the next step, it is not necessary to compare all the reference data, thus improving the efficiency of identity recognition.

[0112] like Figure 8 As shown, in another embodiment of this disclosure, optionally, step S44 further includes:

[0113] S45, determine whether the comparison results of each frame of the attendee's body image are the same. If they are the same, it means that a unique person has been matched. Then, the identity of the attendee is confirmed based on the comparison result, that is, the person ID corresponding to the comparison result is used as the identity recognition result.

[0114] S46. If the comparison results of each frame of the attendee's body image are different and the comparison results are not unique, then the attendee's identity recognition is unsuccessful. In this case, the image data of the attendee whose recognition was unsuccessful is filtered out and manually identified.

[0115] Optionally, if face recognition fails in step S42, body image recognition is not performed, and manual recognition is performed directly.

[0116] The method for comparing the facial image feature vector, body shape image feature vector, and benchmark data in this embodiment is the same as in other embodiments.

[0117] Optionally, the evaluation results may include whether attendance was recorded, the overall attendance rate, or the attendance rate.

[0118] If an attendee's ID is detected during a specific time period, such as the check-in process, their check-in will be automatically confirmed. Based on the checked-in attendees and the pre-defined list of expected attendees, those who did not check in can be identified.

[0119] Optionally, the meeting attendance evaluation results also include the overall attendance rate, used to monitor attendance during the meeting and quantitatively evaluate the attendance of participants. The overall attendance rate is calculated as: (Total meeting duration - Cumulative absence time of the participant) / Total meeting duration. The cumulative absence time of a participant is calculated by defining a recognition interval for each set of images. If a participant is identified within a certain recognition interval, they are considered to have been present continuously during that interval. If, within that recognition interval, the participant is not identified by facial image recognition, body image recognition, or manual recognition, they are considered not present during that interval, and the absence time for this identification equals the recognition interval. The same participant is identified throughout the entire meeting at the set recognition intervals, and the cumulative absence time is obtained by accumulating these intervals.

[0120] Optionally, the identification interval is 1 to 10 minutes, preferably 3 minutes, and can be selected according to the actual situation such as the duration of the meeting.

[0121] Specifically, assuming the recognition interval is set to 3 minutes, if a person is recognized at least once within this interval, they are considered to have been present throughout the 3-minute interval. If, within this 3-minute interval, the person is not recognized by either facial or body image recognition, the system prompts for manual verification. If the person is still not recognized after manual verification, their absence is considered to be 3 minutes. After this 3-minute interval, the system begins recognizing the person for the next 3-minute interval, continuing this process for the entire meeting. The total absence time for the person is then calculated by summing the results for the entire meeting duration.

[0122] The results of manual identification are used to label participants, including those who attended the entire meeting, were late, left early, did not attend the entire meeting, and did not attend at all.

[0123] Optionally, the recognition interval can be shortened during the start or end of the meeting to facilitate the identification of whether participants have attended the entire meeting and to automatically determine whether they are late or have left early, and to modify the manually marked parts.

[0124] By calculating the full attendance rate, the system can automatically and quantitatively evaluate whether attendees participated in the entire meeting, thereby improving the efficiency and quality of meeting management and enhancing the overall meeting experience.

[0125] The attendance rate is calculated as: actual number of attendees / total number of participants.

[0126] Optionally, the face image can be obtained by directly acquiring the face image through the face positioning bounding box at the time of shooting, or by segmenting the captured body image to obtain both the face image and the body image.

[0127] In some embodiments, the settings of the first threshold, the second threshold, and the predetermined threshold can be configured for the face and body shape according to different algorithms called and actual needs, to help the processor filter the results output by the algorithm.

[0128] In the above embodiments, the feature vector of the reference data is obtained in the following way: multiple face images or body images of the participants from different angles, such as front, left, right, back, top, and bottom, are pre-entered, and then the recognition algorithm is called to perform face modeling and body modeling to obtain the reference feature vector of the participants' face image and the reference feature vector of the body image.

[0129] The modeling and feature vector calculation can be achieved by identifying specific points in the image and calculating the proportions.

[0130] Optionally, the facial feature vector includes: face length / width, face aspect ratio, face skin color, and the width or height of local facial details, such as the mouth, nose, and ears. Optionally, the facial recognition algorithm includes: a recognition algorithm based on facial feature points, a recognition algorithm based on the entire face image, a template-based recognition algorithm, and a neural network-based recognition algorithm.

[0131] Optionally, the body image can be a full-body image, a half-body image, or a headshot. The body feature vector includes: head length / width, head length-to-width ratio, shoulder width, head width to shoulder width ratio, vertical distance between the top of the head and the shoulder, vertical distance between the chin and the shoulder, and distance from the shoulder to the ground, etc.

[0132] Optionally, methods for obtaining body shape image feature vectors include inputting sampled photos into a deep learning network model, using big data-driven model self-learning to obtain usable body shape feature vectors; or using grayscale contour acquisition methods.

[0133] Compared with the prior art, the technical solution provided in this disclosure achieves the following beneficial effects:

[0134] 1. At least one set of multi-frame images is used for attendee identification. Because the image capture time interval of multi-frame images is short and the similarity between multi-frame images is high, the impact of changes in shooting background, shooting light intensity, shooting angle, etc. caused by the continuous movement of people during the meeting on subsequent image recognition is effectively avoided, thereby improving the accuracy of automatic meeting check-in and attendance monitoring. Using two or more sets of multi-frame images with a certain time interval for recognition avoids face recognition failure caused by face occlusion, head tilting, head turning, etc. at a certain shooting moment, thus improving the accuracy of personnel identification.

[0135] 2. After facial image recognition fails, body image is used for further identity recognition, which effectively improves the accuracy of personnel identification and the accuracy and efficiency of automatic meeting check-in and meeting monitoring.

[0136] 3. By using facial image feature vectors and body image feature vectors from multiple different angles as benchmark data for comparison, image recognition failures caused by image shooting angles are effectively avoided, improving the accuracy and efficiency of automatic meeting check-in and meeting monitoring.

[0137] 4. If facial or body image recognition fails, the unsuccessful image data will be filtered out for further manual recognition. This improves the efficiency of automatic meeting check-in while ensuring the accuracy of meeting check-in and attendance monitoring.

[0138] 5. In the process of facial and body image recognition, data with high frequency of participant IDs are filtered out for further recognition, which further improves the efficiency of automatic check-in and attendance monitoring.

[0139] 6. The full attendance rate is used to quantitatively evaluate the continued attendance of attendees, which improves the level of meeting management and meeting efficiency.

[0140] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description.

[0141] This disclosure is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A conferencing system, characterized by The method comprises the following steps: An image acquisition unit, a control unit, a reference database, an identity recognition unit and a calculation unit, The image acquisition unit acquires at least one set of images containing meeting participants, and the at least one set of images comprises a plurality of images at a time before or after the start of the meeting; The control unit controls the image acquisition unit to acquire images according to preset image acquisition parameters; The reference database comprises identity information and face image reference feature vectors of the meeting participants; The identity recognition unit extracts feature vectors of each face image in the plurality of images of the meeting participants, and compares the feature vectors with the reference database to obtain face image comparison results, and obtains identity recognition results of the meeting participants according to the comparison results; The calculation unit obtains meeting evaluation results according to the identity recognition results; The step of screening the face image comparison results comprises the following steps: The face image reference feature vectors comprise a plurality of face image reference feature vectors at different angles. The feature vector of the first frame of face image in a certain group of images of a participant obtained is compared with the P2~P i face feature vectors of different angles of the first record stored in the reference database one by one, and the comparison result is compared with a preset first threshold value. If the comparison result is greater than or equal to the first threshold value, it is considered that the identity of the face is recognized in this comparison, and the ID of the record in the current comparison is extracted as the comparison result. If the comparison result is less than the first threshold value, the comparison result is not accepted, and the comparison with the P2~P i face feature vectors of different angles of the second record is continued until the comparison with S records is completed, and the comparison result is obtained. The reference database stores reference information of S conference participants, corresponding to S records, and each record includes P n fields, field P1 is an ID of a conference participant, and fields P2-P i are face image feature vectors of the conference participant at i-1 different angles. By analogy, the feature vectors of the N frames of images in the first participant's image group are compared with the reference data in turn, and D N comparison results are obtained for each image group. If the D N of the same participant is the same, and is the same ID, then the identity of the participant is successfully identified, and it is automatically determined that the participant has signed in.

2. The conferencing system of claim 1, wherein, The reference database further comprises body image reference feature vectors.

3. The conferencing system of claim 1, wherein, The body image reference feature vectors comprise a plurality of body image reference feature vectors at different angles.

4. The conferencing system of claim 3, wherein, The identity recognition unit further extracts feature vectors of each body image in the plurality of images of the meeting participants, and compares the feature vectors with the reference database to obtain body image comparison results, and obtains identity recognition results of the meeting participants.

5. The conferencing system according to any of claims 3-4, characterized in that, The meeting evaluation results comprise whether to sign in, a full-time attendance rate or an attendance rate.

6. The conferencing system of claim 1, wherein, The method comprises the following steps:

7. A conference management method characterized by, S10, acquiring at least one set of images containing meeting participants according to image acquisition parameters, and the at least one set of images comprises a plurality of images at a time before or after the start of the meeting; S20, extracting feature vectors of each face image in the plurality of images of the meeting participants; S30, comparing the feature vectors obtained in step S20 with reference data to obtain comparison results of each face image of the meeting participants; S40, confirming the identity of the meeting participants according to the comparison results of the face images; S50, obtaining meeting evaluation results; The step S30 further comprises the following steps: S31, screening the face image comparison results to obtain results with a frequency greater than a predetermined threshold or a highest frequency; The step S31 comprises the following steps: The step S40 further comprises the following steps: S311, compare the feature vector of the first frame of face image in the obtained one group of images of one participant with the P2~P i face feature vectors of different angles in the first record stored in the reference database one by one, compare the comparison result with the preset first threshold value, if greater than or equal to the first threshold value, it is considered that the identity of the face is recognized this time, and the ID of the record of the current comparison is extracted as the comparison result; if less than the first threshold value, the comparison result is not accepted, and the P2~P i face feature vectors of different angles of the second record are compared one by one until the comparison result is obtained after comparing with S records. The reference database stores reference information of S conference participants, corresponding to S records, and each record includes P n fields, field P1 is an ID of a conference participant, and fields P2-P i are face image feature vectors of the conference participant at i-1 different angles. S313, if the D N of the same participant are the same, and are the same ID, then the identity of the participant is successfully identified, and it is automatically determined that the participant has signed in.

8. The conference management method according to claim 7, characterized by, S41, judging whether the comparison results of each face image of the same meeting participant are the same, and if so, confirming the identity of the meeting participant according to the comparison results; S42, if the comparison results of each face image of the same meeting participant are different, the identity recognition of the meeting participant is unsuccessful, and feature vectors of each body image in the plurality of images of the meeting participant whose identity recognition is unsuccessful are extracted; S43, comparing the feature vectors of the body images obtained in step S42 with reference data to obtain comparison results of each body image of the meeting participant; S44, confirming the identity of the meeting participant according to the comparison results of the body images. ​ 9. The conference management method according to claim 8, characterized by, The step S44 further comprises: S45, judging whether the comparison results of each frame body shape image of the meeting participant are same, if same, confirming the identity of the meeting participant according to the comparison result; S46, if the comparison results of each frame body shape image of the meeting participant are different, the identity recognition of the meeting participant is unsuccessful, then filtering out the image data of the meeting participant whose recognition is unsuccessful, and performing manual recognition.

Images