Self-service intelligent bookshelf book borrowing and returning system based on intelligent library

Abstract

The invention belongs to the field of book intelligent management research, and provides a self-service intelligent bookshelf book borrowing and returning system based on a smart library, which is characterized in that a sensor is deployed at a bookshelf end to collect books, a background system is deployed at a server end to perform management, input user face information is used as an identity label, and face recognition is performed on a user when books are borrowed and returned; book numbers are obtained through the combination of an RF ID technology and an RGB camera, the weight change of a bookshelf is monitored to establish a borrowing and returning behavior session, and book borrowing and returning are judged by comparing the difference before and after a borrowing and returning behavior set; the book circulation state is updated in real time, a warning threshold value is preset, various warning rules are set, overdue conditions and missed recording conditions are calculated, and warning is sent to a background for processing; and summarizing the reported warning conditions, calculating the accuracy of different rule sets, adjusting rule set parameters to generate a new rule set, distributing version numbers for management, establishing a rollback mechanism, comparing different version rule sets, selecting and issuing.

Description

TECHNICAL FIELD

[0001] The application relates to the field of intelligent book management, and particularly relates to a self-service intelligent bookshelf lending and returning system based on a smart library. BACKGROUND

[0002] At present, the management of a library generally relies on manual management. With an increasing number of users borrowing books, the difficulty of book management is also increased, and a large amount of human resources is consumed.

[0003] In China, application No. CN201910257750.X discloses a full-automatic intelligent book library system and a running method. The configuration method comprises a book library server for storing information and coordinating the working order of each device, a self-service device for verifying identity and book information to realize initial operation of book lending and returning, a conveying device for conveying books between the book library server and the self-service device, and an intelligent bookshelf for storing books and automatically taking and returning books. The book library server coordinates the working order and working state of the conveying device and the intelligent bookshelf according to the data returned by the self-service device, improves the management and service level of the library, and realizes quick self-service lending and returning and quick book inventory and arrangement.

[0004] In the field of intelligent book management, although there is a system that logs in an account through a visual module on a self-service device, instructs an intelligent bookshelf to realize query, book lending and returning and automatic arrangement by a book library server, in the scenario of community shared book service, the book lending and returning process relies more on user self-operation. For lending and returning books in a community library, the overall process is complicated and prone to errors, which affects the reading enthusiasm. Meanwhile, the personnel flow of community book lending and returning is large, which leads to a fuzzy book circulation state and many management blind spots. Therefore, a self-service intelligent bookshelf system capable of automatically identifying user lending and returning behaviors, recording a book circulation state in real time and supporting local lightweight deployment is needed, so as to improve the intelligent level and operation efficiency of community book service. SUMMARY

[0005] The application provides a self-service intelligent bookshelf lending and returning system based on a smart library, which aims to solve the problems of complicated manual lending and returning of books and many book management blind spots caused by large personnel flow in a community.

[0006] The technical scheme adopted by the application to solve the above technical problems is as follows: a self-service intelligent bookshelf lending and returning system based on a smart library is provided, which comprises:

[0007] Step one, a lightweight edge system is deployed, and a user identity is registered and identified.

[0008] Step two, book information is automatically sensed and collected, and book lending and returning behaviors are identified.

[0009] Step three, store and update the book circulation status, set the book alert rule set;

[0010] Step four, regularly analyze abnormal data, synchronize to the background iteration rule set.

[0011] As a preferred embodiment, the specific steps of deploying a lightweight edge system are: deploying a server interworking with the community network at the community property room, building a set of background management system for managing and monitoring the book status on the server; deploying a platform type weighing sensor at each layer of the bookshelf partition, installing an RFID antenna on the top of the bookshelf, using the ISBN as the identification for each book in the background management system, using the RFID reader to read the blank tag, mapping the generated tag sequence with the book number, and importing it into the background management system, and installing an RGB camera on the top of the bookshelf to collect human faces, cropping and converting the collected RGB images to grayscale, inputting them into the feature extraction network, aligning the RGB images to the same dimension according to the face key points, extracting different levels of features in the convolutional layer of the feature extraction network, normalizing each level of feature, splicing the normalized multi-level features, compressing them into a one-dimensional vector, and outputting the face feature vector; each bookshelf realizes secure encryption through the application programming interface key API Key and the server, and communicates the data collected at the bookshelf end with the server through the WebSocket communication protocol.

[0012] As a preferred embodiment, the specific steps of registering user identity and identification are: community users register an account through a code scanning applet, collect and input user face information in front of the bookshelf, obtain the face feature vector through the feature extraction network, use the face feature vector as the user identifier in the background management system, and establish a corresponding user ID, and pack and store it in the server; the RGB camera monitors the user approaching the bookshelf, real-time collects a frame of face image, compares the obtained face feature vector with the stored data in the background management system, and if the comparison is successful, establishes a behavior session, and the bookshelf end and the server record the current session synchronously.

[0013] As a preferred embodiment, the specific steps of automatic sensing and collecting book information are: the RFID antenna continuously emits a radio frequency field, the RFID tag on the book enters the radio frequency field, the RFID tag converts the radio frequency capability into direct current to power the chip in the RFID tag, the chip transmits information by adjusting the reflection intensity of the radio frequency field emitted by the RFID antenna, the RFID reader parses the information returned by the chip, and at the same time, combines with the RGB camera to assist in shooting the book cover, calls the OCR technology to obtain the book ISBN; compares the data collected by the RFID reader combined with the RGB camera with the background management system to obtain the book information.

[0014] As a preferred embodiment, the specific steps of identifying the book lending behavior are: real-time monitoring of the current partition weight change through a platform type weighing sensor, pre-setting a weight change threshold, and when the weight loss or gain exceeds the set threshold, judging the lending behavior, initially determining that the weight loss is a book lending behavior and the weight gain is a book returning behavior, recording each lending behavior as a triple, and the formula is:

[0015] j = (b, w, t),

[0016] where j represents a lending behavior, b represents the identification information of the book, w represents the monitored weight change, and t represents the current timestamp;

[0017] Record multiple lending behaviors as a behavior set, and further determine the lending behavior by comparing the differences between the behavior sets before and after the lending behavior, and the behavior set is:

[0018] J before = {j1, j2, j3...j k-1}, J after = {j1, j2, j3...j k-1 , j k},

[0019] where k represents the index of the lending behavior, j k represents the kth lending behavior, J before represents the behavior set of the k-1th lending behavior, and J after represents the behavior set of the kth lending behavior; by comparing the difference set of the behavior set and the occurrence order of the lending behavior, the book lending behavior is confirmed.

[0020] As a preferred embodiment, the specific steps of storing and updating the book circulation state are: designing a data storage model, for the book entity, using ISBN as the unique identifier, designing the book name and author fields to query the book, designing the circulation state field to view whether the book is lent out or returned, and designing the operation time to record the time of the last operation of the book; for the user entity, using the user ID as the unique identifier, designing the face information to match and verify the user identity, designing the name and contact information fields to query the user, for the lending behavior entity, generating a globally unique ID for each lending to identify, using the book ISBN as the foreign key to associate the behavior, and using the user ID as the foreign key to associate the behavior, designing the behavior type to indicate whether it is lent out or returned, and designing the timestamp field to record the behavior occurrence time; when a lending and returning behavior is identified, a new lending behavior data is created, the circulation state and operation time of the book are updated in real time, and the book ISBN, user ID, behavior type, and current timestamp are recorded.

[0021] As a preferred embodiment, the specific steps of setting the book warning rule set are: setting the maximum book lending time in advance, calculating the lending time after the book is lent out, issuing an overdue warning to the administrator in the background management system for the overdue book, and the book lending overdue judgment formula:

[0022] t now -t borrow >thershold max ,

[0023] Where t now represents the current system time stamp, t borrow represents the book lending time stamp, and thershold max represents the pre-set maximum overdue value;

[0024] The average weight of the book is set in advance, and after receiving the lending behavior in the system, the platform type weighing sensor combined with the partition calculates the number of books to be borrowed, and judges the number of books to be borrowed with the actual recorded lending number. The warning of missing record is issued for the case that the number of books is more than 2, and the lending number judgment formula is:

[0025]

[0026] Where Δw represents the weight change value monitored by the partition, w avg represents the pre-set average weight of a single book, and N act represents the actual detected lending number.

[0027] As a preferred embodiment, the specific steps of periodically analyzing abnormal data are: writing the warning information into the abnormal data log in the background management system, each record containing warning ID, book ID, user ID, warning type, warning time, and monthly statistical warning data, calculating the overdue rate and the missing record rate, and the formula is:

[0028]

[0029] Where R overdue represents the overdue rate, R miss represents the missing record rate, Count all represents the total number of warnings, Count overdue represents the number of overdue type warnings, and Count miss represents the number of missing record type warnings.

[0030] As a preferred embodiment, the specific steps of synchronizing to the background iteration rule set are: assigning a version number to the warning rule set, recording the parameters, time and change description of the rule set in each version, evaluating the accuracy of the rule set through historical data, and the formula for calculating the accuracy is:

[0031]

[0032] wherein R Accuracy represents the accuracy of the current rule set, A total represents the total amount of historical alert data, A right represents the amount of historical alert data correctly identified;

[0033] The background management system modifies the parameters in the rule set, generates a new rule set and version number, selects the rule set by comparing the accuracy of the rule sets before and after the version, establishes a rollback mechanism to roll back the historical version, and then issues it to the bookshelf end.

[0034] The beneficial effects of the present application are:

[0035] 1. By deploying a lightweight edge system, the book information collection and user identity recognition are autonomously completed, the borrowing and returning behavior is intelligently judged, the user operation process is simplified, and the use experience is improved.

[0036] 2. Through the configurable alert rule set, various types of abnormalities are automatically detected, located and reported, the effect of the abnormal evaluation rule is regularly analyzed, so as to reduce the book management and operation and maintenance cost.

[0037] Legend

[0038] Figure 1 It is a module diagram of a self-service intelligent bookshelf borrowing and returning system based on a smart library. DETAILED DESCRIPTION

[0039] In order to make the technical means, creative features, purposes and effects realized by the present application easy to understand, the present application will be further described below in combination with specific embodiments, but the following embodiments are only preferred embodiments of the present application, not all. Based on the embodiments in the embodiments, other embodiments obtained by those skilled in the art without creative labor also belong to the protection scope of the present application.

[0040] Example 1, as Figure 1The application discloses a full-automatic intelligent book library system and a running method thereof, and belongs to the technical field of intelligent book library systems.The full-automatic intelligent book library system comprises the following steps: deploying a lightweight edge system, registering a user identity and identifying the same, automatically sensing and collecting book information, identifying a book lending and returning behavior, storing and updating a book circulation state, setting a book warning rule set, regularly analyzing abnormal data, and synchronizing to a background iteration rule set.The specific implementation steps are as follows: a full-automatic intelligent book library system and a running method thereof, wherein the specific steps of deploying a lightweight edge system are as follows: a server that can interwork with a community network is deployed at a community property room, and a background management system for managing and monitoring a book state is constructed on the server; a platform type load cell is arranged at each layer of a bookshelf partition, when the weight of the bookshelf partition changes, a strain gauge in the load cell produces a slight deformation, changes the resistance value thereof, sends a voltage signal through a bridge circuit and an amplifier, a radio frequency identification (RFID) antenna is arranged at the top of the bookshelf, the international standard book number (ISBN) is taken as an identifier for each book in the background management system, a reader-writer device is used to read an RFID blank tag, a generated tag sequence is mapped with the number of the book, and is imported into the background management system; meanwhile, an RGB camera is arranged at the top of the bookshelf to collect a human face, the collected RGB image is cropped and converted into a grayscale, is input into a feature extraction network, the RGB image is aligned to the same dimension according to human face key points, different level features are extracted from the convolutional layer of the feature extraction network, each level feature is normalized, the normalized multi-level features are spliced and compressed into a one-dimensional vector, and the output is a human face feature vector; each bookshelf is connected with the server through an application programming interface key (API Key) to realize secure encryption, when an API request is initiated, the API Key is attached to a request header to be registered at the server side, the data collected by the bookshelf side is communicated with the server through a WebSocket communication protocol, and the WebSocket performs real-time and low-delay data exchange in a full-duplex mode; the specific steps of registering a user identity and identifying the same are as follows: a community user registers an account through a code scanning applet, human face information of the user is collected and input in front of a bookshelf, a human face feature vector is obtained through the feature extraction network, the human face feature vector is taken as a user identifier in the background management system, a corresponding user ID is established to facilitate operation, the human face feature vector is serialized in a fixed format, and a random and unique key vector is generated, and the two are packaged and stored in the server; when the RGB camera monitors that the user approaches the bookshelf, a human face image is collected in real time, the obtained human face feature vector is compared with the data stored in the background management system, a behavior session is established after a successful comparison, and the bookshelf side and the server synchronously record a current session.

[0041] Based on the above steps, the automatic sensing and collecting book information specific steps are: after the user establishes a session, the RFID antenna continuously emits a radio frequency field, when the RFID tag on the book enters the radio frequency field, the RFID tag converts the radio frequency capability into direct current to power the chip in the RFID tag, the chip transmits information by adjusting the reflection intensity of the radio frequency field emitted by the RFID antenna, the RFID reader parses the information returned by the chip, and at the same time, combines with the RGB camera to assist in shooting the book cover, calls the optical character recognition (Optical Character Recognition, OCR) technology to obtain the book ISBN, and the OCR is to locate the frame of lines, paragraphs and single characters or words in the photographed picture, cut into independent characters, and then extract character features, and form text data after splicing; The data collected by the RFID reader combined with the RGB camera is compared with the background management system to obtain the book information;

[0042] The specific steps of identifying the book lending behavior are: the current partition weight change condition is monitored in real time through the platform type weighing sensor, the weight change threshold is set to 50g in advance, when the weight is reduced or increased by more than the set threshold, the lending behavior is judged, the initial judgment is that the weight reduction is a book lending behavior and the weight increase is a book returning behavior, each lending behavior is recorded as a triple, and the formula is:

[0043] j=(b,w,t),

[0044] Where j represents a lending behavior, b represents the identification information of the book, w represents the monitored weight change, and t represents the current timestamp;

[0045] A plurality of lending behaviors are recorded as a behavior set, and the lending behavior is further determined by comparing the differences between the behavior sets before and after the lending behavior, and the behavior set is:

[0046] J before ={j1,j2,j3...j k-1},J after ={j1,j2,j3...j k-1 ,j k},

[0047] Where k represents the index of the lending behavior, j k represents the kth lending behavior, J before represents the behavior set of the k-1th lending behavior, and J after represents the behavior set of the kth lending behavior; By comparing the difference set of the behavior set and the occurrence order of the lending behavior, it is confirmed whether the book is lent out or returned.

[0048] Based on the above steps, the storage and updating of the book circulation state specific steps are: designing a data storage model, for book entity, taking ISBN as the unique identifier, designing book name, author field to query the book, designing circulation state field to view the book lending or returned, designing operation time to record the latest operation time of the book; for user entity, using user ID as the unique identifier, designing face information to match and verify user identity, designing name, contact information field to query user, for borrowing and returning behavior entity, each borrowing and returning generates a globally unique ID for identification, taking book ISBN as the foreign key for behavior association, and user ID as the foreign key for behavior association, designing behavior type to indicate lending or returning, designing timestamp field to record the behavior occurrence time; when identifying lending and returning behavior, a new borrowing and returning behavior data is created, the circulation state and operation time of the book are updated in real time, the book ISBN, user ID, behavior type and current timestamp are recorded; the specific steps of setting the book warning rule set are: setting the maximum lending time of the book as 30 days, calculating the lending duration after the book is lent out, issuing an overdue warning in the background management system to notify the administrator, and reminding the user in the form of short message, the book lending overdue judgment formula:

[0049] t now -t borrow >thershold max ,

[0050] Where t now represents the current system timestamp, t borrow represents the book lending timestamp, and thershold max represents the pre-set maximum overdue value.

[0051] The average weight of the book is set to 200g, and after receiving the lending behavior in the system, the platform type weighing sensor combined with the partition calculates the number of books to be borrowed, and judges the number of books to be borrowed. The actual recorded lending quantity is determined, and the missing record warning is issued for the case that the number of books is more than 2, and the lending quantity judgment formula is:

[0052]

[0053] Where Δw represents the weight change value monitored by the partition, w avg represents the pre-set average single book weight, and N act represents the actual detected lending quantity.

[0054] Based on the above steps, the specific steps of periodically analyzing abnormal data are: writing the warning information into the abnormal data log in the background management system, each record containing warning ID, book ID, user ID, warning type, specific types of overdue warning and missing warning, warning time, etc., counting the warning data by month, calculating the overdue rate and missing rate, the formula is:

[0055]

[0056] wherein R overdue represents the overdue rate, R miss represents the missing rate, Count all represents the total number of warnings, Count overdue represents the number of overdue type warnings, Count miss represents the number of missing type warnings.

[0057] The overdue rate and missing rate of each book are counted; the specific steps of synchronizing to the background iteration rule set are: assigning a version number to the warning rule set, recording the parameters, time, and change description of the rule set in each version, evaluating the accuracy of the rule set through historical data, and the formula for calculating the accuracy is:

[0058]

[0059] wherein R Accuracy represents the accuracy of the current rule set, A total represents the total amount of historical warning data, A right represents the amount of correctly identified historical warning data.

[0060] The parameters in the rule set are modified in the background management system, a new rule set and version number are generated, the current optimal rule set is selected by comparing the accuracy of the rule sets before and after the version, a rollback mechanism is established to roll back to the historical version, and then it is issued to the bookshelf end.

[0061] Embodiment 2, based on the above embodiment 1, the practical application of the self-service intelligent bookshelf lending and returning system based on the smart library in the community shared book service scenario, specifically the following scheme:

[0062] Step 1: Deploy a Linux server in the community library, install NFC, weight, infrared and other sensors at each bookshelf, and use the Message Queuing Telemetry Transport (MQTT) protocol for communication. Community users register for a community library account by swiping a community card or scanning the QR code of the community mini-program. The backend management system uses the community card ID or the mobile phone number corresponding to the mini-program as the user identifier. Scan the code or swipe the community card at the bookshelf to verify the identity. After obtaining the user identifier, it is compared with the backend database. After a successful comparison, a session token for the current user is returned to record user behavior. For users who have difficulty operating, such as the elderly and children, they can go to the manual service center to register directly with their mobile phone number in the backend management system.

[0063] Step 2: Install an RGB-D camera on the side of each bookshelf to simultaneously capture color and depth images. When a new book is placed on the bookshelf, the camera captures the book spine, normalizes and crops the captured image, and uses a spine recognition convolutional neural network (CNN) to obtain the book's ISBN. The processed image is first used as input. After multiple layers of convolution operations, features are extracted from the spine texture, text stripes, and color from high to low. Pooling is performed between each two layers to reduce the spatial dimension. The output of the last convolution layer is flattened to output a fixed-length feature vector, which corresponds to the book ID. The RGB-D camera is used to recognize actions, and images with relative plane motion relative to the bookshelf are found in consecutive frames. The displacement value between the previous and next frames is calculated. A positive value represents a borrowing action, and a negative value represents a return action. The book status is recorded locally and uploaded to the backend.

[0064] Step 3: For each borrowing and returning behavior, create a behavior record, including the book ISBN, user ID, borrowing and returning type, and timestamp, and establish a view in the background database that summarizes all behavior records; define fuzzy variables for parameters with uncertainty in identification, such as book RFID tags, book weight, and spine images, and establish a mapping of variable intervals from 0 to 1 to form a fuzzy set. The confidence level is set by the variable interval on the fuzzy set, and fuzzy rules are constructed based on the confidence levels of different types of identification. For example, if the RFID tag confidence level is high and the book weight confidence level is also high, the corresponding confidence level of the borrowed quantity is the minimum of the two. For the fuzzy rules of the same book, select the one with the highest confidence level for each behavior, and comprehensively view the situation to determine whether to issue a warning. For example, in the fuzzy rule of book A, the confidence level of one book in the borrowing behavior is the highest, but the actual view records that two books were borrowed. In this case, it is determined that the backend should be informed of the quantity discrepancy.

[0065] Step four, the background management system summarizes the feedback of abnormal data, and confirms the abnormal data according to the actual situation of the feedback, adds a label, and uniformly enters the abnormal data table. The total number of abnormal data and the number of accurate judgments of abnormal data are calculated. The average weight of books and the maximum lending time of books are adjusted. Different warning rule sets are generated. The historical abnormal data table is verified in different fusion rule tables. The warning rule set with the highest accurate probability is selected as the latest version. It is issued to the bookshelf end and continues to count the abnormal data table for iteration.

[0066] The above describes the embodiments of the present application, and in the specific operation, the skilled in the art can make data modifications and mode changes based on this place without departing from the embodiments of the present application and its broader aspects. The appended claims are intended to include all such data modifications and mode changes that do not depart from the embodiments of the present application.

Claims

1. A self-service intelligent bookshelf borrowing and returning book system based on a smart library, characterized by: include: Deploy a lightweight edge system to register and identify users. This involves deploying sensors on the bookshelf to collect books, deploying a backend system on the server to manage them, and using the recorded user facial information as an identity identifier to perform facial recognition on users when borrowing and returning books. Automatically sense and collect book information and identify book borrowing and returning behaviors. This is done by combining RFID technology with an RGB camera to obtain book numbers, monitor weight changes on the bookshelf to establish a borrowing and returning behavior session, and compare the differences before and after the borrowing and returning behavior set to determine the book borrowing and returning behavior. Storing and updating the book circulation status and setting the book warning rule set means designing a database model, specifying the primary key and foreign key associations for each table, updating the book circulation status in real time for each borrowing and returning behavior, pre-setting warning thresholds, setting rules for various types of warnings, calculating overdue and missed records, and sending warnings to the backend for processing; Regularly analyze abnormal data and synchronize it to the background iterative rule set, which is to summarize the reported warning situations, calculate the accuracy of different rule sets, adjust the rule set parameters to generate new rule sets, assign version numbers for management, establish a rollback mechanism, compare the accuracy of different versions of rule sets, and select rule sets to be sent to the bookshelf.

2. The self-service intelligent bookshelf borrowing and returning book system based on the smart library according to claim 1, characterized in that: The specific steps of deploying the lightweight edge system are as follows: deploying a server that is interconnected with the community network in the community property computer room, and building a background management system on the server for managing and monitoring the status of books; deploying a platform weighing sensor at each shelf of the bookshelf, installing a radio frequency identification RFID antenna on the top of the bookshelf, identifying each book by ISBN in the background management system, using a reader / writer device to read the RFID blank tag, mapping the generated tag sequence with the book number, and importing it into the background management system, while installing an RGB camera on the top of the bookshelf to capture faces, cropping and converting the captured RGB image into grayscale, inputting it into a feature extraction network, aligning the RGB image to the same dimension according to the key points of the face, extracting different levels of features in the convolution layer of the feature extraction network, normalizing the features of each level, splicing the normalized multi-level features, compressing them into a one-dimensional vector, and outputting them as a face feature vector; each bookshelf is securely encrypted with the server through the application programming interface key API Key, and the data collected by the bookshelf is communicated with the server through the WebSocket communication protocol.

3. The self-service intelligent bookshelf borrowing and returning book system based on the smart library according to claim 1 is characterized by: The specific steps of registering and identifying the user identity are as follows: the community user registers an account through the QR code scanning applet, collects and enters the user's facial information in front of the bookshelf, obtains the facial feature vector through the feature extraction network, uses the facial feature vector as the user identifier in the background management system, and creates a corresponding user ID, which is packaged and stored in the server; The RGB camera detects the user approaching the bookshelf and captures a facial image in real time. The acquired facial feature vector is traversed and compared with the data stored in the background management system. If the comparison is successful, a behavioral session is established, and the bookshelf and server synchronously record the current session.

4. The self-service intelligent bookshelf borrowing and returning book system based on a smart library according to claim 1, characterized in that: The specific steps of the automatic sensing and collection of book information are as follows: the RFID antenna continuously emits a radio frequency field, the RFID tag on the book enters the radio frequency field, the RFID tag converts the radio frequency capability into direct current to power the chip in the RFID tag, the chip transmits information by adjusting the reflection intensity of the radio frequency field emitted by the RFID antenna, the RFID reader parses the information returned by the receiving chip, and at the same time uses an RGB camera to assist in photographing the book cover, calling OCR technology to obtain the book's ISBN; the data collected by the RFID reader and the RGB camera are compared with the background management system to obtain the book information.

5. The self-service intelligent bookshelf borrowing and returning book system based on the smart library according to claim 1 is characterized by: The specific steps of identifying book borrowing and returning behavior are: The platform-type weighing sensor is used to monitor the weight change of the current partition in real time. The weight change threshold is preset. When the weight decreases or increases beyond the set threshold, the borrowing and returning behavior is judged. The initial judgment is that the weight decrease is a borrowing behavior, and the weight increase is a returning behavior. Each borrowing and returning behavior is recorded as a triplet. The formula is: j=(b,w,t), Where j represents a borrowing and returning behavior, b represents the identification information of the book, w represents the monitored weight change, and t represents the current timestamp.

6. The self-service intelligent bookshelf borrowing and returning book system based on the smart library according to claim 5, characterized in that: The specific steps of identifying book borrowing and returning behavior also include: Multiple borrowing and returning behaviors are recorded as a behavior set. The borrowing and returning behaviors are further determined by comparing the differences in the behavior sets before and after the borrowing and returning behaviors. The behavior set is: J before ={j1,j2,j3...j k-1 },J after ={j1,j2,j3...j k-1 ,j k }, Where k represents the total number of borrowing and returning behaviors, J before represents the set of k-1 borrowing and returning behaviors, J after represents a set of k borrowing and returning behaviors, j1,j2,j3...j k-1 ,j k They represent the first, second, ...kth borrowing and returning behaviors respectively; the book borrowing and returning behaviors are confirmed by comparing the difference set of the behavior set and the order of occurrence of the borrowing and returning behaviors therein.

7. The self-service intelligent bookshelf borrowing and returning book system based on the smart library according to claim 1 is characterized by: The specific steps of storing and updating the book circulation status are as follows: Design a data storage model. For book entities, use ISBN as the unique identifier. Design title and author fields to query books. Design a circulation status field to check whether the book is borrowed or returned. Design an operation time to record the time of book operation. For user entities, the user ID is used as the unique identifier, facial information is designed to match and verify the user's identity, and name and contact information fields are designed to query users. For borrowing and returning behavior entities, a globally unique ID is generated for each borrowing and returning behavior. The book's ISBN is used as a foreign key to associate the behavior with the user ID as a foreign key. The behavior type is designed to indicate whether it is borrowing or returning, and a timestamp field is designed to record the time when the behavior occurred. When a borrowing and returning behavior is identified, a new borrowing and returning behavior data is created, the circulation status and operation time of the book are updated in real time, and the book ISBN, user ID, behavior type and current timestamp are recorded.

8. The self-service intelligent bookshelf borrowing and returning book system based on a smart library according to claim 1, characterized in that: The specific steps for setting the book warning rule set are: Pre-set the book lending time, calculate the lending time after the book is borrowed, and issue an overdue warning in the background management system to notify the administrator for overdue books. The formula for judging overdue book lending is as follows: t now -t borrow >thershold max , where t now Indicates the timestamp of the current system, t borrow Indicates the timestamp of book borrowing, max Indicates the pre-set book loan time; The average weight of books is pre-set. When the system receives a borrowing action, it uses the platform-type weighing sensor of the partition to calculate the number of books that should be borrowed. This is compared with the actual number of books borrowed. If the combined number exceeds two, a missing book warning is issued. The formula for determining the borrowing quantity is: Where Δw represents the weight change value monitored by the partition, w avg Indicates the preset average weight of a single book, N act Indicates the actual detected loan quantity.

9. The self-service intelligent bookshelf borrowing and returning book system based on the smart library according to claim 1 is characterized by: The specific steps of regularly analyzing abnormal data are as follows: In the background management system, the warning information is written into the abnormal data log. Each record contains the warning ID, book ID, user ID, warning type, warning time, and monthly statistics of warning data. The overdue rate and omission rate are calculated using the following formula: where R overdue represents the overdue rate, R miss Indicates the omission rate, Count all Indicates the total number of warnings, Count overdue Indicates the number of overdue warnings, Count miss Indicates the number of alerts of the missed type.

10. The self-service intelligent bookshelf borrowing and returning book system based on the smart library according to claim 1, characterized in that: The specific steps of synchronizing to the background iteration rule set are: Assign version numbers to the alert rule set, record the parameters, time, and change description of the rule set in each version, and evaluate the accuracy of the rule set based on historical data. The formula for calculating the accuracy is: where R Accuracy Indicates the accuracy of the current rule set, A total Indicates the total amount of historical warning data, A right Indicates the amount of historical warning data that is correctly identified; Modify the parameters in the rule set in the background management system to generate a new rule set and version number. Select the rule set by comparing the accuracy of the rule sets of the previous and next versions, and establish a rollback mechanism to roll back the historical version, and then send it to the bookshelf end.

Images