Attendance management method and system based on Internet of Things positioning technology

Abstract

The invention provides an attendance management method and system based on the Internet of Things positioning technology, and relates to the technical field of attendance management, and the method comprises the steps of initialization setting, time judgment, position judgment, attendance machine judgment, invalid output and the like. The system comprises an initialization setting module, a time judgment module, a position judgment module, an attendance machine judgment module and an invalid output module. According to the invention, the workload of medical staff can be reduced, and the attendance success rate is improved.

Description

Technical Field

[0001] The present application relates to the technical field of attendance management, and in particular to an attendance management method and system based on Internet of Things positioning technology. Background Art

[0002] Currently, most hospitals follow a strict attendance system that requires medical staff to clock in four times a day, corresponding to the morning shift and the afternoon shift. However, the nature of medical staff's work determines that they have to face a high-intensity and high-pressure work environment on a daily basis. They are often overwhelmed by emergency treatment, patient consultation, meeting participation and other emergencies, making it difficult for them to clock in on time.

[0003] Traditional attendance methods, especially fingerprint punching and face recognition punching, face many challenges and inconveniences in practical applications. These challenges are not only related to the efficiency and accuracy of attendance management, but also directly affect the work experience of medical staff and the overall operational efficiency of the hospital. This situation not only frequently leads to missed punching, but also often occurs when the punching time exceeds the prescribed range, which in turn has an adverse impact on the performance evaluation of medical staff and dampens their work enthusiasm.

[0004] In view of the above problems, it is urgent to introduce an efficient, convenient and imperceptible attendance management method. Summary of the invention

[0005] In order to reduce the workload of medical staff and improve the success rate of attendance, the present application provides an attendance management method and system based on Internet of Things positioning technology.

[0006] In the first aspect, the present application provides an attendance management method based on Internet of Things positioning technology, which adopts the following technical solutions: An attendance management method based on Internet of Things positioning technology, comprising: Initialization settings: including setting area and setting information; Set area: Set the check-in area; Setting information: Setting personnel information, punch-in time requirement information and punch-in device information; Time and location judgment: Based on the IoT positioning technology, judge whether the current time and location meet the expectations. If so, record the clock-in time and clock-in location; if not, execute the steps judged by the attendance machine; Attendance machine judgment: judge whether the current person has completed the clocking in at the attendance machine. If yes, record the clocking in time and location; if no, execute the invalid output step; Invalid output: Send an invalid punch-in signal.

[0007] By adopting the above technical solution, the initialization setting step (including setting the area, personnel information, punching time requirements, and punching device information) provides the system with a high degree of flexibility and configurability, so that the system can adapt to the attendance management needs of different hospitals or departments. Whether it is geographical location, personnel changes or working hours adjustment, it can adapt to its changes through simple configuration. The time and location judgment step determines whether the punching time and punching location are in line with expectations, that is, the punching behavior is only considered during the set working and off-duty time, which effectively reduces the mistaken punching or invalid punching during non-working hours, improves the accuracy of attendance data, and increases the rigor of punching based on the location judgment of the Internet of Things positioning technology. The time and location judgment step requires that the punching device must be in the designated punching area to complete the punching, which effectively prevents the behavior of remote punching, and further improves the fairness and effectiveness of attendance management. This application not only supports the punching method based on the Internet of Things positioning technology, but also is compatible with the traditional attendance machine punching method, taking into account the actual needs in different scenarios. For example, when the network is unstable, medical staff forget to bring the punching device or the punching device fails, medical staff can still complete the punching through the attendance machine, ensuring the continuity of attendance work. When the clock-in device is not in the clock-in area and the attendance machine does not record the clock-in information, the invalid output steps will be executed and an invalid clock-in signal will be sent to the existing attendance management platform. This timely exception handling mechanism helps managers to detect and handle attendance anomalies in a timely manner.

[0008] Optionally, the method further comprises: Data collection: collecting historical duty information, including historical duty time, historical duty location and historical duty plan; First modeling: Building a BI-LSTM model based on deep learning algorithm; First training: Use historical duty information to train the BI-LSTM model; Prediction: The current duty time and the current duty location are used as the input of the trained BI-LSTM model and the first duty plan is output.

[0009] By adopting the above technical solution, this application establishes a BI-LSTM model for analyzing and processing historical duty information. The BI-LSTM model can consider the time information before and after at the same time, and is suitable for processing complex scenarios that need to consider historical and future information at the same time, such as the prediction of duty plans. The BI-LSTM model is trained using historical duty information, and the patterns and rules in the historical duty information are learned by continuously adjusting the parameters of the BI-LSTM model to minimize the prediction error. Effective training can significantly improve the prediction ability of the BI-LSTM model. Taking the current duty time and location as input, the BI-LSTM model outputs the first duty plan. Since the BI-LSTM model can capture the complex relationships in historical data, the duty plan it outputs has high accuracy and practicality. Through automated and intelligent duty plan prediction, the workload of managers can be greatly reduced, the efficiency of scheduling can be improved, and the smooth progress of duty work can be ensured.

[0010] Optionally, after the step of performing the prediction, the method further comprises: Second modeling: Building an optimization model based on reinforcement learning algorithm; Setting space and function: including defining state space, defining action space and setting reward function; Define state space: Integrate all the information required for the current duty plan into the state space of the optimization model; Define action space: Integrate all historical duty plans into the action space of the optimization model; Set the reward function: Set the reward function , the calculation model is as follows: ; S is the current state; A is the action taken; is the reward for the i-th positive influencing factor; is the penalty for the jth negative influencing factor; Second training: Use historical duty information to train the optimization model; Generate a duty plan: Use the first duty plan as the input of the trained optimization model to generate the optimal duty plan.

[0011] By adopting the above technical solutions, the reinforcement learning algorithm learns through trial and error, so that the optimization model can continuously optimize its decision-making process in a dynamic environment. In the scenario of duty plan optimization, reinforcement learning can automatically adjust the strategy to generate a better duty plan based on historical data and real-time conditions. The optimization model is trained with historical duty information so that the optimization model can learn the laws and patterns in historical data. Through continuous trial and error and adjustment, the optimization model can gradually optimize its decision-making process and improve the accuracy and efficiency of generating duty plans. Taking the first duty plan as the input of the trained optimization model, the optimal duty plan can be generated based on the current state and historical experience. This solution not only takes into account the accuracy of the prediction, but also incorporates the decision-making wisdom of the optimization model, which is more in line with actual needs and scenario characteristics. The prediction model is responsible for providing a preliminary duty plan based on historical data, while the optimization model further optimizes and adjusts on this basis. The two work together to improve the overall performance and practicality of the system.

[0012] Optionally, after executing the step of setting the space and the function and before executing the step of second training, the method further includes: Initial generation plan: A greedy algorithm is used to combine the current duty information to initially generate a duty plan, which is recorded as the second duty plan; Pre-training: Use the second duty plan to pre-train the optimization model, use the pre-trained optimization model as a new optimization model, and execute the second training steps.

[0013] By adopting the above technical solution, a preliminary duty plan (second duty plan) is quickly generated by using a greedy algorithm combined with the current duty information. The introduction of this step provides a more reasonable starting point for the subsequent optimization model. Although the greedy algorithm may not necessarily find the global optimal solution, it can give a good approximate solution in a relatively short time, thereby improving the efficiency of the entire process. The optimization model is trained using the preliminary generated second duty plan. The purpose of this step is to allow the optimization model to have a preliminary understanding and adaptation to the generation process of the duty plan before formal training. Through pre-training, the optimization model can enter the learning state faster and reduce the trial and error cost during the formal training process. At the same time, pre-training can also improve the stability and convergence speed of the optimization model, making it easier for the optimization model to find the optimal solution in subsequent training. The prediction model first generates the first duty plan based on historical data, and then the greedy algorithm quickly generates the second duty plan as the preliminary input of the optimization model. Then, the optimization model preliminarily optimizes the second duty plan through pre-training, and further fine-tunes it in the subsequent second training stage. This phased, step-by-step optimization strategy ensures that the entire process remains efficient while also ensuring that the resulting duty plan is of high quality and practicality.

[0014] Optionally, after executing the step of generating a duty plan, the method further includes: Get location: Get the location information of the on-duty personnel based on the clock-in device; Get path: Generate duty path based on location information at different times; Area division: Divide the check-in area into multiple sub-areas; Path judgment: judge whether the duty path covers all sub-areas. If so, execute the step of obtaining the location; if not, execute the step of sending patrol information; Send patrol information: Send patrol information to the duty personnel to indicate the uncovered sub-areas.

[0015] By adopting the above technical solution, the duty path is generated using the location information at different times, which helps the duty personnel to perform their tasks more efficiently. The duty path is generated based on the location information to ensure that the duty personnel patrol in the predetermined order and route to avoid omissions or duplications. The clock-in area is divided into multiple sub-areas to achieve refined management of the duty area. This division method helps to clarify the responsible person and patrol requirements of each sub-area and improve the pertinence and effectiveness of the duty work. By checking whether the duty path covers all sub-areas, the comprehensiveness and completeness of the duty work are ensured. If the path does not cover all sub-areas, the system will send patrol information to instruct the duty personnel to patrol the uncovered areas, thereby reducing the patrol blind spots. By sending clear instructions to the duty personnel, they are reminded to patrol the uncovered sub-areas. This timely reminder mechanism helps the duty personnel to adjust the patrol plan in time so that all areas receive full attention and patrol.

[0016] Optionally, after executing the step of initialization setting and before executing the step of time and position determination, the method further includes: Status acquisition: Get the status of the clock-in device; Status judgment: judge whether the punching device is in the awake state. If so, execute the location judgment step; if not, execute the attendance machine judgment step.

[0017] By adopting the above technical solution, when the clock-in device is in the awake state, the location information is directly acquired and processed based on the current state of the clock-in device. When the clock-in device is not in the awake state, the attendance machine judgment step is executed, thereby reducing the phenomenon of false clocking in caused by medical staff placing the clock-in device in the clock-in area for a long time and increasing the authenticity of the attendance.

[0018] Optionally, after executing the attendance machine judgment step and before executing the invalid output step, the method further includes: Deployment of equipment: Deployment of clock-in equipment at the entrance and / or exit of the clock-in area; Device judgment: judge whether the punch-in device has captured the face information. If so, execute the matching step; if not, re-execute the device judgment step; Matching: matching the facial information with the personnel information of the attendance management platform to obtain the matching degree; Information judgment: judge whether the matching degree is greater than a preset threshold. If so, the time when the facial information is captured is recorded as the punch-in time. If not, execute the invalid output step.

[0019] By adopting the above technical solution, the probability of missed clocking-in events is reduced by deploying clocking-in devices at the entrance and / or exit of the clocking-in area. The application also obtains the information of passers-by through face recognition, and makes intelligent judgments based on whether the matching degree is greater than a preset threshold to avoid erroneous recording of attendance data. By introducing face recognition technology and automated matching and verification processes, the attendance management platform reduces employees' clocking-in time and operation steps, and improves the overall employee experience.

[0020] Optionally, the clock-in areas are divided into level one, level two and level three according to the security level from high to low. The level one clock-in area is provided with a level one personnel information whitelist, the level two clock-in area is provided with a level two personnel information whitelist, and the level three clock-in area is provided with a level three personnel information whitelist. Personnel on the level one personnel information whitelist can enter the level one, level two and level three clock-in areas, and persons on the level two personnel information whitelist can enter the level two and level three clock-in areas.

[0021] By adopting the above technical solution, by dividing the clock-in area into different security levels and setting up a corresponding whitelist system, it is possible to accurately control which personnel can enter which areas. This refined management method effectively reduces the risk of unauthorized personnel entering sensitive areas. Different security levels correspond to different permission ranges, thereby improving overall security. People on the first-level personnel information whitelist can enter all levels of clock-in areas, while people on the second-level personnel information whitelist can only enter the second-level and third-level clock-in areas. This clear division of permissions helps reduce permission confusion and abuse and improves the rational allocation and use of resources.

[0022] Optionally, after executing the step of position determination and before executing the step of attendance machine determination, the method further includes: Security judgment: Determine whether there are people in the low-level personnel information whitelist entering the high-level clocking area. If so, an alarm is issued; if not, the attendance machine judgment steps are executed.

[0023] By adopting the above technical solution and through security judgment steps, it is possible to monitor and identify in real time whether there are people in the low-level personnel information whitelist entering the high-level clocking-in area without authorization. This real-time security monitoring helps to promptly discover and respond to potential security risks and prevent the leakage of sensitive information or the occurrence of security accidents.

[0024] In the second aspect, the present application provides an attendance management system based on Internet of Things positioning technology, which adopts the following technical solutions: An attendance management system based on Internet of Things positioning technology, comprising: An initialization setting module includes a setting area unit and a setting information unit; Setting area unit, used to set the clock-in area; Setting information unit, used to set personnel information, punch-in time requirement information and punch-in device information; The time and location judgment module is used to judge whether the current time and location meet expectations; The attendance machine judgment module is used to judge whether the current person has completed the clocking in at the attendance machine; Invalid output module, used to send invalid punch-in signal.

[0025] By adopting the above technical solution, the system realizes comprehensive, accurate and real-time monitoring and management of the attendance process through the close cooperation of multiple modules such as the initialization setting module, the time and location judgment module, the attendance machine judgment module and the invalid output module. The system not only improves the efficiency and accuracy of attendance work, but also enhances security management, effectively prevents illegal clocking behavior, and provides an instant feedback mechanism to quickly respond to attendance problems. At the same time, its good scalability and maintainability enable the system to adapt to changes in different attendance needs.

[0026] In summary, the present application includes at least one of the following beneficial technical effects: 1. The initialization setting steps (including setting the area, personnel information, punch-in time requirements, and punch-in device information) provide the system with a high degree of flexibility and configurability, enabling the system to adapt to the attendance management needs of different hospitals or departments. Whether it is geographical location, personnel changes or working hours adjustments, it can adapt to its changes through simple configuration. The time and location judgment step determines whether the punch-in time and punch-in location are in line with expectations, that is, the punch-in behavior is only considered during the set working and leaving hours, which effectively reduces the mistaken punch-in or invalid punch-in during non-working hours, improves the accuracy of attendance data, and increases the rigor of punching based on the location judgment of the Internet of Things positioning technology.

[0027] 2. The time and location judgment step requires that the clock-in device must be in the designated clock-in area to complete the clock-in, which effectively prevents the behavior of remote clock-in and further improves the fairness and effectiveness of attendance management. This application not only supports the clock-in method based on IoT positioning technology, but is also compatible with the traditional clock-in method of attendance machines, taking into account the actual needs in different scenarios. For example, when the network is unstable, medical staff forget to bring the clock-in device, or the clock-in device fails, medical staff can still complete the clock-in through the attendance machine, ensuring the continuity of attendance work. When the clock-in device is not in the clock-in area and the attendance machine does not record the clock-in information, the invalid output step will be executed to send an invalid clock-in signal to the existing attendance management platform. This timely exception handling mechanism helps managers to promptly discover and handle attendance anomalies. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flow chart of the method of Example 1 of the present application; Figure 2 is a flow chart of the method from S2 time judgment to S3 position judgment of Example 1 of the present application; Figure 3 It is a flow chart of the method from S601 data collection to S610 generation of duty plan in Example 2 of the present application; Figure 4 This is a flow chart of the S7 patrol method of Example 2 of the present application. DETAILED DESCRIPTION

[0029] The following combination Figures 1 to 4 This application is described in further detail.

[0030] With traditional fingerprint or face recognition clocking in, medical staff need to take the initiative to go to the attendance machine to operate. However, medical staff are busy and stressed at work, and sometimes they may need to deal with some emergencies. They often forget to clock in or do not have time to clock in at the attendance machine. If they forget to clock in or miss a clock in, if they do not submit a replacement card application in time, it will affect the performance of medical staff. With fingerprint or face recognition clocking in, they need to take off their gloves or masks every time they clock in, which invisibly increases the workload of medical staff and easily causes the spread of germs.

[0031] Furthermore, when doctors or nurses encounter special circumstances (for example, a patient needs emergency rescue) and are unable to go to the punch-in location to punch in, as long as the medical staff wears a badge with them, they can realize non-perceptual punch-in. This application plans and deploys the IoT gateway according to the activity areas of medical staff's daily work, such as emergency rooms, operating rooms, pharmacies and other locations. When medical staff arrive at the designated rescue area to carry out rescue work and have no time to punch in, they can use the punch-in device to realize non-perceptual attendance punch-in, which is convenient for timely deployment of manpower and recording attendance. The punch-in device will broadcast a signal in real time. When the IoT base station receives the signal broadcast by the punch-in device, it can be determined that the punch-in device appears within the coverage of the base station. The IoT base station will also receive the MAC address information of the punch-in device. The MAC address corresponds one-to-one with the medical staff, so that the punch-in personnel information can be determined and the punch-in can be completed.

[0032] Therefore, the present application provides an attendance method and system based on Internet of Things positioning technology to solve the above problems.

[0033] Embodiment 1: This embodiment discloses an attendance management method based on Internet of Things positioning technology, referring to Figure 1 The method includes the following steps: S1 initialization setting, S2 time and location judgment, S3 attendance machine judgment and S4 invalid output. First, the punch-in area and related information are initially set, and then it is judged whether the punch-in time and punch-in location meet the expectations. If the medical staff does not carry a punch-in device, the attendance machine is used to punch in. The specific process of this embodiment is as follows: S1 initialization setting, including S11 setting area and S12 setting information.

[0034] S11 sets the area. According to the actual layout of the hospital and the attendance management needs, different clock-in areas are demarcated by deploying different IoT devices. The clock-in areas include: offices where medical staff are located, inpatient departments, operating rooms, security booths, etc.

[0035] The clock-in area is divided into the first-level clock-in area, the second-level clock-in area and the third-level clock-in area according to the security level from high to low. The first-level clock-in area is set with a first-level personnel information whitelist, the second-level clock-in area is set with a second-level personnel information whitelist, and the third-level clock-in area is set with a third-level personnel information whitelist. People on the first-level personnel information whitelist can enter the first-level clock-in area, the second-level clock-in area and the third-level clock-in area, and people on the second-level personnel information whitelist can enter the second-level clock-in area and the third-level clock-in area, and people on the third-level personnel whitelist can only enter the third-level clock-in area. When the third-level clock-in area enters the second-level clock-in area, and the second-level clock-in area enters the first-level clock-in area, there are multiple verification checkpoints. The clock-in device information is used to determine whether the current person belongs to the whitelist of the clock-in area at each level, or whether the current person belongs to the whitelist of a higher-level person. If so, the verification checkpoint will directly release the person, otherwise, the verification checkpoint will prohibit the current person from passing.

[0036] S12 sets information, including personnel information, punch-in time requirement information, and punch-in device information.

[0037] Enter all the personnel information that needs to participate in attendance in the clock-in device, including basic information such as name, work number and department. According to the personnel's job responsibilities and actual needs, associate each person with the corresponding clock-in area by setting permissions, roles or assigning area codes. For example: the nurse is associated with the IoT of the inpatient department. The nurse can only clock in if the clock-in device enters the IoT coverage of the inpatient department.

[0038] According to the company's work and rest system, set the time range for clocking in on each workday (such as 0:00 to 8:00 in the morning and 1:00 to 1:30 after get off work) and the time range for clocking out off work (such as 12:00 to 1:30 in the morning and 5:00 to 23:59 in the afternoon).

[0039] Register and configure all time clock devices (such as badges), including their unique identifiers, the types of functions they support (such as positioning, time stamp recording, etc.), and the status of the time clock device (whether it is online, whether it needs maintenance, etc.).

[0040] S2 time and location judgment, based on the Internet of Things positioning technology, judge whether the current time and location meet the expectations. If so, record the clock-in time and clock-in location; if not, execute S3 attendance machine judgment.

[0041] Judging whether the current time meets expectations means: judging whether the current time belongs to the time to punch in or the time to punch out. Judging whether the current location meets expectations means: judging whether the punch-in device appears in the punch-in area based on the Internet of Things positioning technology (such as GPS, Wi-Fi positioning, Bluetooth beacons, etc.). In this embodiment, it is possible to first judge whether the punch-in time meets expectations and then judge whether the punch-in location meets expectations, or it is possible to first judge whether the punch-in location meets expectations and then judge whether the punch-in time meets expectations, or it is possible to judge whether the punch-in time and the punch-in location meet expectations at the same time. If and only if both meet expectations at the same time, the punch-in time and the punch-in location will be recorded; otherwise, the S3 attendance machine judgment is executed.

[0042] Within the valid clock-in time range, medical staff wear badges to enter the clock-in area. The IoT gateway will detect the presence of the badges and set the personnel information bound to the badges. The monitored time information, personnel information, and clock-in location information will be pushed to the existing attendance management platform. Medical staff do not need to perform any operations during the entire process, and attendance data is collected without perception, making attendance clock-in more convenient and faster, and reducing the workload of medical staff. When medical staff have completed clocking in on the attendance machine and completed non-perceptual attendance clocking in through the badge, the earliest clock-in time and the latest clock-in time are obtained by comparing the two times. The earliest clock-in time is used as the clock-in time for work, and the latest clock-in time is used as the clock-in time for leaving work. The earliest clock-in time and the latest clock-in time refer to the earliest clock-in time and the latest clock-in time within the valid clock-in time range.

[0043] S3 attendance machine judgment, for situations where the validity of the clock-in cannot be confirmed by S2 time and location judgment (such as employees forget to bring the clock-in device and use the attendance machine to complete the clock-in), automatically check the attendance machine records to confirm whether the current employee has completed the clock-in operation on the attendance machine. If so, record the clock-in time and clock-in location; if not, execute S4 invalid output.

[0044] S4 invalid output, when it is judged as invalid punch-in, it automatically sends an invalid punch-in signal to the attendance management platform.

[0045] Take nursing management as an example: Nurses play an increasingly important role in hospitals, such as cleaning operating rooms, transferring patients, accompanying patients, etc. However, nurses are often highly mobile and their work areas are not fixed, which puts a lot of burden on the hospital's management. If the traditional management method is adopted, not only the fingerprint information of the workers must be collected and maintained, but it is also difficult to supervise the work process of the nurses. This application distributes clock-in devices (such as badges or bracelets) to nurses, and binds the badges or bracelets to the nurses on the existing attendance management platform. When the nurses enter the wards, operating rooms and other areas, they can automatically clock in, and can also view the working hours and working hours of the workers' work orders on the attendance management platform.

[0046] This embodiment realizes the contactless clocking in of medical staff through the Internet of Things positioning technology. By cooperating with the clocking in equipment, the workload of medical staff is reduced and safety is improved.

[0047] like Figure 2 As shown, in other embodiments, after executing S1 initialization setting and before executing S2 time and position determination, it also includes: S51 status acquisition, obtains the status of the clock-in device. The status information includes: the device's power status, network connection status, sensor status (such as GPS, Bluetooth, etc.), and whether it is in sleep or wake-up mode.

[0048] S52 status judgment, judge whether the punching device is in the awake state and can respond to the punching request immediately. If so, execute S2 time and location judgment to judge whether the current time is within the punching time requirement, and judge whether the punching device is located in the designated punching area based on the Internet of Things positioning technology; if not, it proves that the current punching device is unavailable and the location of the punching device cannot be obtained, then execute S3 attendance machine judgment to verify the punching behavior of medical staff.

[0049] Taking clocking in after work as an example, there are two situations in which medical staff punch in without feeling using a clocking-in device (taking a badge as an example): The first type: medical staff wear badges when leaving get off work. During the effective time range for clocking out, when the IoT gateway detects that the badge disappears from the clocking-in area, it will report the time when the badge disappears and the information of the medical staff bound to it. At this time, the time when the medical staff last left the clocking-in area will be used as the time when they clock out. The second type: after leaving get off work, medical staff place their badges on their workstations and leave on their own. After the badges are stationary for a period of time, the badges will go from awake to dormant. During the effective time range for clocking out, when the IoT gateway detects that the badge goes from awake to dormant and remains in dormant for a period of time, the IoT gateway will push the time when the badge goes from awake to dormant and the information of the medical staff bound to the badge to the attendance management platform. The system will use the time when the badge last went from awake to dormant as the time when they clock out.

[0050] The system will match the medical staff's punch-in data with the day's shift schedule, automatically match the employee's punch-in time to the corresponding shift, and accurately determine whether the employee's attendance is normal.

[0051] After executing S2 time and location judgment and before executing S3 attendance machine judgment, it also includes: S21 security judgment, through the comprehensive geographical location information, employee authority level and the authority level of the clocking area, to judge whether there is a person in the low-level personnel information whitelist entering the high-level clocking area. If so, an alarm is issued to warn the management personnel of the risk. If not, the S3 attendance machine judgment is executed.

[0052] Take the operating room as an example: The operating room is an important area in the hospital, and there is strict control over the entry and exit of personnel. When authorized medical staff enter the operating room area, in order to avoid contact with the attendance machine and contamination of bacteria, thereby increasing the risk of surgery, they can complete the clocking in through non-perception technology to accurately record the time when the medical staff participated in the operation. At the same time, the system can also be bound to the access control system. When the clocking in is successful, the door of the area will open automatically, and the medical staff does not have to open the door manually, simplifying the operation process. When unauthorized personnel enter the operating room area, the system can also trigger an alarm reminder to notify security personnel to go to the scene for disposal, and at the same time, the alarm information can be pushed to the access control system to automatically lock the door.

[0053] After executing S3 attendance machine judgment and before executing S4 invalid output, it also includes: S31 deploys equipment, deploys clock-in equipment (specifically equipment such as cameras with facial recognition functions) at the entrance and / or exit of the clock-in area to facilitate medical staff to complete the clock-in when entering or leaving the work area.

[0054] S32 device judgment, judging whether the punch-in device has captured the face information, if so, executing S33 matching; if not, re-executing S32 device judgment. The face information can be a face image of a medical staff member wearing a mask.

[0055] S33 matching, matching the facial information with the personnel information stored in the attendance management platform to obtain a matching degree. The matching process calculates the similarity or matching degree of facial features to determine whether the captured face belongs to an employee in the system.

[0056] The attendance management platform stores a large amount of facial image data of people wearing masks and people not wearing masks, and annotates the data, and records the data as first data. The first data is used to train a convolutional neural network (CNN) or other deep learning model so that it can learn the characteristics of faces wearing masks, and the trained convolutional neural network (CNN) or other deep learning model is deployed in the clock-in device for real-time face detection and recognition.

[0057] S34 information judgment, judging whether the matching degree is greater than a preset threshold, if so, the time when the facial information is captured is recorded as the punch-in time, if not, executing S4 invalid output.

[0058] Take the example of a medical staff clocking in through facial recognition without wearing a badge: When medical staff do not wear badges and it is inconvenient for them to clock in through the attendance machine, they can clock in automatically and without perception through a facial recognition camera. Facial recognition cameras can be deployed in the medical staff's work area or at the entrance and exit. When the camera captures the medical staff's facial information, it will be compared with the facial database in the attendance management platform to confirm the medical staff's identity information and record the facial capture time as the time the medical staff clocks in or out of get off work.

[0059] This application also sets up an electronic fence in the hospital. During working hours, if medical staff leave their daily work area, the IoT gateway will push the information of the medical staff who left and the location where they crossed the boundary to the attendance management platform, providing administrative personnel with timely disposal.

[0060] By combining IoT positioning technology and face recognition technology, this embodiment enables the system to more accurately determine whether an employee punches in within a designated area and verify the identity of the person punching in, thereby avoiding problems such as punching in for someone else or false punching in, and improving the accuracy and credibility of attendance data. Punching equipment is deployed at the entrance and / or exit of the punching area, so that employees can conveniently punch in when entering or leaving the work area, improving the convenience and efficiency of punching in. At the same time, the ability to recognize faces wearing masks also meets the needs of medical facilities and improves the user experience.

[0061] Example 2: Figure 3As shown, the difference from Example 1 is that the method further comprises: S601 data collection, collecting historical duty information, the historical duty information includes: historical duty time, historical duty location and historical duty plan.

[0062] Historical duty time records the specific time of each duty, including date, start time and end time.

[0063] Historical duty locations indicate the specific location or area of ​​each duty.

[0064] The historical duty plan describes in detail the specific plans for each duty, such as personnel deployment and task allocation.

[0065] S602 first models a BI-LSTM model based on a deep learning algorithm. The BI-LSTM model can process sequence data and capture long-term dependencies in the data. It is suitable for analyzing time series data (such as duty hours) and contextual information (such as duty locations).

[0066] S603 is the first training, in which the BI-LSTM model is trained using historical duty information. During the training process, the BI-LSTM model will learn the patterns in the historical data.

[0067] S604 predicts, taking the current duty time and the current duty location as inputs of the trained BI-LSTM model, performing predictions through the trained BI-LSTM model, and outputting a first duty plan.

[0068] S605 second modeling, based on the reinforcement learning algorithm, establishes an optimization model. Reinforcement learning allows the optimization model to learn the optimal strategy by continuous trial and error in the environment, which is used to optimize the first duty plan.

[0069] S606 sets the space and function, including defining the state space, defining the action space and setting the reward function.

[0070] S6061 Define state space: Integrate all the information required for the current duty plan (such as personnel status, task progress, resource allocation, etc.) into the state space of the optimization model.

[0071] S6062 defines the action space: all historical duty plans (or plans that have undergone certain transformations) are integrated into the action space of the optimization model, and the action space is the set of all possible actions (i.e., duty plans) that the optimization model can choose.

[0072] S6063 Set reward function, set reward function , the calculation model is as follows: ; S is the current state; A is the action taken; is the reward for the i-th positive influencing factor; is the penalty for the jth negative influencing factor.

[0073] The reward function is used to evaluate the quality of the action (duty plan) selected by the model. It calculates the reward or penalty based on the current state (S) and the action taken (A). The reward function usually contains multiple positive influencing factors (such as task completion efficiency, personnel satisfaction) and negative influencing factors (such as the ratio of cost to benefit, risk level).

[0074] For example: the task completion efficiency is 80%, the staff satisfaction is 90%, the cost-benefit ratio is 0.5, and the risk level is 30%, then .

[0075] S607 generates a preliminary plan, using a greedy algorithm combined with current duty information to preliminarily generate a duty plan, which is recorded as the second duty plan.

[0076] S608 pre-training, using the second duty plan to pre-train the optimization model, using the pre-trained optimization model as a new optimization model, and executing S609 second training.

[0077] S609 Second training: Use historical duty information to train the optimization model. During the training process, the optimization model will continuously try different duty plans and evaluate the quality of these plans based on the reward function, so as to gradually learn how to generate the optimal duty plan.

[0078] S610 generates a duty plan: uses the first duty plan as an input of the trained optimization model to generate an optimal duty plan.

[0079] For example, suppose a hospital needs to develop a duty plan for its emergency department. The duty plan for the emergency department needs to take into account the professional skills of doctors, the shift system, the urgency of patients, and the impact of special events (such as an epidemic outbreak).

[0080] A. Data Collection,The hospital collected the emergency department duty information in the past year,,including: Historical duty hours: 8:00 to 20:00, Monday to Sunday, divided into four time periods (8:00-12:00, 12:00-16:00, 16:00-20:00, night shift).

[0081] Historical duty locations: different areas within the emergency department (such as emergency room, observation room, operating room, etc.).

[0082] Historical duty plan: records the specific list of doctors on duty each time, their professional background, the number and type of cases handled, etc.

[0083] B. First, modeling and training, using the BI-LSTM model to learn the relationship between historical duty time and location and duty plan. The model input includes time (encoded as day of the week and time period) and location (encoded as different areas of the emergency department), and the output is a preliminary duty plan prediction, including a list of recommended doctors and possible shift arrangements.

[0084] The BI-LSTM model learns patterns by traversing historical data. For example, it may find that more experienced doctors are needed on weekends and evenings, while fewer are needed during the day on weekdays.

[0085] Assume that it is Friday night (16:00-20:00) and the emergency department needs to make a duty plan. The BI-LSTM model predicts a preliminary duty plan based on this time and location, including several recommended doctors on duty.

[0086] C. Build an optimization model based on reinforcement learning, the goal of which is to adjust the preliminary duty plan to maximize efficiency, patient satisfaction and minimize cost.

[0087] Status space: includes information such as current time, location, doctor status (such as availability, professional skills), and patient waiting status.

[0088] Action space: includes all possible combinations of doctors and shift arrangements, which can be extracted from historical data and transformed appropriately.

[0089] Reward function: includes positive rewards and negative penalties.

[0090] Positive rewards: number of emergency cases successfully handled, patient satisfaction scores, doctor work efficiency, etc.

[0091] Negative penalties: risk of doctor burnout, long waiting times for patients, costs exceeding budget, etc.

[0092] D. Use a greedy algorithm to combine current conditions (such as the possible increase in patient volume on Friday night) to initially generate a duty plan. Then, use this plan to pre-train the reinforcement learning model to quickly adapt to the current environment.

[0093] The reinforcement learning model is trained by simulating different duty plans and observing their impact on the reward function. The model will gradually learn which duty plans are optimal and continuously adjust to cope with new situations in actual applications.

[0094] E. The preliminary duty plan predicted by the BI-LSTM model is used as the input of the reinforcement learning model. After further adjustment and optimization of the optimization model, an optimal duty plan is generated.

[0095] This embodiment predicts future duty information by combining historical duty information with deep learning algorithms and reinforcement learning algorithms. Specifically, this embodiment uses the BI-LSTM model to mine the impact pattern of time and place on the duty plan from historical data, and preliminarily predicts the duty arrangement that meets conventional needs. Subsequently, the preliminary plan is finely optimized through the optimization model, taking into account more dynamic factors and constraints in actual implementation, such as the doctor's professional skills, the patient's urgency, and the hospital's resource allocation. This dual-model combination not only improves the scientificity and rationality of the duty plan, but also enhances the adaptability and flexibility of the duty plan.

[0096] like Figure 4 As shown, in other embodiments, after executing the step of generating a duty plan, the method further includes: S7 patrol: includes S71 obtaining position, S72 obtaining path, S73 area division, S74 path judgment and S75 sending patrol information.

[0097] S71 Obtaining location: Obtaining the location information of the on-duty personnel based on the clock-in device.

[0098] S72 Obtain Path: Generate a duty path based on location information at different times, and connect discrete location points into a continuous duty path through the Geographic Information System (GIS).

[0099] S73 Area division: Divide the clock-in area into multiple sub-areas. The clock-in area in this step specifically refers to the area that needs to be patrolled by the on-duty personnel.

[0100] S74 path judgment: judge whether the duty path covers all sub-areas. If so, it means that the current patrol task has been completed, then execute S71 to obtain the location; if not, it means that there are security risks or management loopholes, then execute S75 to send patrol information.

[0101] S75 Send patrol information: Send patrol information to the on-duty personnel to indicate the uncovered sub-areas.

[0102] Take security and / or cleaning attendance management as an example: When security guards and / or cleaning staff are active in the designated work area of ​​the hospital, they will automatically clock in to ensure the cleanliness of the hospital environment and the normal development of safety work. Especially for security guards, it can also realize the visual patrol function. Traditional security guards need to carry patrol sticks when patrolling. When arriving at the patrol point, they need to bring the patrol stick close to the information button of the patrol point for induction. The patrol is more effective only after successful recognition. The above operation needs to be repeated every time a patrol point is reached. The entire patrol process is very cumbersome, and there is a lack of supervision over the patrol process and patrol trajectory. This application deploys an IoT base station at the patrol location in advance. When the security guard arrives at the patrol point with a clock-in device, the signal sent by the clock-in device will be received by the IoT base station. Successful reception indicates that the patrol of the patrol point is completed. The security personnel do not need to perform any operation during this patrol process, and the entire process is completed without perception.

[0103] This embodiment uses the clock-in device to obtain the position of the on-duty personnel in real time, generates a patrol path, and subdivides the patrol area into multiple sub-areas for monitoring. The system can automatically determine whether the path fully covers all sub-areas. If there are omissions, the patrol information is immediately sent to the on-duty personnel to instruct them to go to the uncovered area to ensure the comprehensiveness and efficiency of the patrol work, thereby improving the overall safety management level.

[0104] The clock-in device can not only be used for medical staff's non-perception attendance clock-in, but can also be used as a medical staff employee card, access card, and canteen consumption card. By connecting the clock-in device with the access control system and the canteen consumption system, the all-in-one card function can be realized.

[0105] Embodiment 3: This embodiment discloses an attendance management system based on Internet of Things positioning technology, the system comprising: Initialization setting module: includes setting area unit and setting information unit.

[0106] The area setting unit is used to set the clock-in area. The clock-in area refers to the physical space where medical staff need to clock in and out. The hospital can flexibly set multiple clock-in areas according to actual conditions, such as office floors, department areas, etc.

[0107] The information setting unit is used to set personnel information, punch-in time requirement information and punch-in device information.

[0108] The time and location judgment module is used to judge whether the current time and location meet expectations, that is, to judge whether the current time is the time to punch in or the time to punch out. This module judges in real time whether the current time is the time to punch in or the time to punch out according to the punch time requirement information in the set information unit. This module ensures the accuracy and timeliness of time judgment by synchronizing with the system clock. If the current time meets the punching requirements, the system will allow employees to punch in; otherwise, it will prompt employees that it is not the time to punch in. This module also judges whether the punching device appears in the punching area based on the Internet of Things positioning technology. If the punching device appears in the set punching area, the punching is considered valid.

[0109] The attendance machine judgment module is used to determine whether the current person has completed clocking in at the attendance machine.

[0110] Invalid output module, used to send invalid punch-in signal to the attendance management platform.

[0111] This embodiment provides an attendance management system based on the Internet of Things positioning technology. The initialization setting module includes setting the area and personnel, time, equipment and other information to ensure that the attendance management platform can accurately identify the punch-in area, employee identity and punch-in rules. The system comprehensively verifies the validity of the punch-in through the time judgment, location judgment and attendance machine judgment modules. If invalid behaviors such as non-specified time, non-designated area or failure to punch in through the attendance machine are found, the invalid output module will send a signal to the management system and record it to ensure the strictness and fairness of attendance management.

[0112] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. An attendance management method based on Internet of Things positioning technology, characterized in that: include: Initialization settings: including setting area and setting information; Set area: Set the check-in area; Setting information: Setting personnel information, punch-in time requirement information and punch-in device information; Time and location judgment: Based on the IoT positioning technology, judge whether the current time and location meet the expectations. If so, record the clock-in time and clock-in location; if not, execute the steps judged by the attendance machine; Attendance machine judgment: judge whether the current person has completed the clocking in at the attendance machine. If yes, record the clocking in time and location; if no, execute the invalid output step; Invalid output: Send an invalid punch-in signal.

2. The attendance management method based on Internet of Things positioning technology according to claim 1 is characterized in that: The method further comprises: Data collection: collecting historical duty information, including historical duty time, historical duty location and historical duty plan; First modeling: Building a BI-LSTM model based on deep learning algorithm; First training: Use historical duty information to train the BI-LSTM model; Prediction: The current duty time and the current duty location are used as the input of the trained BI-LSTM model and the first duty plan is output.

3. The attendance management method based on Internet of Things positioning technology according to claim 2 is characterized in that: After the step of performing the prediction, the method further comprises: Second modeling: Building an optimization model based on reinforcement learning algorithm; Setting space and function: including defining state space, defining action space and setting reward function; Define state space: Integrate all the information required for the current duty plan into the state space of the optimization model; Define action space: Integrate all historical duty plans into the action space of the optimization model; Set the reward function: Set the reward function , the calculation model is as follows: ； S is the current state; A is the action taken; is the reward for the i-th positive influencing factor; is the penalty for the jth negative influencing factor; Second training: Use historical duty information to train the optimization model; Generate a duty plan: Use the first duty plan as the input of the trained optimization model to generate the optimal duty plan.

4. The attendance management method based on Internet of Things positioning technology according to claim 3 is characterized in that: After executing the step of setting the space and the function and before executing the step of the second training, the method further includes: Initial generation plan: A greedy algorithm is used to combine the current duty information to initially generate a duty plan, which is recorded as the second duty plan; Pre-training: Use the second duty plan to pre-train the optimization model, use the pre-trained optimization model as a new optimization model, and execute the second training steps.

5. The attendance management method based on Internet of Things positioning technology according to claim 4 is characterized in that: After executing the step of generating the duty plan, it also includes: Get location: Get the location information of the on-duty personnel based on the clock-in device; Get path: Generate duty path based on location information at different times; Area division: Divide the check-in area into multiple sub-areas; Path judgment: judge whether the duty path covers all sub-areas. If so, execute the step of obtaining the location; if not, execute the step of sending patrol information; Send patrol information: Send patrol information to the duty personnel to indicate the uncovered sub-areas.

6. The attendance management method based on Internet of Things positioning technology according to any one of claims 1 to 5, characterized in that: After executing the step of initialization setting and before executing the step of time and position determination, the method further includes: Status acquisition: Get the status of the clock-in device; Status judgment: judge whether the punching device is in the awake state. If so, execute the time and location judgment steps; if not, execute the attendance machine judgment steps.

7. The attendance management method based on Internet of Things positioning technology according to claim 6 is characterized in that: After executing the step of attendance machine judgment and before executing the step of invalid output, it also includes: Deployment of equipment: Deployment of clock-in equipment at the entrance and / or exit of the clock-in area; Device judgment: judge whether the punch-in device has captured the face information. If so, execute the matching step; if not, re-execute the device judgment step; Matching: matching the facial information with the personnel information of the attendance management platform to obtain the matching degree; Information judgment: judge whether the matching degree is greater than a preset threshold. If so, the time when the facial information is captured is recorded as the punch-in time. If not, execute the invalid output step.

8. The attendance management method based on Internet of Things positioning technology according to claim 7 is characterized in that: The clock-in areas are divided into level 1, level 2 and level 3 according to the security level from high to low. The level 1 clock-in area is equipped with a level 1 personnel information whitelist, the level 2 clock-in area is equipped with a level 2 personnel information whitelist, and the level 3 clock-in area is equipped with a level 3 personnel information whitelist. People on the level 1 personnel information whitelist can enter the level 1, level 2 and level 3 clock-in areas, and people on the level 2 personnel information whitelist can enter the level 2 and level 3 clock-in areas.

9. The attendance management method based on Internet of Things positioning technology according to claim 8 is characterized in that: After executing the step of determining the time and location and before executing the step of determining the attendance machine, the method further includes: Security judgment: Determine whether there are people in the low-level personnel information whitelist entering the high-level clocking area. If so, an alarm is issued; if not, the attendance machine judgment steps are executed.

10. An attendance management system based on Internet of Things positioning technology, the system being applicable to the method according to any one of claims 1 to 9, comprising: An initialization setting module includes a setting area unit and a setting information unit; Setting area unit, used to set the clock-in area; Setting information unit, used to set personnel information, punch-in time requirement information and punch-in device information; The time and location judgment module is used to judge whether the current time and location meet expectations; The attendance machine judgment module is used to judge whether the current person has completed the clocking in at the attendance machine; Invalid output module, used to send invalid punch-in signal.

Images