Queuing and calling method and device, equipment, storage medium and program product

Abstract

The invention provides a queuing and calling method and device, equipment, a storage medium and a program product, and relates to the field of artificial intelligence. According to the method, when a queuing machine generates a queuing number for a target user of a to-be-handled service, pedestrian features of the target user are extracted based on a pedestrian re-identification model according to an authorized image corresponding to the target user, and target features are obtained; when it is determined that the predicted waiting time corresponding to the target user is smaller than or equal to a preset threshold value, whether the target user is in a target area or not is determined based on a pedestrian re-recognition model according to the target features; wherein the target area comprises a business handling area and a waiting area; if the target user is not in the target area, delaying the queuing number of the target user, and sending prompt information to the target user; wherein the prompt information is used for prompting the target user to return to the target area, so that continuous waiting for the target user is avoided, time is saved, and service processing efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a queuing method, apparatus, device, storage medium, and program product. Background Art

[0002] The queuing system is widely used in various places where queuing is required, such as banks, hospitals, supermarkets, service agencies, etc., to efficiently manage waiting time, improve user experience and optimize service processes.

[0003] In traditional queuing systems, users typically obtain a queue number from a number machine upon arriving at a service location, and are then called on a first-come, first-served basis, resulting in low service processing efficiency. Queuing systems based on artificial intelligence or big data can intelligently queue and triage based on factors such as user type and service urgency, significantly improving service processing efficiency.

[0004] However, due to the randomness of user behavior, business processing efficiency is still low. Summary of the Invention

[0005] The present application provides a queuing calling method, apparatus, device, storage medium and program product for improving business processing efficiency.

[0006] In a first aspect, the present application provides a queuing method, comprising:

[0007] When the number-picking machine generates a queue number for a target user to be processed, the pedestrian features of the target user are extracted based on the authorized image corresponding to the target user and the pedestrian re-identification model to obtain the target features;

[0008] When it is determined that the expected waiting time corresponding to the target user is less than or equal to a preset threshold, determining whether the target user is in a target area based on the pedestrian re-identification model and the target characteristics; wherein the target area includes a business processing area and a waiting area;

[0009] If the target user is not in the target area, the queuing number of the target user is postponed and a prompt message is sent to the target user; wherein the prompt message is used to prompt the target user to return to the target area.

[0010] In a second aspect, the present application provides a queuing and calling device, comprising:

[0011] A target feature extraction module is used to extract the pedestrian features of the target user based on the pedestrian re-identification model according to the authorized image corresponding to the target user when the number-taking machine generates a queue number for the target user to be processed, thereby obtaining the target features;

[0012] a target user query module, configured to determine, when it is determined that the expected waiting time corresponding to the target user is less than or equal to a preset threshold, whether the target user is in a target area based on the person re-identification model and the target characteristics; wherein the target area includes a business processing area and a waiting area;

[0013] The queuing number postponing module is used to postpone the queuing number of the target user if the target user is not in the target area, and send a prompt message to the target user; wherein the prompt message is used to prompt the target user to return to the target area.

[0014] In a third aspect, the present application provides an electronic device, comprising: a processor, and a memory communicatively connected to the processor;

[0015] The memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method according to the first aspect.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method described in the first aspect.

[0017] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which implements the method described in the first aspect when executed by a processor.

[0018] The present application provides a queue calling method, device, equipment, storage medium and program product. When a number-picking machine generates a queue number for a target user to be processed, the method extracts the target user's pedestrian features based on the authorized image corresponding to the target user and a pedestrian re-identification model to obtain target features; when it is determined that the expected waiting time corresponding to the target user is less than or equal to a preset threshold, the method determines whether the target user is in a target area based on the pedestrian re-identification model and the target features; wherein the target area includes a business processing area and a waiting area; if the target user is not in the target area, the queue number of the target user is postponed, and a prompt message is sent to the target user; wherein the prompt message is used to prompt the target user to return to the target area. In this method, by postponing the queue number of a target user who is not in the target area, continuous waiting for the target user is avoided, time is saved, and business processing efficiency is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0020] Figure 1 A flowchart of a queuing method provided in an embodiment of the present application;

[0021] Figure 2 A schematic flow chart of a method for extracting target features provided in an embodiment of the present application;

[0022] Figure 3 A schematic diagram of the structure of a pedestrian re-identification model provided in an embodiment of the present application;

[0023] Figure 4 A schematic flow chart of a method for determining a target image provided in an embodiment of the present application;

[0024] Figure 5 A schematic diagram of the structure of a target detection model provided in an embodiment of the present application;

[0025] Figure 6 A schematic diagram of the structure of a time prediction model provided in an embodiment of the present application;

[0026] Figure 7 A schematic diagram of the structure of a queuing and calling device provided in an embodiment of the present application;

[0027] Figure 8 A schematic structural diagram of an ultrasonic flow meter control unit provided in an embodiment of the present application.

[0028] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0029] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0030] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0031] In addition, this application involves big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.), and the use of artificial intelligence technology for automated decision-making, and a technical solution for making decisions that have a significant impact on personal rights and interests based on the results of automated decision-making. The application provides users with corresponding operation entrances for users to choose to agree or reject the results of automated decision-making; if the user chooses to reject, the expert decision-making process will be entered.

[0032] It should be noted that the queuing and calling method, device, equipment, storage medium and program product provided in this application can be used in the field of artificial intelligence, and can also be used in any field other than the field of artificial intelligence. This application does not limit the application field of the queuing and calling method, device, equipment, storage medium and program product.

[0033] The queuing system is widely used in various places where queuing is required, such as banks, hospitals, supermarkets, service agencies, etc., to efficiently manage waiting time, improve user experience and optimize service processes.

[0034] In traditional queuing systems, users typically obtain a queue number from a number machine upon arriving at a service location, and are then called on a first-come, first-served basis, resulting in low service processing efficiency. Queuing systems based on artificial intelligence or big data can intelligently queue and triage based on factors such as user type and service urgency, significantly improving service processing efficiency.

[0035] However, due to the random nature of user behavior, service processing efficiency remains low. For example, in some queuing systems, after a user successfully obtains a number, a prediction of the user's estimated waiting time is made. Based on this estimated waiting time, the user may schedule other activities instead of waiting for their number to be called. This can result in the user being absent when their number is called and unable to access the service processing area, leading to a prolonged wait. Alternatively, after a period of waiting, the user's number may be canceled, forcing the user to obtain a number again upon returning, resulting in low overall service processing efficiency.

[0036] In order to solve the above technical problems, an embodiment of the present application proposes a queuing number calling method. When the number-picking machine generates a queuing number for a target user to be processed, the method extracts the pedestrian features of the target user based on the authorized image corresponding to the target user and the pedestrian re-identification model to obtain the target features; when it is determined that the expected waiting time corresponding to the target user is less than or equal to the preset threshold, the method determines whether the target user is in the target area based on the pedestrian re-identification model and the target features; wherein the target area includes the business processing area and the waiting area; if the target user is not in the target area, the queuing number of the target user is postponed, and a prompt message is sent to the target user; wherein the prompt message is used to prompt the target user to return to the target area. In this method, by postponing the queuing number of the target user who is not in the target area, continuous waiting for the target user is avoided, time is saved, and business processing efficiency is improved.

[0037] The following specific embodiments describe in detail the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0038] Figure 1 This is a flow chart of a queuing method provided by an embodiment of the present application. The execution subject of the method can be an electronic device for implementing the abnormality detection method, which is implemented by software and hardware, and can be a server deployed locally or in the cloud. Figure 1 As shown, the specific steps of this method are as follows:

[0039] S101. When a number-picking machine generates a queue number for a target user to process a service, pedestrian features of the target user are extracted based on an authorized image corresponding to the target user and a pedestrian re-identification model to obtain target features.

[0040] The authorized image corresponding to the target user includes the target user, and this image can be obtained through various means. For example, a camera can be used to capture the number-taking machine and the user currently taking a number. When the number-taking machine generates a queue number, the current frame of the video stream captured by the camera is used as the authorized image corresponding to the target user. For example, a camera can be installed on the number-taking machine. When the number-taking machine generates a queue number, the camera takes a photo of the target user, and the resulting photo is used as the authorized image corresponding to the target user.

[0041] It should be noted that the target user's authorization can be obtained by asking the target user before obtaining the number.

[0042] Among them, pedestrian characteristics may include but are not limited to pedestrian appearance characteristics (such as height, body shape, clothing, hairstyle, etc.), facial features (such as nose, eyes, etc.), gait characteristics, and body part characteristics (such as head, upper body, lower body, shoes, etc.).

[0043] The person re-identification model may be any model for extracting pedestrian features of a person and performing person re-identification based on the extracted pedestrian features, and this embodiment does not limit this. For example, the person re-identification model may be a model based on an attention mechanism, a model based on a convolutional neural network, or a model based on metric learning.

[0044] In this embodiment, the target user goes to the location of the number-taking machine to take a number. The number-taking machine responds to the target user's number-taking operation and generates a queue number for the target user.

[0045] In this step, when the number-taking machine generates a queue number, it obtains the authorized image corresponding to the target user, obtains the image area where the target user is located from the authorized image corresponding to the target user, and uses the image area where the target user is located as the target image; the target image is input into the pedestrian re-identification model, and the pedestrian re-identification model processes the target image and extracts the pedestrian features of the target user in the target image.

[0046] Acquiring the image region where the target user is located from the authorized image corresponding to the target user can be accomplished in a variety of ways. For example, a target detection model can be used to detect the target user and output the image region where the target user is located. For example, if the authorized image corresponding to the target user is a video frame image determined from a video stream captured by a camera at a fixed shooting position, the image region of the number-taking machine in the video frame image is fixed, and the target user is located at the location of the number-taking machine when the machine generates the queue number. Therefore, the image region where the number-taking machine is located in the video frame image can be used as the image region where the target user is located.

[0047] S102. When it is determined that the expected waiting time corresponding to the target user is less than or equal to a preset threshold, based on the pedestrian re-identification model and according to the target characteristics, determine whether the target user is in the target area; wherein the target area includes the business processing area and the waiting area.

[0048] The expected waiting time refers to the time obtained by predicting the target user's queue waiting time according to the target user's queue number (i.e., queue order) after the target user takes a number. It should be noted that any method for predicting the target user's queue waiting time can be used to predict the expected waiting time, and this embodiment does not limit this. For example, any time prediction model for predicting the queue waiting time according to the user's queue order can be used to predict the expected waiting time, such as a time prediction model based on machine learning, a time prediction model based on a support vector machine, etc. For example, a statistical method can also be used to predict the expected waiting time. Specifically, the queue number is used as an independent variable and the expected waiting time is used as a dependent variable to establish a linear relationship between the queue number and the expected waiting time, so that the expected waiting time can be predicted according to the queue number.

[0049] In this embodiment, at least one shooting device (such as a high-definition camera) is used to monitor the target area. The business processing area refers to the area where the target user is located when processing business, and the waiting area refers to the area where the target user queues before processing business.

[0050] In this step, when it is determined that the expected waiting time corresponding to the target user is less than or equal to the preset threshold, it means that it is about to be or has been the target user's turn to handle the business, then the current video frame image is obtained from the video stream captured by the shooting device, and the pedestrian re-identification model is used to extract the pedestrian features of each person contained in the current video frame image, and the similarity between each person's pedestrian features and the target features is calculated to obtain multiple similarities.

[0051] If the maximum similarity among multiple similarities is greater than or equal to the similarity threshold, the pedestrian corresponding to the maximum similarity is determined to be the target user, and the target user is determined to be in the target area. If the maximum similarity among multiple similarities is less than the similarity threshold, the target user is determined not to be in the target area.

[0052] The preset threshold and the similarity threshold are both real numbers greater than or equal to 0, and can be set according to actual application needs, which is not limited in this embodiment. It should be noted that the pedestrian re-identification model in this step is the same as the pedestrian re-identification model in the aforementioned S101.

[0053] S103: If the target user is not in the target area, the queuing number of the target user is postponed, and a prompt message is sent to the target user; wherein the prompt message is used to prompt the target user to return to the target area.

[0054] In this step, if it is determined that the target user is not in the target area, it means that the target user cannot go to the business processing area in time to handle the business. The target user's queue number will be postponed and a prompt message will be sent to the target user to prompt the target user to return to the target area to handle the business.

[0055] In this embodiment, the queue numbers of the target users are arranged in the order of number taking. In this step, delaying the queue number of the target user means delaying the position of the queue number in the queue number sequence.

[0056] The number of digits by which the target user's queue number is postponed can be set according to actual application needs, and this embodiment does not limit this. For example, the target user's queue number is postponed by three digits, that is, the position of the target user's queue number in the queue number sequence is postponed by three digits.

[0057] It should be noted that after the target user's queue number is postponed, the estimated waiting time corresponding to the target user can be re-predicted based on the postponed queue number to obtain a new estimated waiting time, and the new estimated waiting time is added to the prompt information and sent to the target user, so that the target user can reasonably arrange the itinerary according to the estimated waiting time, thereby optimizing the user experience.

[0058] Optionally, if the preset threshold is set to 0, when it is determined that the expected waiting time corresponding to the target user is equal to the preset threshold, it means that it is the target user's turn to handle the business. If it is determined that the target user is in the target area, the target detection model is used to detect whether the target user is in the business handling area.

[0059] Specifically, the target detection model is used to output the image area of ​​the target user in the current video frame image determined by the pedestrian re-identification model. If the image area of ​​the target user in the current video frame image coincides with the image area where the business processing area is located, it means that the target user is in the business processing area; if the image area of ​​the target user in the current video frame image does not coincide with the image area where the business processing area is located, it means that the target user is not in the business processing area.

[0060] The target detection model may be any model for detecting pedestrians and outputting the image region where the pedestrians are located, and this embodiment does not limit this. For example, the target detection model may be a Transformer-based target detection model, a Faster R-CNN (Faster Region-based Convolutional Neural Network), or the like. Optionally, the target detection model used here may be the same as the target detection model used in S402 below.

[0061] If the target user is not in the business processing area, the target user's location and notification information are output; wherein the notification information is used to notify the staff to guide the target user to the business processing area.

[0062] Specifically, if the target user is not in the business processing area, it means that the target user is in the waiting area and has not gone to the business processing area in time to handle the business. The target user's location and notification message will be output to notify the staff to go to the target user's location and guide the target user to the business processing area to handle the business.

[0063] In this embodiment, when it is determined that it is the target user's turn to handle business, if it is determined that the target user is in the target area but not in the business handling area, it means that the target user is stranded. The target user's location and notification information are output to notify the staff to go to the target user's location and guide the target user to the business handling area to handle business, so as to avoid the target user being stranded, thereby saving time and improving the overall efficiency of business processing.

[0064] In an embodiment of the present application, when the number-taking machine generates a queue number for a target user to be processed, the method extracts the target user's pedestrian features based on the authorized image corresponding to the target user and the pedestrian re-identification model to obtain the target features; when it is determined that the expected waiting time corresponding to the target user is less than or equal to the preset threshold, the method determines whether the target user is in the target area based on the pedestrian re-identification model and the target features; wherein the target area includes the business processing area and the waiting area; if the target user is not in the target area, the queue number of the target user is postponed, and a prompt message is sent to the target user; wherein the prompt message is used to prompt the target user to return to the target area. In this method, by postponing the queue number of the target user who is not in the target area, continuous waiting for the target user is avoided, time is saved, and business processing efficiency is improved. In addition, the user is reminded and does not need to re-take a number, which optimizes the user experience.

[0065] Optionally, the queuing number calling method can also be integrated with number-taking devices other than traditional number-taking machines, such as tablets, mobile terminals, etc., to reduce the hardware installation and setting costs such as traditional number-taking machines.

[0066] Figure 2 A flow chart of a method for extracting target features provided in an embodiment of the present application. In an optional implementation, in the aforementioned S101, when the number-picking machine generates a queue number for the target user to be processed, the pedestrian features of the target user are extracted based on the authorized image corresponding to the target user and the pedestrian re-identification model to obtain the target features, such as Figure 2 As shown, this can be achieved by taking the following steps:

[0067] S201. Obtain number-taking information sent by a number-taking machine located in a target area; wherein the number-taking information is used to indicate that the target user has successfully taken a number, and the number-taking information includes a queue number.

[0068] The number-taking information may include but is not limited to the queue number generated by the number-taking machine for the target user, and a sign or symbol indicating that the target user has successfully taken the number, to indicate that the target user's indication is successful.

[0069] In this step, when a target user successfully obtains a number from a number-picking machine located in the target area, the number-picking information is transmitted to the electronic device used to implement the anomaly detection method. At this point, the target user's location coincides with the number-picking machine's location. In subsequent steps, the target user who successfully obtained a number can be determined based on the corresponding video frame image and the number-picking machine's location.

[0070] S202. When the number-taking information is obtained, a target image corresponding to the target user is determined from a video stream captured by a camera; wherein the camera is used to capture the number-taking machine and the user currently taking the number.

[0071] In this step, when the number-taking information is obtained, it means that the target user has successfully taken a number and is currently located at the location of the number-taking machine. At this time, the current video frame image can be obtained from the video stream captured by the shooting device, and the image area where the target user is located in the current video frame image can be used as the target image.

[0072] S203: After obtaining authorization from the target user, extract the target user's pedestrian features based on the target image and the person re-identification model to obtain target features.

[0073] Specifically, the target image is input into the pedestrian re-identification model, processed by the pedestrian re-identification model, and the pedestrian features of the target user are obtained, and the pedestrian features of the target user are used as the target features.

[0074] Optionally, the target user's queue number can be associated with the target feature, or a unique number or identifier can be assigned to the target feature, so that when the target user is subsequently queried in the target area based on the target feature, the target feature corresponding to the target user can be queried based on the queue number or the unique number or identifier.

[0075] The person re-identification model may be any model for extracting pedestrian features and performing person re-identification based on the extracted pedestrian features, and this embodiment does not limit this. For example, the person re-identification model may be a model based on an attention mechanism, a model based on a convolutional neural network, or a model based on metric learning.

[0076] Optionally, the person re-identification model includes L transformer modules stacked sequentially.

[0077] In this step, based on the target image and the person re-identification model, the pedestrian features of the target user are extracted. When the target features are obtained, the following methods can be used to implement them:

[0078] The target image is subjected to L-1 reduction operations in sequence to obtain L-1 first images of successively reduced sizes; and the target image is subjected to L-2 reduction operations in sequence to obtain L-2 second images of successively reduced sizes; wherein the L-1 first images are arranged from large to small according to image size, and the L-2 second images are arranged from large to small according to image size, and the size of the L-1th first image is the same as the size of the L-2th second image.

[0079] Wherein, L is a positive integer greater than or equal to 3, which can be set according to the needs of actual application and is not limited in this embodiment.

[0080] Specifically, the target image is scaled down once to obtain a first image of a first size. The first image is scaled down once to obtain a second image of a second size. This is repeated until the L-1th first image of the L-1th size is obtained. The degree or multiple of reduction in each scale-down operation can be the same or different. The first size is larger than the second size, the second size is larger than the third size, and so on, until the L-2th size is larger than the L-1th size. Thus, L-1 first images of successively smaller sizes are obtained.

[0081] The target image is scaled down once to obtain a first second image of a second size; the target image is scaled down once to obtain a second second image of a third size; and so on, the target image is scaled down once to obtain an L-2th second image of an L-1th size. Thus, L-2 second images of successively smaller sizes are obtained.

[0082] For the first transformer module, the target image is input into the first transformer module to obtain the output of the first transformer module.

[0083] Specifically, the target image is input into the first transformer module, processed by the first transformer module, and the output of the first transformer module is obtained.

[0084] Optionally, before inputting the target image into the first converter module, the target image can be processed using a convolution module (such as adjusting the image size, dimension, etc.) so that the processed target image can be better processed by the first converter block; then the output of the convolution module (i.e., the processed target image) is input into the first converter module for processing to obtain the output of the first converter module.

[0085] For the second transformer module, the output of the first transformer module and the first first image are spliced ​​together and then input into the second transformer module to obtain the output of the second transformer module.

[0086] The output of the first transformer module and the first first image may be spliced ​​using any splicing operation, which is not limited in this embodiment. For example, the output of the first transformer module (which is an image) and the first first image may be spliced ​​together according to the channel dimension of the image to obtain a new image; or the output of the first transformer module and the first first image may be spliced ​​together according to the spatial dimension (such as width or height) to obtain a new image.

[0087] Optionally, before splicing the output of the first transformer module and the first first image, the output of the first transformer module and the first first image are respectively input into the convolution module for processing so that the output of the processed first transformer module and the processed first first image have the same size; then the output of the processed first transformer module and the processed first first image are added element by element or weighted added to obtain a new image.

[0088] In this step, the new image is input into the second transformer module for processing to obtain the output of the second transformer module.

[0089] For the mth transformer module among the third to Lth transformer modules, the output of the previous transformer module, the m-1th first image and the m-2th second image are spliced ​​together and input into the mth transformer module to obtain the output of the mth transformer module; and the output of the Lth transformer module is used as the target feature.

[0090] Here, m is a positive integer greater than or equal to 3 and less than or equal to L.

[0091] Any splicing operation may be used to splice the output of the previous converter module, the (m-1)th first image, and the (m-2)th second image, which is not limited in this embodiment.

[0092] For example, the output of the previous converter module, the m-1th first image, and the m-2th second image can be spliced ​​together according to the channel dimension to obtain a new image; or the output of the previous converter module, the m-1th first image, and the m-2th second image can be spliced ​​together according to the spatial dimension (such as width or height) to obtain a new image.

[0093] Optionally, before splicing the output of the previous converter module, the m-1th first image and the m-2th second image, the output of the previous converter module, the m-1th first image and the m-2th second image are respectively input into the convolution module for processing so that the processed output of the previous converter module, the processed m-1th first image and the processed m-2th second image have the same size; then the processed output of the previous converter module, the processed m-1th first image and the processed m-2th second image are added element by element or weighted added to obtain a new image.

[0094] It should be noted that each converter module from the first converter module to the Lth converter module can use different module parameters (such as the number of self-attention heads, hidden layer parameters, etc.), each convolution module can use different module parameters (such as convolution kernel size, number of convolution kernels, etc.), and each splicing operation can use different splicing operations.

[0095] In this step, the new image is input into the mth transformer module for processing to obtain the output of the mth transformer module. m is incremented from 3 to L, and so on. When m is L, the output of the L-1th transformer module, the L-1th first image, and the L-2th second image are concatenated and input into the Lth transformer module to obtain the output of the Lth transformer module, which is the pedestrian feature of the target user, also known as the target feature.

[0096] For example, Figure 3 This is a schematic diagram of the structure of a pedestrian re-identification model provided in an embodiment of the present application. Figure 3 As shown, the pedestrian re-identification model includes four transformer modules (including the first transformer module to the fourth transformer module), nine convolution modules (including the first convolution module to the ninth convolution module) and three splicing modules (including the first splicing module to the third splicing module).

[0097] Among them, the transformer module is used to perform feature extraction processing on the input image, the convolution module is used to perform size processing (enlarging or reducing the size) or dimension processing (increasing or reducing the dimension) on the input image, and the stitching module is used to stitch multiple input images.

[0098] It should be noted that the converter modules from the first converter module to the fourth converter module can use different module parameters (such as the number of self-attention heads, hidden layer parameters, etc.), the first convolution module to the ninth convolution module can use different module parameters (such as convolution kernel size, number of convolution kernels, etc.), and the first splicing module to the third splicing module can use different splicing operations.

[0099] Specifically, the target image is processed by the first convolution module to obtain a processed target image; the processed target image is input into the first transformer module to obtain the output of the first transformer module.

[0100] The output of the first transformer module is processed by the second convolution module to obtain the processed output of the first transformer module; the target image is reduced to obtain a first first image of a first size, and the first first image is processed by the third convolution module to obtain a processed first first image; the processed output of the first transformer module and the processed first first image are spliced ​​by the first splicing module to obtain a first new image.

[0101] The first new image is input into the second transformer module to obtain the output of the second transformer module, and the output of the second transformer module is processed by the fourth convolution module to obtain the processed output of the second transformer module; the first first image is reduced to obtain a second first image of the second size, and the second first image is processed by the fifth convolution module to obtain the processed second first image; the target image is processed by the sixth convolution module to obtain the processed first second image; the processed output of the second transformer module, the processed second first image and the processed first second image are spliced ​​by the second splicing module to obtain a second new image.

[0102] The second new image is input into the third transformer module to obtain the output of the third transformer module, and the output of the third transformer module is processed by the seventh convolution module to obtain the processed output of the third transformer module; the second first image is reduced to obtain a third first image of the third size, and the third first image is processed by the eighth convolution module to obtain the processed third first image; the target image is reduced to obtain a second second image of the third size, and the second second image is processed by the ninth convolution module to obtain the processed second second image; the processed output of the third transformer module, the processed third first image, and the processed second second image are spliced ​​by the third splicing module to obtain the third new image.

[0103] The third new image is input into the fourth transformer module to obtain the output of the fourth transformer module, that is, the pedestrian features of the target user corresponding to the target image, that is, the target features.

[0104] It should be noted that the person re-identification model can be trained in the following way:

[0105] From the video stream captured by the camera, frames are extracted at a fixed length (e.g., one video frame image is extracted every ten frames) to obtain multiple video frames. For any of the multiple video frames obtained, the position of a specific pedestrian (which can be selected arbitrarily) in the video frame image is annotated in the image to obtain a dataset annotated with the specific pedestrian. Repeating this annotation process to annotate another specific pedestrian will obtain a dataset annotated with another specific pedestrian. In practical applications, the above annotation process is repeated as many times as the number of datasets for specific pedestrians required.

[0106] In a training of the pedestrian re-identification model, for any specific pedestrian, a first specific video frame image is arbitrarily selected, and based on the position of the specific pedestrian in the first specific video frame image, the pedestrian re-identification model is used to extract the pedestrian features of the specific pedestrian in the first specific video frame image (called the first feature). Then, a second specific video frame image is arbitrarily selected, and the pedestrian features of each pedestrian in the second specific video frame image are extracted using the pedestrian re-identification model (called the second feature). The training goal of the pedestrian re-identification model is to make the similarity between the pedestrian features corresponding to the specific pedestrian in the second feature and the first feature greater, and the similarity between the pedestrian features that do not correspond to the specific pedestrian in the second feature (i.e., the pedestrian features of non-specific pedestrians) and the first feature smaller. Therefore, based on the first feature and the second feature, the weighted sum of the contrast loss, triplet loss, improved triplet loss and quadruple loss can be calculated, and the model parameters of the pedestrian re-identification model can be back-propagated and optimized. For any pedestrian with any characteristics, the above single training process can be repeated; for pedestrians with different characteristics, the above single training process can be repeated; until the weighted sum of the contrast loss, triplet loss, improved triplet loss and quadruple loss no longer decreases, the training can be stopped to obtain the trained pedestrian re-identification model.

[0107] In this embodiment, the target image that is successively reduced and the target image that is directly reduced are spliced ​​and then input into the converter module for processing, thereby improving the accuracy of extracting pedestrian features of the target user and obtaining target features.

[0108] In this embodiment, by obtaining the current video frame image when the target user successfully takes a number, and using the image area of ​​the user where the current video frame image overlaps with the image area where the number-taking machine is located as the target image corresponding to the target user, the target user and the pedestrian image of the target user are accurately determined; then the pedestrian features of the target user (i.e., target features) are extracted through the pedestrian re-identification model, thereby improving the accuracy of the target features corresponding to the target user.

[0109] Figure 4A flow chart of a method for determining a target image provided by an embodiment of the present application. In an optional implementation, in the aforementioned S202, a target image corresponding to a target user is determined from a video stream captured by a shooting device, such as Figure 4 As shown, this can be achieved by taking the following steps:

[0110] S401: Determine, from a video stream captured by a capturing device, a corresponding video frame image when a target user successfully obtains a number.

[0111] When the target user successfully obtains a number, the target user is located at the location of the number-taking machine. In subsequent steps, a pedestrian image of the target user, ie, a target image, can be determined from the video frame image corresponding to the target user successfully obtaining a number.

[0112] S402: Determine a target image from the video frame image based on a target detection model.

[0113] The target detection model may be any model for detecting pedestrians and outputting the image region where the pedestrians are located, and this embodiment does not limit this. For example, the target detection model may be a Transformer-based target detection model, a Faster Region-based Convolutional Neural Network (Faster R-CNN), or the like.

[0114] In this step, the video frame image corresponding to the target user successfully taking the number is input into the target detection model. The target detection model detects pedestrians in the video frame image, obtains the bounding box of each pedestrian in the video frame image, and uses the image area corresponding to the bounding box of the pedestrian that overlaps with the image area where the number-taking machine is located as the target image.

[0115] Optionally, the target detection model includes N transformer encoding modules, N transformer decoding modules, and a classification layer and a boundary regression layer stacked in sequence.

[0116] This step can be implemented in the following ways:

[0117] The video frame image corresponding to the target user successfully obtaining a number is input into the target detection model, and is processed in sequence by N transformer encoding modules and N transformer decoding modules to obtain the output of the Nth transformer decoding module.

[0118] Wherein, N is a positive integer greater than 0, which can be set according to the needs of actual application and is not limited in this embodiment. For example, N can be set to 4, 6 or 8.

[0119] Specifically, the video frame image corresponding to the target user successfully obtaining the number is input into the target detection model, and is first processed by N converter encoding modules in sequence (extracting the features of the video frame image) to obtain the output of the Nth converter encoding module; then the output of the Nth converter encoding module is input into the converter decoding module, and is processed by N converter decoding modules in sequence (generating candidate areas) to obtain the output of the Nth converter decoding module.

[0120] It should be noted that the output of the Nth transformer decoding module includes multiple candidate regions, wherein the multiple candidate regions refer to different image regions in the video frame image corresponding to when the target user successfully obtains a number.

[0121] The output of the Nth transformer decoding module is input into the classification layer for confidence calculation, and the confidence of each user in the image corresponding to the target user is output; the confidence is used to measure the possibility that a user is human.

[0122] Specifically, for any candidate region, the classification layer calculates its confidence. The greater the confidence of the candidate region, the more likely it is to contain a person. This allows us to obtain the confidence of the candidate region corresponding to each user in the image corresponding to the target user, i.e., the confidence of each user.

[0123] The confidence of the user who overlaps with the area where the number-picking machine is located is determined from the confidence of each user. If the determined confidence exceeds the confidence threshold, the output of the Nth transformer decoding module is input into the boundary regression layer for processing, and the image area where the user who overlaps with the area where the number-picking machine is located is output to obtain the target image.

[0124] The confidence threshold can be set according to actual application needs, and this embodiment does not limit this.

[0125] Specifically, the user who overlaps with the area where the number-picking machine is located is the target user. If the confidence corresponding to the target user exceeds the confidence threshold, it means that the target user is a human. The image area where the user who overlaps with the area where the number-picking machine is located can be determined as the pedestrian image of the target user, and the pedestrian image of the target user is used as the target image.

[0126] It should be noted that when outputting the image area of ​​the user that coincides with the area where the number-taking machine is located, the image area can be represented by its coordinates within the entire video frame. The specific coordinate representation method can be set according to actual application needs and is not limited in this embodiment. For example, the image area can be represented by the coordinates of the lower left corner and upper right corner of the image area, or by the coordinates of the upper left corner and lower right corner of the image area, or by the center coordinates of the image area and its offset on the x-axis and y-axis.

[0127] For example, Figure 5 This is a schematic diagram of the structure of a target detection model provided in an embodiment of the present application. Figure 5 As shown in Figure 1, the target detection model includes four transformer encoding modules, four transformer decoding modules, a classification layer and a boundary regression layer.

[0128] Among them, the transformer encoding module is used to extract the features of the video frame image, the transformer decoding module is used to generate candidate regions based on the features of the video frame image, the classification layer is used to calculate the confidence of the candidate region, and the boundary regression layer is used to output the bounding box of the candidate region when the confidence of the candidate region exceeds the confidence threshold.

[0129] Specifically, the video frame image is input into the target detection model, where it is processed sequentially by four transformer encoding modules to extract the features of the video frame image. The features of the video frame image are then input into the transformer decoding module, where they are processed sequentially by four transformer decoding modules to generate multiple candidate regions. These candidate regions are then input into the classification layer, where the confidence score of each candidate region is calculated. For candidate regions that overlap with the area where the ticket machine is located, if the confidence score of the candidate region exceeds the confidence threshold, the candidate region is input into the boundary regression layer to obtain the bounding box corresponding to the candidate region, i.e., the bounding box of the target image. This process results in the target image.

[0130] It should be noted that the target detection model can be trained using images with bounding box annotation information, wherein the images with bounding box annotation information can be obtained by extracting video frame images of a specific length from the video stream captured by the shooting device and annotating the video frame images.

[0131] In this embodiment, the target detection model uses N transformer encoding modules and N transformer decoding modules to extract features from video frame images and generate candidate regions, combining the transformer network structure (with global modeling capabilities) and the encoder-decoder structure (with local feature recovery capabilities) to improve detection accuracy and speed.

[0132] In this embodiment, compared with determining the target image by intercepting the image area of ​​the area where the number-taking machine is located in the video frame image, the target detection model is used to improve the accuracy of determining the pedestrian image corresponding to the target user.

[0133] In an optional implementation, after the number-picking machine generates a queue number for the target user to be processed, or after the queue number of the target user is postponed if the target user is not in the target area, the expected waiting time of the target user can be predicted. Specifically, this can be achieved in the following manner:

[0134] Based on the time prediction model, the target user's expected waiting time is predicted according to the target user's queue number and the pending business information of each user before the queue number.

[0135] The time prediction model may be any model used to predict the expected waiting time for a user corresponding to a queue number based on the queue number and the pending services of each user preceding the queue number, and this embodiment is not limited thereto. For example, the time prediction model may be a Transformer-based prediction network, a linear regression model, a decision tree regression model, or the like.

[0136] Specifically, for any user before the queue number in the queue, based on the time prediction model, the time required to handle the business to be handled by the user is predicted according to the business information to be handled by the user; the sum of the time required to handle the business corresponding to each user before the queue number is used as the estimated waiting time of the target user.

[0137] Optionally, the time prediction model includes a time series processing module, a static information processing module and a time prediction module.

[0138] Based on the time prediction model, the target user's expected waiting time is predicted according to their queue number and the pending services of the users before their queue number. This can be achieved in the following ways:

[0139] A business data sequence is constructed based on time-related data in the business information; wherein the business data sequence is used to indicate the dynamic relationship of business information changing over time.

[0140] Service information may include, but is not limited to, the current queue user traffic, the processing speed of service providers serving queue users, and the types of services awaited by each queue user. Time-related data in service information refers to data that changes over time within a specific time period, such as the current queue user traffic and the types of services awaited by each queue user.

[0141] Specifically, a specific time period (which can be set as needed) is divided into multiple sub-time periods. For any sub-time period, the time-related data in the business information within the sub-time period is used as the data corresponding to the sub-time period. The multiple data corresponding to the sub-time periods are arranged according to the starting time of the sub-time period to obtain a business data sequence.

[0142] It should be noted that, in the business data sequence, one sub-time period corresponds to one time step, and the time information of the time step is the start time of the corresponding sub-time period. One time step contains the data corresponding to the corresponding sub-time period.

[0143] For example, in one time step, there are three users in the queue, and the types of services to be processed, in order of queue, are deposits, withdrawals, and loans. In the next time step, there are two users in the queue, and the types of services to be processed, in order of queue, are deposits and loans. Based on these two time steps, a service data sequence consisting of two time steps is formed.

[0144] The business data sequence is input into the time series processing module to obtain the dynamic characteristics of the business information, and the time-independent data in the business information is input into the static information processing module to obtain the static characteristics of the business information.

[0145] Among them, the timing processing module may include one or more arbitrary network structures for processing time series, and this embodiment does not limit this. For example, the timing processing module may include a long short-term memory network (Long Short-Term Memory, referred to as LSTM), a recurrent neural network, etc. Optionally, the timing processing model may also include a normalization processing module to normalize the extracted dynamic features, so that the time required for each queue user before the queue number to handle the business can be better predicted based on the normalized dynamic features, thereby more accurately predicting the expected waiting time of the target user. Optionally, the business data sequence can be standardized before entering the business data sequence into the timing processing module.

[0146] The time-independent data in the business information refers to data that does not change with time within a specific time period, such as the areas of expertise of personnel providing business processing services to queuing users.

[0147] The static information processing module may include one or more network structures for extracting features from structured data, which is not limited in this embodiment. For example, the static information processing module may include multiple residual processing network structures, multi-layer perceptron structures, etc.

[0148] Optionally, before inputting the time-independent data in the business information into the static information processing module, the time-independent data in the business information may be standardized first.

[0149] According to the dynamic features and static features, based on the time prediction module, the service processing time of each user before the queue number is predicted in turn, and the sum of the service processing time of each user is used as the expected waiting time.

[0150] Among them, the time prediction module includes multiple dense blocks (i.e., the specific number can be set as needed), a fully connected layer and a softmax layer.

[0151] Specifically, dynamic and static features are fused to generate a fused feature. This fused feature is then fed into the time prediction module to determine the service processing time of the last queued user in the sub-time period corresponding to the last time step in the service data sequence (i.e., the last queued user in a specific time period). Repeating this time prediction process yields the service processing time of each user preceding the queue number. The sum of the service processing times for each user is used as the estimated waiting time.

[0152] There are many ways to fuse dynamic features and static features. For example, dynamic features and static features can be spliced ​​together, or dynamic features and static features can be input into the attention module for fusion.

[0153] For example, Figure 6 This is a schematic diagram of the structure of a time prediction model provided in an embodiment of the present application. Figure 6 As shown, the time prediction model includes a timing processing module, a static information processing module, an attention module and a time prediction module.

[0154] The time series processing module includes a stacked long short-term memory network module, a normalization module, and a transformer module. The business data sequence is input into the time series processing module and processed sequentially by the long short-term memory network module, the normalization module, and the transformer module to obtain dynamic features. The static information processing module includes three stacked residual modules. Time-independent data from the business information is input into the static information processing module and processed sequentially by the three residual modules to obtain static features. The static and dynamic features are input into the attention module and processed by the attention module to obtain fused features.

[0155] The time prediction module consists of three dense blocks, a fully connected layer, and a softmax layer stacked in sequence. The fused features are input into the time prediction module and processed by the three dense blocks, the fully connected layer, and the softmax layer in sequence to obtain the service processing time of the last queued user in a specific time period.

[0156] Optionally, the time prediction model can be trained using a training set labeled with the user's actual business processing time. Optionally, the loss function used during training can be a root mean square error function used to measure the difference between the predicted business processing time predicted by the time prediction model and the actual business processing time. Optionally, the Adam optimizer can be used to update the model parameters during training to accelerate model convergence. During the training process, the model's hyperparameters, such as hidden layer dimensions, number of attention heads, etc., can be adjusted according to the training results (such as prediction accuracy) to optimize model performance and training results.

[0157] Optionally, if there is a large gap between the predicted business processing time and the actual business processing time during the operation of the queuing system, the time prediction model can be automatically further trained based on the actual business processing time and predicted business processing time data generated by the queuing system when it is working, so as to adjust the model parameters and thus improve the prediction accuracy.

[0158] In this embodiment, the data in the business information are divided into time-related data and time-independent data. The time-related data are constructed into time series data and then input into the time series processing module to capture dynamic features; the time-independent data are input into the static information processing module to capture static features; finally, the dynamic features and the static features are combined to predict the business processing time of the user before the queue number in turn, and the sum of the business processing time of each user before the queue number is used as the expected waiting time of the target user, thereby improving the accuracy of the predicted expected waiting time.

[0159] Output the estimated waiting time for the target user.

[0160] Specifically, the expected waiting time of the target user may be outputted via an output device (such as an electronic screen) provided in the target area.

[0161] In this embodiment, the estimated waiting time of the target user is predicted by a time prediction model, and the estimated waiting time of the target user is output, so that the target user can reasonably arrange his time according to the estimated waiting time, thereby improving the user experience.

[0162] Figure 7 This is a structural diagram of a queuing and calling device provided in an embodiment of the present application. Figure 7 As shown, the queuing number calling device 700 includes: a target feature extraction module 701, a target user query module 702 and a queuing number delay module 703.

[0163] Among them, the target feature extraction module 701 is used to extract the pedestrian features of the target user based on the pedestrian re-identification model according to the authorized image corresponding to the target user when the number-taking machine generates a queue number for the target user to handle the business, and obtain the target feature.

[0164] The target user query module 702 is used to determine whether the target user is in the target area based on the pedestrian re-identification model and the target characteristics when it is determined that the expected waiting time corresponding to the target user is less than or equal to the preset threshold; wherein the target area includes the business processing area and the waiting area.

[0165] The queuing number delay module 703 is used to delay the queuing number of the target user if the target user is not in the target area, and send a prompt message to the target user; wherein the prompt message is used to prompt the target user to return to the target area.

[0166] In an optional implementation, when the number-picking machine generates a queue number for a target user to be processed, pedestrian features of the target user are extracted based on the authorized image corresponding to the target user and the person re-identification model. When the target features are obtained, the target feature extraction module 701 is specifically configured to:

[0167] Obtain the number-taking information sent by the number-taking machine located in the target area; wherein the number-taking information is used to indicate that the target user has successfully taken the number, and the number-taking information includes the queue number; when the number-taking information is obtained, determine the target image corresponding to the target user from the video stream captured by the shooting device; wherein the shooting device is used to shoot the number-taking machine and the user currently taking the number; after obtaining the authorization of the target user, according to the target image, based on the pedestrian re-identification model, the pedestrian features of the target user are extracted to obtain the target features.

[0168] In an optional implementation, when determining a target image corresponding to a target user from a video stream captured by a shooting device, the target feature extraction module 701 is specifically configured to:

[0169] Determine the corresponding video frame image when the target user successfully obtains the number from the video stream; and determine the target image from the video frame image based on the target detection model.

[0170] In an optional implementation, the target detection model includes N transformer encoding modules, N transformer decoding modules, a classification layer, and a boundary regression layer stacked in sequence. When determining a target image from a video frame image based on the target detection model, the target feature extraction module 701 is specifically configured to:

[0171] The video frame image is input into the target detection model, and processed in sequence by N transformer encoding modules and N transformer decoding modules to obtain the output of the Nth transformer decoding module; the output of the Nth transformer decoding module is input into the classification layer for confidence calculation, and the confidence of each user in the image corresponding to the target user is output; wherein, the confidence is used to measure the possibility that a certain user is a human; the confidence of the user overlapping with the area where the number-taking machine is located is determined from the confidence of each user. If the determined confidence exceeds the confidence threshold, the output of the Nth transformer decoding module is input into the boundary regression layer for processing, and the image area where the user overlapping with the area where the number-taking machine is located is output to obtain the target image.

[0172] In an optional implementation, the person re-identification model includes L transformer modules stacked sequentially. Based on the target image and the person re-identification model, the target feature extraction module 701 extracts the target user's pedestrian features. When the target features are obtained, the target feature extraction module 701 is specifically configured to:

[0173] The target image is successively subjected to L-1 reduction operations to obtain L-1 first images of successively reduced sizes; and the target image is respectively subjected to L-2 reduction operations to obtain L-2 second images of successively reduced sizes; wherein, the L-1 first images are arranged from large to small according to image size, and the L-2 second images are arranged from large to small according to image size, and the size of the L-1th first image is the same as the size of the L-2th second image; for the first converter module, the target image is input into the first converter module to obtain the output of the first converter module; for the second converter module, the output of the first converter module and the first first image are concatenated, and then input into the second converter module to obtain the output of the second converter module; for the mth converter module among the third to Lth converter modules, the output of the previous converter module, the m-1th first image and the m-2th second image are concatenated, and then input into the mth converter module to obtain the output of the mth converter module; and the output of the Lth converter module is used as the target feature.

[0174] In an optional implementation, the target user query module 702 is further configured to:

[0175] Based on the time prediction model, the target user's queue number and the pending business information of each user before the queue number are predicted, and the estimated waiting time of the target user is output.

[0176] In an optional implementation, the time prediction model includes a time sequence processing module, a static information processing module, and a time prediction module. Based on the time prediction model, when predicting the target user's expected waiting time according to the target user's pending service information and queue number, the target user query module 702 is specifically configured to:

[0177] Based on the time-related data in the business information, a business data sequence is constructed; wherein the business data sequence is used to indicate the dynamic relationship of the business information changing over time; the business data sequence is input into the timing processing module to obtain the dynamic characteristics of the business information, and the time-independent data in the business information is input into the static information processing module to obtain the static characteristics of the business information; based on the dynamic characteristics and the static characteristics, the business processing time of each user before the queue number is predicted in turn based on the time prediction module, and the sum of the business processing time of each user is used as the expected waiting time.

[0178] In an optional implementation, the queue number delay module 703 is further configured to:

[0179] If the target user is in the target area, the target detection model is used to detect whether the target user is in the business processing area; if the target user is not in the business processing area, the target user's location and notification information are output; among them, the notification information is used to notify the staff to guide the target user to the business processing area.

[0180] Figure 8 This is a schematic diagram of the structure of an ultrasonic flow meter control unit provided in an embodiment of the present application. Figure 8 As shown, the ultrasonic flow meter control unit 800 includes a memory 801 and a processor 802. The memory 801 is used to store a computer program, and the processor 802 implements the method of any of the above embodiments when executing the computer program. A communication link is provided between the memory 801 and the processor 802. For example, the memory 801 and the processor 802 can communicate via a communication bus 803.

[0181] Optionally, the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), or application-specific integrated circuits (ASIC). A general-purpose processor may be a microprocessor or any conventional processor. The steps in the method embodiments disclosed in this application may be directly implemented by a hardware processor or implemented by a combination of hardware and software modules in the processor.

[0182] An embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, a method in any of the above method embodiments is implemented.

[0183] An embodiment of the present application further provides a computer program product, including a computer program, which implements the method in any of the above method embodiments when the computer program is executed by a processor.

[0184] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all optional embodiments, and the actions and modules involved are not necessarily required by this application.

[0185] It should be understood that the above-described device embodiments are merely illustrative, and the device of the present application may also be implemented in other ways. For example, the division of units / modules in the above-described embodiments is merely a logical functional division, and actual implementations may employ other division methods. For example, multiple units, modules, or components may be combined or integrated into another system, or some features may be omitted or not implemented.

[0186] In the above embodiments, the description of each embodiment has its own emphasis. For parts not described in detail in a particular embodiment, please refer to the relevant description of other embodiments. The technical features of the above embodiments can be combined in any way. To keep the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0187] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0188] It should be understood that the present application is not limited to the exact structure described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A queuing method, characterized in that: include: When the number-picking machine generates a queue number for a target user to be processed, the pedestrian features of the target user are extracted based on the authorized image corresponding to the target user and the pedestrian re-identification model to obtain the target features; When it is determined that the expected waiting time corresponding to the target user is less than or equal to a preset threshold, determining whether the target user is in a target area based on the pedestrian re-identification model and the target characteristics; wherein the target area includes a business processing area and a waiting area; If the target user is not in the target area, the queuing number of the target user is postponed and a prompt message is sent to the target user; wherein the prompt message is used to prompt the target user to return to the target area.

2. The method according to claim 1, characterized in that When the number-picking machine generates a queue number for a target user to be processed, pedestrian features of the target user are extracted based on the authorized image corresponding to the target user and the pedestrian re-identification model to obtain target features, including: Obtaining number-taking information sent by a number-taking machine located in the target area; wherein the number-taking information is used to indicate that the target user has successfully taken a number, and the number-taking information includes the queue number; When the number-taking information is obtained, a target image corresponding to the target user is determined from a video stream captured by a camera; wherein the camera is used to capture the number-taking machine and the user currently taking the number; After obtaining authorization from the target user, pedestrian features of the target user are extracted according to the target image and based on the person re-identification model to obtain the target features.

3. The method according to claim 2, characterized in that The determining the target image corresponding to the target user from the video stream captured by the shooting device includes: Determine from the video stream a video frame image corresponding to when the target user successfully obtains a number; Based on the target detection model, the target image is determined from the video frame image.

4. The method according to claim 3, characterized in that The target detection model includes N transformer encoding modules, N transformer decoding modules, a classification layer and a boundary regression layer stacked in sequence; The determining the target image from the video frame image based on the target detection model includes: Inputting the video frame image into the object detection model, processing it sequentially through the N transformer encoding modules and the N transformer decoding modules, and obtaining an output of the Nth transformer decoding module; Inputting the output of the Nth transformer decoding module into the classification layer for confidence calculation, and outputting the confidence of each user in the image corresponding to the target user; wherein the confidence is used to measure the possibility that a user is a human; The confidence of the user who overlaps with the area where the number-taking machine is located is determined from the confidence of each user. If the determined confidence exceeds the confidence threshold, the output of the Nth converter decoding module is input into the boundary regression layer for processing, and the image area where the user who overlaps with the area where the number-taking machine is located is output to obtain the target image.

5. The method according to claim 2, characterized in that The pedestrian re-identification model includes L transformer modules stacked in sequence; The extracting pedestrian features of the target user based on the target image and the person re-identification model to obtain target features includes: Performing L-1 reduction operations on the target image in sequence to obtain L-1 first images of successively reduced sizes; and performing L-2 reduction operations on the target image in sequence to obtain L-2 second images of successively reduced sizes; wherein the L-1 first images are arranged from largest to smallest in terms of image size, and the L-2 second images are arranged from largest to smallest in terms of image size, and the size of the L-1th first image is the same as the size of the L-2th second image; For the first converter module, inputting the target image into the first converter module to obtain an output of the first converter module; For the second converter module, the output of the first converter module and the first first image are spliced ​​together, and the spliced ​​images are input into the second converter module to obtain the output of the second converter module; For the mth converter module among the third to Lth converter modules, the output of the previous converter module, the m-1th first image and the m-2th second image are spliced ​​together and input into the mth converter module to obtain the output of the mth converter module; and the output of the Lth converter module is used as the target feature.

6. The method according to claim 1, characterized in that Also includes: Based on the time prediction model, the target user's expected waiting time is predicted according to the target user's queue number and the pending services of each user before the queue number; The estimated waiting time is output.

7. The method according to claim 6, characterized in that The time prediction model includes a time series processing module, a static information processing module and a time prediction module; The time prediction model is based on the target user's pending service information and queue number, and predicts the target user's expected waiting time, including: Constructing a business data sequence based on the time-related data in the business information; wherein the business data sequence is used to indicate the dynamic relationship of the business information changing over time; Inputting the service data sequence into the time sequence processing module to obtain dynamic features of the service information, and inputting the time-independent data in the service information into the static information processing module to obtain static features of the service information; According to the dynamic features and the static features, based on the time prediction module, the service processing time of each user before the queue number is predicted in sequence, and the sum of the service processing time of each user is used as the expected waiting time.

8. The method according to any one of claims 1 to 7, characterized in that Also includes: If the target user is in the target area, detecting whether the target user is in the business handling area based on the target detection model; If the target user is not in the business processing area, the location and notification information of the target user are output; wherein the notification information is used to notify staff to guide the target user to the business processing area.

9. A queuing number calling device, characterized in that: include: A target feature extraction module is used to extract the pedestrian features of the target user based on the pedestrian re-identification model according to the authorized image corresponding to the target user when the number-taking machine generates a queue number for the target user to be processed, thereby obtaining the target features; a target user query module, configured to determine, when it is determined that the expected waiting time corresponding to the target user is less than or equal to a preset threshold, whether the target user is in a target area based on the person re-identification model and the target characteristics; wherein the target area includes a business processing area and a waiting area; The queuing number postponing module is used to postpone the queuing number of the target user if the target user is not in the target area, and send a prompt message to the target user; wherein the prompt message is used to prompt the target user to return to the target area.

10. An electronic device, characterized in that: include: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 8 when executed by a processor.

12. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 8 when executed by a processor.

Images