Personnel statistics management method and system based on state machine mechanism

Abstract

The invention belongs to the technical field of access management equipment, and particularly relates to a personnel statistics management method and system based on a state machine mechanism. The method comprises the following steps: arranging at least two RFID reading modules in a passage through which a person passes, and reading RFID tag information carried by the person; the server carries out personnel state judgment according to a preset state rule and carries out one-time personnel access data statistics according to a state judgment result; an image monitoring module is arranged in a channel, dynamic image information of personnel is captured, and secondary personnel access data statistics is carried out according to a formed personnel track; and periodically fusing, comparing and counting the two-time personnel access data to calculate the number of people, and outputting the number of people in the area and summarizing the access condition in real time. The method is used for solving the problem that people counting is difficult for large-scale people. Personnel statistics can be better carried out in a complex scene, the accuracy of the number of personnel is ensured, and the image monitoring module is combined with the in-out direction to carry out personnel in-out condition summary statistics.

Description

Technical Field

[0001] The present invention belongs to the technical field of access management devices, and particularly relates to a personnel statistics management method and system based on a state machine mechanism. Background Art

[0002] The factory operation environment is complex with numerous workshops and a large flow of people. In order to better build factory safety and ensure the safety of workers, management personnel need to constantly pay attention to the situation of personnel staying in the operation area, understand the personnel entry and exit conditions of each operation area and workshop, so as to control the personnel distribution in each area.

[0003] At present, RFID (Radio Frequency Identification) technology has been widely used in personnel and goods access management systems. Its basic principle is to achieve automatic identity recognition and data recording through the interaction between RFID tags and reading and writing devices. The mainstream RFID systems on the market usually use a single-direction detection method: when personnel or goods carry active / passive RFID tags and pass through the detection device, the system only identifies it as "entry", and does not effectively determine the "exit" direction.

[0004] In order to achieve two-way determination, some current detection devices use multiple sets of devices for assistance. For example, the solution of patent code CN201020192378 uses an external infrared device to assist the RFID antenna for direction determination, and judges the entry and exit directions by detecting the action sequence of personnel in front of the infrared sensor. However, this solution still requires additional deployment of sensor peripherals, and it is impossible to accurately judge the personnel entry and exit conditions in an environment with a large flow of people; Another example: the solution of patent code CN201710037460 designs two RFID detection devices, and deduces the moving direction of personnel by collecting the change of tag signal strength (RSSI) within a period of time. Its implementation mechanism includes: the RFID reader collects the tag signal strength value received by the antenna and combines it with the directionality algorithm for direction determination. However, this mechanism only describes judging personnel entry and exit by signal strength, and the definition of personnel position status information is vague. This mechanism is difficult to cope with complex on-site scenarios, such as the behavior of personnel with different speeds, wandering or repeatedly entering and exiting in a short time, which may lead to misjudgment of direction and is not applicable to places with large-scale crowd flow. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, a personnel statistics management method and system based on a state machine mechanism are provided to solve the problem of difficult statistics of a large number of people.

[0006] The technical solution of the present invention to solve the above technical problems is as follows: A personnel statistics management method based on a state machine mechanism, which includes: By arranging at least two RFID reading modules in the passage where people pass through, the RFID tag information carried by the people is read and the state conversion signal is synchronously triggered; The data read by the RFID reading module is transmitted to the server in real time. The server determines the personnel status according to the preset status rules and performs personnel entry and exit data statistics based on the status determination results. By placing image monitoring modules in the channel, dynamic image information of personnel is captured, and secondary personnel entry and exit data statistics are performed based on the formed personnel trajectory; Utilizing the data information from the RFID reading module and the image monitoring module, combined with status rules, the data of two personnel entry and exit are periodically fused and compared to calculate the number of people entering and leaving, and the number of people in the area and the summary of entry and exit situations are output in real time.

[0007] Compared with the prior art, the above technical solution has the following beneficial effects: The RFID reading module is set to two channels, and the two RFID reading modules are set at intervals. This can accurately determine the direction of personnel entry and exit. In conjunction with the state machine mechanism, the status of personnel carrying RFID tags can be switched in real time, so that personnel statistics can be better performed in complex scenarios. In addition, the secondary personnel entry and exit data detected by the image monitoring module can be integrated and reviewed to ensure the accuracy of the number of personnel. In addition, the image monitoring module combines the entry and exit directions to summarize and count the entry and exit of personnel.

[0008] Based on the above technical solution, the embodiment of the present application can also be improved as follows: Furthermore, the personnel status determination is performed according to the preset status rules as follows: First, the personnel status is abstracted into at least four states that are switched in sequence along the passage direction, and the four states are triggered to switch by the two RFID reading modules respectively; If a person enters the fourth state from the first state, or returns to the first state from the fourth state, a person entry and exit data statistics is performed; If a person switches from the first state or the fourth state to the second state or the third state and stays there for more than a certain period of time or returns, it is determined to be an abnormal state, recorded and an alarm is issued.

[0009] Furthermore, the periodic fusion comparison and statistics of two personnel entry and exit data specifically include: Align the data of RFID reading module and image monitoring module through timestamp to achieve time synchronization; Map the position of the person carrying the RFID tag to the plane coordinate system monitored by the image monitoring module, and compare the trajectory range detected by the RFID reading module and the image monitoring module; Associate the data detected by the RFID reading module and the image detection module through the characteristics of time and space.

[0010] Furthermore, the periodic fusion and comparison for statistically analyzing the personnel in-and-out data twice further includes: Compare the number of people counted by the RFID reading module and the image monitoring module. If they are consistent, adopt it; Compare the trajectory directions detected by the RFID reading module and the image monitoring module. If they are consistent, adopt it; otherwise, take the direction detected by the image monitoring module as the standard.

[0011] Furthermore, the periodic fusion, comparison, and statistical analysis of the personnel in-and-out data twice further includes: Decode the video to obtain image frames, input the image frames into the training model, and obtain the target bounding boxes and confidence levels; Based on the multi-object tracking algorithm, filter the active trajectories for matching and updating the trajectory status; Update the total number of people in the area according to the in-and-out directions of the trajectories, and periodically output the statistical results per unit time.

[0012] Furthermore, after obtaining the target bounding boxes and confidence levels, filter the active trajectories through the following steps: Classify the targets into high-confidence targets and low-confidence targets according to the confidence levels; Use the Kalman filter to predict the next position of the trajectory, calculate the intersection over union (IoU) between the existing trajectory and the high-confidence targets. If the detection match is successful, mark the trajectory status as a high-confidence trajectory; If not successful, send the trajectory to the low-confidence for matching, calculate the IoU between this trajectory and the low-confidence targets. If the detection match is successful, mark the trajectory status as a low-confidence trajectory; If not successful, mark it as an invalid trajectory.

[0013] Furthermore, the specific steps for updating the total number of people in the area according to the in-and-out directions of the trajectories specifically include: Define at least two virtual judgment lines in the monitoring area, calculate the center points of the areas delimited by the virtual judgment lines, perform crossing detection based on the center points, mark the in-and-out directions of the personnel according to the direction in which the center point of the personnel crosses the judgment line, and update the total number of people in the area according to the in-and-out directions of the trajectories.

[0014] Furthermore, the RFID reading module regularly clears redundant data through a cache cleaning mechanism according to the following measures to ensure real-time data update; The RFID reading module reads the same person carrying an RFID tag once within the first time threshold; For data determined to be in an abnormal state exceeding the time threshold, cache cleaning is performed on the data for which the personnel in-and-out data statistics have been completed.

[0015] Further, configure the RFID tag on the personnel's clothing or personal belongings to ensure that the tag moves synchronously with the personnel; Use the multi-target tracking algorithm to determine the state consistency of the target personnel between multiple frames, and automatically trigger an alarm and record for the personnel detected and determined not to carry an RFID tag and abnormal states.

[0016] The present invention also discloses a personnel statistics management system based on a state machine mechanism, which includes: An RFID tag for the personnel to carry with them; An RFID reading device, which includes at least two detection modules. The detection module is used to form a state detection area. When the RFID tag passes through the state detection area, the RFID reading device is used to read the tag information of the RFID tag and trigger a state conversion signal at the same time; An image monitoring module for synchronously capturing the dynamic image information of the personnel and counting the personnel in and out information according to the data information; A server for receiving the data transmitted by the RFID reading device and the image monitoring module. The server is further configured to perform the following steps: By arranging at least two RFID reading modules in the passage where the personnel pass, read the RFID tag information carried by the personnel and trigger the state conversion signal synchronously; Transmit the data read by the RFID reading module to the server in real time. The server determines the personnel state according to the preset state rules and performs a statistics of the personnel in and out data according to the state determination result; By arranging an image monitoring module in the passage, capture the dynamic image information of the personnel and perform a secondary statistics of the personnel in and out data according to the formed personnel trajectory; Utilize the data information of the RFID reading module and the image monitoring module, combine with the state rules, periodically fuse and compare the two statistics of the personnel in and out data to calculate the number of people in and out, and output the number of people and the passage in the area in real time.

[0017] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. Through the state machine mechanism, precise management of the four states of personnel can be realized, and complex behavior scenarios such as wandering and abnormal return can be dealt with, ensuring accurate judgment of the internal and external states.

[0018] 2. Introduce an AI camera as an image monitoring module for assistance. Through the cross-verification of the pedestrian flow statistics algorithm and the RFID data, the risk of missed reading and misreading is significantly reduced, and the robustness of the system is improved.

[0019] 3. The signal reading, cooperation caching, and cleaning mechanism formed by the dual RFID reading modules avoid the system load problem caused by repeated signal reading. Even in high-frequency usage scenarios, the system can still operate stably.

[0020] 4. The safety helmet with a passive RFID tag does not require a power supply, is easy to install, and has a low cost, making it suitable for real-time personnel management in large-scale construction sites or shipyard environments. Description of the Drawings

[0021] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0022] Figure 1 It is a flowchart of the method in the present invention.

[0023] Figure 2 It is a flowchart of the state transition of the outer antenna detection in the present invention.

[0024] Figure 3 It is a flowchart of the state transition of the inner antenna detection in the present invention.

[0025] Figure 4 It is a schematic diagram of the positional relationship corresponding to the four states in the present invention.

[0026] Figure 5 It is a schematic diagram of the structure of the system in the present invention.

[0027] Figure 6 For Figure 5 the schematic diagram of the structure from the bottom-up perspective.

[0028] Reference Signs: 1. Detection device; 2. Suspension ring; 3. Control box; 4. Mobile chassis; 5. Display device; 6. Image monitoring module; 7. Supplementary light; 8. Detection module. Detailed Embodiments

[0029] The following will describe in detail the embodiments of the technical solutions of the present invention in conjunction with the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention, so they are only examples and cannot be used to limit the protection scope of the present invention. It should be noted that unless otherwise specified, the technical terms or scientific terms used in this application should have the ordinary meaning understood by those skilled in the art to which the present invention belongs.

[0030] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention.

[0031] In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. In the description of the present invention, the meaning of "a plurality" is more than two, unless otherwise specifically defined.

[0032] In the present application, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed", etc. should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0033] As Figures 1-4 shown, a personnel statistics management method based on a state machine mechanism provided by the present invention includes the following steps: S1. By arranging at least two RFID reading modules in the passage where personnel pass, read the RFID tag information carried by the personnel, and synchronously trigger a state conversion signal.

[0034] The tag information includes information such as the number, name, work area, and department of the personnel. The two RFID reading modules are used to read the RFID tag information of the personnel, that is, the relevant data frames and upload them. The RFID reading module can be implemented in an antenna mode, and the RFID tag is detected wirelessly. Specifically, it is divided into an outer antenna and an inner antenna. Normally, when a person enters the internal area, they pass through the outer antenna and the inner antenna in sequence. On the contrary, when a person leaves the internal area, the person passes through the inner antenna and the outer antenna in sequence. The two antennas work together, effectively utilizing the logical method of triggering the update of the personnel status by the signal, and triggering the state conversion signal through the antenna to ensure the real-time update of the personnel status.

[0035] S2. Transmit the data read by the RFID reading module to the server in real time. The server determines the personnel status according to the preset status rules, and conducts a statistics of the personnel in and out data according to the status determination result.

[0036] Among them, the determination of the personnel status according to the preset status rules is specifically as follows: First, the personnel status is at least abstracted into four states that are sequentially switched along the passage direction. The four states are respectively triggered and switched by the two RFID reading modules.

[0037] As Figure 4 shown, the four states from the first state to the fourth state can be respectively set as: ① Outer state: The personnel are completely in the outer area; ② Outer intermediate state: The personnel start to enter but do not touch the inner antenna; ③ Inner intermediate state: The personnel start to leave but do not touch the outer antenna; ④ Inner state: The personnel are completely inside the inner area.

[0038] If the personnel enter the fourth state in sequence from the first state, or return from the fourth state to the first state in sequence, then a statistics of the personnel in and out data is performed; If the personnel switch from the first state or the fourth state to the second state or the third state, and stay for more than a certain time or return, then it is determined as an abnormal state, recorded and alarmed.

[0039] Each antenna triggers a state transition according to the detected personnel tag signal (data frame). Specifically, it can be divided into two cases: First, when the personnel are in the outer area and the specific state is the outer state, as Figure 2 shown, the state determination is carried out through the following steps: S210. First, judge whether to pass through the outer antenna; if so, trigger the state transition signal and enter S211; if not, the state remains the outer state; S211. Update the state to the outer intermediate state and enter S212; S212. Judge whether to pass through the inner antenna; if so, trigger the state transition signal and enter S213; if not, enter S214; S213. Update the state to the inner state and upload the normal entry information; S214. Keep the state as the outer intermediate state and enter S215; S215. Judge again whether to pass through the outer antenna; if so, trigger the state transition signal and enter S216; if not, enter S217; S216. Update the state to the outer state and upload the information of abnormal entry; S217. Keep the state as the outer intermediate state and judge whether to maintain this state for 30 seconds; if so, issue an alarm; if not, clear the current state and return to S211.

[0040] Second, when the personnel are in the inner area and the specific state is the inner state, asFigure 3 As shown, the status determination is carried out through the following steps: S220. First, determine whether it passes through the inner antenna; if so, trigger the status conversion signal and enter S221. If not, the status remains the inner status. S221. Update the status to the inner intermediate state and enter S222. S222. Determine whether it passes through the outer antenna. If so, trigger the status conversion signal and enter S223. If not, enter S224. S223. Update the status to the outer status and upload the normal departure information. S224. Keep the status as the inner intermediate state and enter S225. S225. Determine again whether it passes through the inner antenna. If so, trigger the status conversion signal and enter S226. If not, enter S227. S226. Update the status to the inner status and upload the abnormal departure information. S227. Keep the status as the inner intermediate state and determine whether to maintain this status for 30 seconds. If so, issue an alarm and enter S228. If not, enter S228. S228. Clear the current status and return to S221.

[0041] Through the cyclic monitoring of the two states, the accuracy and security of personnel flow data are ensured. In the face of complex behavior scenarios, such as incomplete entry and exit, lingering and staying, etc., a state machine mechanism is adopted to manage the status of personnel in real time, effectively avoiding misjudgment and missed reports, and improving the reliability and response speed of the system.

[0042] S3. By arranging an image monitoring module in the passage, capturing the dynamic image information of personnel, and conducting secondary statistics on the personnel entry and exit data according to the formed personnel trajectories. In order to further enhance the reliability of the system, an image monitoring module is introduced. Specifically, an AI camera module is used as an auxiliary tool, which specifically includes the following functions: 1. Pedestrian flow statistics: The AI camera is built-in with a pedestrian flow statistics algorithm, which is periodically compared with the entry and exit number statistics results of RFID to verify the data accuracy.

[0043] 2. Abnormality recognition: When there are data deviations in the RFID signal (such as missed reading or misreading), the data of the AI camera can be used to correct the results.

[0044] When an abnormality recognition situation occurs, data comparison is carried out, such as: Time synchronization: According to the RFID reading time and the camera detection time, match the same time window. Number comparison: Compare the number of people entering and exiting counted by the RFID system and the camera. Trajectory verification: matching camera trajectory data with entry / exit events from RFID tags; Identify anomalies; through the combination of intelligent AI cameras and RFID technology, abnormal identification conditions can be corrected to ensure the accuracy of the identification results.

[0045] 3. Behavior recognition: AI cameras detect human behavior, assist in determining whether there is an abnormal situation of not wearing a helmet, and trigger an alarm. Because it is used in a crowded environment, the ByteTrack multi-target tracking algorithm continues to track the movement trajectory of personnel to ensure that the helmet status of each target is consistent to avoid misjudgment between frames. If the target is detected without a helmet in N consecutive frames, it is judged as abnormal and the alarm device is triggered to alarm. At the same time, abnormal behaviors (wandering, intrusion, climbing) are predefined, and video action recognition models (such as I3D, SlowFast) are used to identify specific actions. If they meet the predefined abnormal behaviors, an alarm is triggered.

[0046] S4. Utilize the data information of the RFID reading module and the image monitoring module, combined with the status rules, periodically integrate and compare the data of two personnel entry and exit to calculate the number of people entering and leaving, and output the number of people in the area and the summary of entry and exit in real time.

[0047] Among them, the data processing of the RFID data detected by the RFID reading module and the camera data of the camera is carried out in the data processing module, and the data calculation is completed through fusion, comparison, and statistics to provide data accuracy.

[0048] Regarding RFID data, it can specifically include: tag ID, antenna location, read timestamp, signal strength (RSSI); Regarding camera data, it may include: target ID (generated by tracking algorithm), target trajectory (motion path), entry and exit direction, and timestamp.

[0049] S401. First, the detection data of the RFID system and the detection results of the AI ​​camera are integrated and compared in multiple dimensions using data fusion technology, so as to realize data verification, correction and abnormality identification; The specific steps include: Time synchronization: The data of the RFID reading module and the image monitoring module are aligned through the timestamp to perform time synchronization operations. By defining a time window Δt (such as ±1 second), the RFID reading records in the same time window are matched with the camera detection results: |T rfid −T camera ∣≤Δt; Among them, T rfid is the time when the RFID reads the record, Tcamera It is the time of the camera detection result.

[0050] Spatial matching: Map the position of the person carrying the RFID tag to the plane coordinate system monitored by the image monitoring module, and compare the trajectory ranges detected by the RFID reading module and the image monitoring module. It is required that the two trajectories are within the matching distance threshold: Distance(P rfid ,P camera )≤D threshold ; P rfid : The center point of the RFID antenna; P camera : The average position of the camera trajectory; D threshold : The matching distance threshold.

[0051] Data association: Correlate the data detected by the RFID reading module and the image detection module through features in time and space, such as the time window, spatial proximity of the target position, and different dimensional features such as matching direction and trajectory form.

[0052] Furthermore, the periodic fusion and comparison for statistically counting the data of personnel entering and leaving twice also includes: S402. Secondly, use the data comparison logic to compare and correct the data collected by the RFID reading module and the camera; Specifically, it includes the following steps: Number comparison: Compare the number of people counted by the RFID reading module and the image monitoring module. If they are consistent, adopt it; Trajectory verification: Compare the trajectory directions detected by the RFID reading module and the image monitoring module. If they are consistent, adopt it. For example, when the state judged by the RFID reading module is the in-state, the camera trajectory should also show that the target moves from the entrance to the inside. Otherwise, it is first marked as an abnormal state; After that, correct the data for the abnormal state. The abnormal state includes the following two situations: If the camera detects the target trajectory but the RFID reading module has no corresponding tag record, insert a virtual RFID record to complete the missing reading record; If the direction determined by the RFID reading module according to the state machine mechanism does not match the camera trajectory direction, take the camera direction as the standard and correct the RFID-determined direction.

[0053] Take the direction detected by the image monitoring module as the standard.

[0054] S403. Statistically calculate the number of people entering and leaving, and output the summary of the number of people and the situation of entering and leaving in the area in real time; Specifically, it includes the following steps: 1. Capture video frames: Decode the video to obtain image frames. Specifically, collect the video stream through the camera, use the RTSP or USB interface to obtain the video stream in real time, and decode the video stream into single-frame images; Frame processing: Scale the frame to the model input size (640x640), then perform normalization to scale the pixel values to the range [0,1]. After adjusting the frame data, use Gaussian filtering to reduce the impact of noise on the frame data and extract the monitored area data. Eliminate the irrelevant background and perform lightweight processing on the data; 2. Perform object detection: Input the image frame into the trained model to obtain the object bounding boxes and confidence scores. Specifically, load the lightweight object detection model YOLOv5n, pre-trained on the COCO dataset and adjusted to adapt to the construction site scenario, where the adapted scenario can be set differently according to the actual situation. Then deploy the edge computing device to optimize the model inference speed through TensorRT. Input the preprocessed image frame into the trained model, and after processing, output the object bounding boxes, class labels, and confidence scores, and perform confidence score filtering and bounding box filtering on the output results; Among them, the confidence score filtering and bounding box filtering of the output results are specifically implemented through the following steps: Based on the ByteTrack multi-object tracking algorithm, by processing the low-confidence detection results (referred to as "byte" tracks), the tracking performance in crowded scenes is improved.

[0055] For the results obtained from the input object detection, specifically including the bounding box positions [x min , y min , x max , y max and the confidence score [0,1], where according to the confidence score, the detection results can also be divided into: High-confidence objects (conf≥τ h ): Main objects; Low-confidence objects (τ l ≤conf<τ h ): Potential objects; Among them, τ h and τ l are the confidence score thresholds.

[0056] 3. Perform multi-object tracking: Based on the multi-object tracking algorithm, screen the active tracks for matching and updating the track status; Update the total number of people in the area according to the entry and exit directions of the tracks, and periodically output the statistical results per unit time.

[0057] Specifically, after obtaining the object bounding boxes and confidence scores, screen the active tracks through the following steps: Classify the objects into high-confidence objects and low-confidence objects according to the confidence score, that is, distinguish the main objects and potential objects to achieve priority stratification; Use Kalman filter to predict the next position of the trajectory. Subsequently, according to the method of high-confidence matching, calculate the IOU between the existing trajectory and the high-confidence target. IOU is the intersection over union, which in this embodiment is: the ratio of the intersection to the union of the trajectory box and the detection box. The trajectory box corresponds to the predicted trajectory, and the detection box corresponds to the existing trajectory. Subsequently, apply the Hungarian algorithm for trajectory-target matching. If the detection match is successful, that is, IOU > δ (δ = 0.3), update the trajectory state, including position, speed, and confidence, and mark the trajectory state as a high-confidence trajectory; If it is not successful, send the trajectory to low confidence for matching. Low-confidence matching mainly attempts to associate the unmatched trajectory with the low-confidence target, mainly by recalculating the IOU for the unmatched trajectory and the confidence detection box, and applying the Hungarian algorithm for matching again. If the low-confidence target still fails to match the existing trajectory, it is initialized as a new trajectory. Otherwise, if the match is successful, update the trajectory information but do not increase the confidence, and mark the trajectory state as a low-confidence trajectory.

[0058] In terms of trajectory management, if a trajectory fails to be successfully matched for multiple consecutive frames, it is marked as "invalid" and regarded as an invalid trajectory. The "invalid" trajectory is removed after 30 frames.

[0059] Take the trajectory of the high-confidence target as the active trajectory for output (which specifically includes information such as ID, position, speed, etc.).

[0060] Furthermore, updating the total number of people in the area according to the entry and exit directions of the trajectory specifically includes the steps: Define at least two virtual judgment lines in the monitoring area. The two judgment lines are located at the entrance and exit positions of the monitoring area, specifically on the central axis of the entrance and exit. Set the coordinates of the judgment lines as (x start , y start ) and (x end , y end ), so as to frame the monitoring area; By calculating the center point of the area delimited by the virtual judgment lines, the center point position is: (x c , y c ) = (x min + x max ) / 2, (y min + y max ) / 2; Based on the center point, perform crossing detection. Mark the entry and exit directions of the personnel according to the direction in which the center point of the personnel crosses the judgment line, and update the total number of people in the area according to the entry and exit directions of the trajectory. When the center point of the target moves from the outside of the judgment line to the inside, it is marked as entering, and when it moves in the opposite direction, it is marked as leaving.

[0061] Update the total number of people in the area according to the entry and exit directions of the trajectory. Among them, the current number of people = the number of people entering - the number of people leaving, and the calculation result is output periodically, and the statistical result is output every minute.

[0062] To avoid the system load problem caused by repeated reading of data by the RFID reading module, a cache cleaning mechanism is established. The RFID reading module regularly clears redundant data according to the following measures through the cache cleaning mechanism to ensure real-time data update; The RFID reading module reads the same person carrying an RFID tag once within the first time threshold, and this time threshold can be set to 2 seconds, that is, the antenna reads the same person only once within 2 seconds, avoiding an excessive reading frequency from increasing the system burden.

[0063] At the same time, the cache records in the unfinished state will be automatically cleared regularly to keep the system stable.

[0064] Set a regular cleaning task, and set this time period to 30 seconds. Check the cache records every 30 seconds: For data determined to be in an abnormal state exceeding the time threshold, if the status of the target is not the internal status or the external status, and the cache time exceeds the set maximum expiration time (such as 30 seconds), it is considered that the data of this target has expired. Subsequently, an alarm is issued and the record is deleted, and then the status is refreshed again; Clean the cache for the data of the completed personnel entry and exit statistics. If the signal of a certain target is not updated within a certain time window (for example, no signal is detected within 10 seconds), then judge its status according to the status mechanism when the signal last appeared. If the status when it last appeared shows a completed state, that is, a normal entry state or a normal exit state, and after being registered in the statistical personnel information, this record information will be cleared.

[0065] Configure the RFID tag on the person's clothing or personal belongings to ensure that the tag moves synchronously with the person. In this embodiment, the RFID tag is configured in safety equipment such as safety helmets to ensure that personnel wear safety helmets at all times in the factory environment.

[0066] Use a multi-target tracking algorithm to determine the state consistency of the target personnel between multiple frames, and automatically trigger an alarm and record for the personnel and abnormal states determined to be not carrying an RFID tag.

[0067] Specifically: For the processed video frames, input the image frames into the trained multi-class object detection model for detecting personnel and safety helmets. Among them, the label of the personnel is "person", and the label of the safety helmet is "helmet". Input the processed frames into the model, and output the detection box, class label and confidence of each target; Use the intersection over union (IoU) to determine the relationship between a person and a safety helmet. If the "helmet" detection box is located in the head area of the "person" box and the IoU > δ (threshold δ = 0.5), it is considered that the person is wearing a safety helmet. If no safety helmet associated with the person box is detected, it is marked as the "not worn" state. Use the multi-object tracking algorithm (ByteTrack) to assign a unique ID to each detection box. Track the state consistency of the tracking target between multiple frames to avoid misjudgment in a single frame. If a certain target is detected as not wearing a safety helmet for N consecutive frames, it is confirmed as the "not worn" state and an alarm is triggered.

[0068] As Figures 4-6 shown, the present invention also discloses a personnel statistics management system based on a state machine mechanism, which includes: RFID tags, RFID reading devices, an image monitoring module, and a server.

[0069] The RFID reading device and the image monitoring module are installed in the passage of a detection device. The detection device can adopt a barrier-free passage design such as a booth with a passage. A control box is configured outside the detection device, and a controller and a power module are installed in the control box. The power module is used to provide power for the devices in the booth. The controller is connected to the server and the devices in the booth to implement the processing and sending of local control instructions.

[0070] The RFID tags are for personnel to carry with them. Specifically, refer to Embodiment 1 and configure them in safety equipment such as safety helmets for convenient identification by the camera.

[0071] The detection modules are arranged at intervals in the passage of the detection device. The detection modules can form a detection area in the passage. The detection modules are used to identify and record the tag information of the personnel carrying RFID tags. Specifically, the detection modules can adopt detection antennas that cooperate with RFID and are installed on the top of the passage in the booth. It consists of an RFID reading device (RFID antenna + reading and writing module) for detecting RFID tags to achieve non-contact identification. Two detection antennas arranged at intervals cooperate to detect the entry and exit of personnel. The data of personnel entry and exit are recorded only when the personnel completely pass through the two detection antennas. The two antennas can conveniently detect the direction of personnel entry and exit.

[0072] The RFID reading device includes at least two detection modules. The detection modules are used to form a state detection area. When the RFID tag passes through the state detection area, the RFID reading device is used to read the tag information of the RFID tag and simultaneously trigger a state conversion signal.

[0073] The two detection modules are arranged at intervals along the traveling direction of the channel, and the channel is sequentially divided into at least four regions. Two regions are formed on both sides of one detection module, and four regions can be divided by using two detection modules. The controller is used to add different status tags correspondingly when the person carrying the RFID tag is in different regions, and cooperate with the detection module to abstract four states of the person carrying the RFID tag. Corresponding to the four regions from the outside to the inside, the four states are respectively: Outer state: The person is completely outside the sentry box area, and this state corresponds to the outermost first region; Outer intermediate state: The person starts to enter but does not touch the inner antenna, and this state corresponds to the second region from the outside to the inside; Inner intermediate state: The person starts to leave but does not touch the outer antenna, and this state corresponds to the third region from the outside to the inside; Inner state: The person is completely inside the sentry box area, and this state corresponds to the fourth region from the outside to the inside.

[0074] Through the cooperation of the detection module and the RFID tag, combined with the state machine mechanism, the real-time management of the personnel state is realized, and it is better to cope with complex behavior scenarios, such as incomplete entry and exit, wandering and staying, etc.

[0075] The image monitoring module is implemented by an AI camera, which is used to synchronously capture the dynamic image information of the personnel, and count the personnel entry and exit information according to the data information. There are two cameras, which are respectively aimed at the entrances and exits of the sentry box. The cameras can be used for functions such as safety equipment detection, passenger flow detection and face recognition.

[0076] In this embodiment, a display device is further included. The display device is connected to the server and is used to display the data information processed by the controller. The display device can be implemented by a display screen, etc., and can be configured on the sentry box. The data information for display can include the number of people, the number of abnormal detections, the detection duration, the number of alarms, etc.

[0077] To improve the detection effect of the camera, a supplementary light is set in the channel of the detection device, which is used to assist the image monitoring module to collect information, and assist the AI camera to detect personnel in an environment with insufficient light, so as to ensure the acquisition of high-quality image information.

[0078] The bottom of the detection device is detachably connected with a mobile chassis. The mobile chassis can adopt a wheeled or tracked chassis, etc., and is set at the bottom of the sentry box. The sentry box can be moved to the target position, without being limited to a certain specific location, and can be used as needed, improving the mobility of the detection device.

[0079] Lifting rings are provided at the tops on both sides of the detection device. In complex road conditions where the mobile chassis has difficulty moving, mechanical equipment such as lifting tools and cranes can be used to lift the entire guard booth through the lifting rings and transport the guard booth to the target location, providing a second means for the handling and deployment of the guard booth detection device.

[0080] The server is used to receive the data transmitted by the RFID reading device and the image monitoring module, and the server is further configured to perform the following steps: By arranging at least two RFID reading modules in the passage where people pass, the RFID tag information carried by people is read, and the status conversion signal is synchronously triggered; The data read by the RFID reading module is transmitted to the server in real time. The server determines the personnel status according to the preset status rules and conducts a primary statistics of the personnel in and out data based on the status determination result; By arranging an image monitoring module in the passage, the dynamic image information of people is captured, and a secondary statistics of the personnel in and out data is conducted according to the formed personnel trajectory; Using the data information of the RFID reading module and the image monitoring module, combined with the status rules, the in and out data of personnel are periodically fused and compared to calculate the number of people in and out, and the number of people and the passage in the area are output in real time.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A personnel statistics management method based on a state machine mechanism, characterized in that: include: By arranging at least two RFID reading modules in the passage where people pass through, the RFID tag information carried by the people is read and the state conversion signal is synchronously triggered; The data read by the RFID reading module is transmitted to the server in real time. The server determines the personnel status according to the preset status rules and performs personnel entry and exit data statistics based on the status determination results. By placing image monitoring modules in the channel, dynamic image information of personnel is captured, and secondary personnel entry and exit data statistics are performed based on the formed personnel trajectory; Utilizing the data information from the RFID reading module and the image monitoring module, combined with status rules, the data of two personnel entry and exit are periodically fused and compared to calculate the number of people entering and leaving, and the number of people in the area and the summary of entry and exit situations are output in real time.

2. The personnel statistics management method according to claim 1, characterized in that: According to the preset status rules, the personnel status is determined as follows: First, the personnel status is abstracted into at least four states that are switched in sequence along the passage direction, and the four states are triggered to switch by the two RFID reading modules respectively; If a person enters the fourth state from the first state, or returns to the first state from the fourth state, a person entry and exit data statistics is performed; If a person switches from the first state or the fourth state to the second state or the third state and stays there for more than a certain period of time or returns, it is determined to be an abnormal state, recorded and an alarm is issued.

3. The personnel statistics management method according to claim 1, characterized in that: The periodic fusion comparison and statistics of two personnel entry and exit data specifically include: Align the data of RFID reading module and image monitoring module through timestamp to achieve time synchronization; Map the position of the person carrying the RFID tag to the plane coordinate system monitored by the image monitoring module, and compare the trajectory range detected by the RFID reading module and the image monitoring module; The data detected by the RFID reading module and the image detection module are associated through the characteristics of time and space.

4. The personnel statistics management method according to claim 3, characterized in that: The periodic fusion comparison and statistics of two personnel entry and exit data also include: Compare the number of people counted by the RFID reading module and the image monitoring module, and adopt them if they are consistent; Compare the trajectory directions detected by the RFID reading module and the image monitoring module. If they are consistent, adopt them. Otherwise, the direction detected by the image monitoring module shall prevail.

5. The personnel statistics management method according to claim 4, characterized in that: The periodic fusion, comparison and statistics of the two personnel entry and exit data also include: Decode the video to obtain image frames, input the image frames into the training model, and obtain the target bounding box and confidence; Based on the multi-target tracking algorithm, active trajectories are screened to match and update trajectory status; The total number of people in the area is updated according to the entry and exit directions of the trajectory, and the statistical results per unit time are output periodically.

6. The personnel statistics management method according to claim 4, characterized in that: After obtaining the target bounding box and confidence, the following steps are performed to filter active tracks: According to the confidence level, the targets are divided into high-confidence targets and low-confidence targets; Use Kalman filtering to predict the next position of the trajectory, calculate the intersection-over-union ratio of the current trajectory and the high-confidence target, and if the detection match is successful, mark the trajectory state as a high-confidence trajectory; If it is unsuccessful, the trajectory is sent to the low confidence level for matching, and the intersection-over-union ratio between the trajectory and the low confidence target is calculated. If the detection and matching is successful, the trajectory status is marked as a low confidence trajectory. If unsuccessful, it is marked as a failed track.

7. The personnel statistics management method according to claim 4, characterized in that: The updating of the total number of people in the area according to the entry and exit directions of the trajectory specifically includes the following steps: Define at least two virtual judgment lines in the monitoring area, calculate the center point of the area demarcated by the virtual judgment line, perform crossing detection based on the center point, mark the entry and exit direction of the person according to the direction in which the person's center point crosses the judgment line, and update the total number of people in the area according to the entry and exit direction of the trajectory.

8. The personnel statistics management method according to claim 2, characterized in that: The RFID reading module uses the cache cleaning mechanism to regularly clear redundant data according to the following measures to ensure real-time data updates; The RFID reading module reads the same person carrying the RFID tag once within the first time threshold; For data that exceeds the time threshold and is judged to be in an abnormal state, the cache of the data that has completed the statistics of personnel entry and exit is cleaned up.

9. The personnel statistics management method according to claim 2, characterized in that: Place RFID tags on people's clothing or personal belongings to ensure that the tags move synchronously with the people; The multi-target tracking algorithm is used to determine the consistency of the target person's status between multiple frames. For people detected as not carrying RFID tags and abnormal status, alarms are automatically triggered and recorded.

10. A personnel statistics management system based on a state machine mechanism, characterized in that: include: An RFID tag, wherein the RFID tag is used for a person to carry with him; An RFID reading device, comprising at least two detection modules, wherein the detection modules are used to form a state detection area, and when the RFID tag passes through the state detection area, the RFID reading device is used to read the tag information of the RFID tag and trigger a state transition signal; Image monitoring module, used to synchronously capture dynamic image information of personnel and count personnel entry and exit information based on data information; A server is used to receive data transmitted by the RFID reading device and the image monitoring module, and the server is further configured to perform the following steps: By arranging at least two RFID reading modules in the passage where people pass through, the RFID tag information carried by the people is read and the state conversion signal is synchronously triggered; The data read by the RFID reading module is transmitted to the server in real time. The server determines the personnel status according to the preset status rules and performs personnel entry and exit data statistics based on the status determination results. By placing image monitoring modules in the channel, dynamic image information of personnel is captured, and secondary personnel entry and exit data statistics are performed based on the formed personnel trajectory; Utilizing the data information from the RFID reading module and the image monitoring module, combined with the status rules, the data of two personnel entry and exit are periodically fused and compared to calculate the number of people entering and leaving, and the number of people and channels in the area are output in real time.

Images