Millimeter wave radar dormitory personnel activity monitoring method and system for smart campus

Abstract

The invention discloses a millimeter wave radar dormitory personnel activity monitoring method and system for a smart campus. Intelligent management of dormitory personnel activities is realized through a non-contact monitoring technology. The system obtains high-precision real-time point cloud data through a 60G millimeter wave radar, effectively filters environmental noise in combination with coordinate conversion and a dynamic point cloud segmentation algorithm, realizes stable and reliable personnel detection in a complex dormitory environment, and accurately distinguishes seat and bed body area activities through accurate area division. According to the method, effective and intelligent multi-target tracking is realized by adopting sliding window statistics and a self-adaptive threshold triggering mechanism. The system only processes point cloud data, does not collect biological characteristic information, fully guarantees the privacy of students while guaranteeing the monitoring effect, can achieve 24-hour automatic monitoring, greatly reduces the workload of manual inspection, supports the extended functions of attendance statistics, safety early warning, behavior analysis and the like, provides reliable technical support for constructing a safe and intelligent campus environment, and has a wide application prospect. And the method has remarkable social benefits and application values.

Description

Technical Field

[0001] The present invention relates to the field of millimeter-wave radar sensing technology, and in particular to a millimeter-wave radar dormitory personnel activity monitoring method and system for smart campuses. Background Art

[0002] In recent years, with the rapid advancement of smart campus construction, intelligent management of student dormitories has become a key issue in the development of educational informatization. Traditional dormitory management primarily relies on manual inspections, access control systems, and video surveillance. These methods have significant limitations: manual inspections are inefficient and lack 24 / 7 coverage; access control systems only record entry and exit information but cannot monitor indoor activities; and video surveillance faces ethical concerns regarding privacy, particularly during nighttime rest periods, which may infringe on students' personal privacy. Furthermore, existing technologies exhibit a lag in responding to emergencies, making it difficult to detect sudden incidents within dormitories in a timely manner. Millimeter-wave radar technology offers a new approach to addressing these issues. Its operating principle is to transmit high-frequency electromagnetic waves and receive signals reflected from the human body. It can not only penetrate common dormitory objects such as curtains and thin partitions, but also accurately capture micro-motion signatures through the Doppler effect.

[0003] The current technical challenges in dormitory activity monitoring lie primarily in three areas: First, existing infrared sensors are susceptible to interference from ambient temperature, resulting in a high false alarm rate; second, bed exit detection systems based on pressure pads are complex to deploy and cannot identify specific activity types; and third, video surveillance presents privacy concerns. Millimeter-wave radar, however, utilizes micro-Doppler effect analysis to provide strong anti-interference capabilities. It can adapt to the confined and complex environments of dormitories and discern subtle human movements without any interference. It can distinguish between regular activities like studying and sleeping, while also detecting abnormal events. Compared to technologies like UWB and Wi-Fi sensing, millimeter-wave radar offers higher spatial resolution and can maintain relatively stable target tracking capabilities even in mixed-use scenarios. Importantly, this technology only collects human outline and movement information, without involving biometric recognition, fundamentally addressing privacy concerns.

[0004] Therefore, a millimeter-wave radar dormitory personnel activity monitoring method and system for smart campuses is needed. Summary of the Invention

[0005] In view of this, the present invention aims to provide a millimeter-wave radar dormitory personnel activity monitoring method and system for smart campuses. Through the non-contact sensing capability of 60GHz millimeter-wave radar, it can accurately monitor the location, activities, and behavior of dormitory personnel while fully protecting students' privacy. This method can not only track personnel dynamics in real time and identify abnormal behavior, but also automatically generate extended functions such as student dormitory attendance statistics and work and rest pattern analysis based on long-term monitoring data, providing multi-dimensional data support for smart campus management, including security warnings, behavior analysis, and dormitory attendance. It solves the privacy leakage problems of traditional video surveillance and the defects of the single monitoring function and insufficient accuracy of technologies such as infrared sensing, and realizes modern dormitory management that is non-sensitive, intelligent, and multifunctional.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The millimeter-wave radar dormitory personnel activity monitoring method for smart campuses provided by the present invention includes the following steps:

[0008] S1. Obtain 3D point cloud data streams through the millimeter-wave radar deployed on the roof of the dormitory; perform preliminary processing on the real-time point cloud data, including coordinate conversion, boundary filtering, and dynamic point cloud segmentation;

[0009] S2. Extract and simulate the monitoring areas in the dormitory, divide the key monitoring areas into regional blocks, and combine the obtained point cloud dataset to divide them into seating areas and bed areas;

[0010] S3. Performing human presence detection on the processed point cloud data through sliding window statistics and threshold triggering;

[0011] S4. Perform multi-target tracking on the point cloud data after personnel presence detection.

[0012] Furthermore, the method further comprises the following steps:

[0013] S5. The real-time monitoring data obtained through the above steps is used to continuously monitor and analyze the daily behavior of dormitory residents, such as entering and exiting the dormitory, getting in and out of bed, etc., and the analysis results are combined with timestamp information accurate to milliseconds to generate a structured monitoring data report.

[0014] Furthermore, the preliminary processing in step S1 is performed in the following manner:

[0015] S11 coordinate conversion: uses a transformation algorithm to convert the original radar polar coordinates into a Cartesian coordinate system (x, y, z);

[0016] S12 boundary filtering: Based on the pre-set dormitory three-dimensional space boundary parameters, a spatial filtering model is established to automatically eliminate invalid point cloud data outside the monitoring area;

[0017] S13 dynamic point cloud segmentation: By calculating the velocity components (vx, vy, vz) of each point cloud in the three coordinate axis directions, a reasonable velocity threshold is set to distinguish between static background points and moving dynamic target points.

[0018] Furthermore, the division of the monitoring area in step S2 is performed in the following manner:

[0019] S21 determines the dormitory space measurement information and the installation location information of the millimeter wave radar in the dormitory, including height and angle.

[0020] S22 performs spatial distribution analysis on the collected point cloud data and divides the dormitory area into multiple functional sub-areas, such as the seating area for study monitoring and the bed area for rest monitoring.

[0021] Furthermore, the person presence detection in step S3 is performed in the following manner:

[0022] S31 uses a sliding time window technique to count the number of point clouds and spatial distribution density characteristics of the most recent frames in each area in real time based on the division of the dormitory area in step S2, and uses historical monitoring data to perform data smoothing to eliminate instantaneous noise interference;

[0023] S32 sets relevant thresholds and performs experimental data calibration, setting static clutter point count frames when no one is present and dynamic point count frames when someone is present. The cumulative number of points and point cloud density in the most recent consecutive frames are counted through a sliding window. Based on the comprehensive average judgment of the experimental data, if the threshold is exceeded, it is detected as a person.

[0024] Furthermore, in step S4, multi-target tracking is performed on the point cloud data after the presence of personnel is detected. First, the next state of the existing tracked target is predicted and estimated using the Kalman filter algorithm. Then, the Mahalanobis distance between the predicted target and each point cloud cluster in the current frame is calculated to evaluate their association possibility. Finally, a probability-weighted data fusion method is used to update the target state information using the successfully matched observation data.

[0025] Furthermore, the multi-target cluster tracking of personnel in step S4 is performed in the following manner:

[0026] S41 Kalman filter prediction: Establish a linear motion state equation to predict the position coordinates, motion speed and other state parameters of each tracking target at the next moment, and update the uncertainty covariance matrix of the state estimation.

[0027] S42 Data Association: Matches radar point clouds to target trajectories through a two-stage data association process: global matching followed by local compensation. First, existing targets are associated, and then unmatched points are reassigned based on distance thresholds.

[0028] S43 Data Management: Use the DBSCAN clustering algorithm to spatially cluster unassociated radar point clouds, initializing point clusters that meet density conditions and are far away from existing targets as new tracking targets, achieving new detection and lifecycle management of dynamic targets;

[0029] S44 Target Update: This update module updates the motion state of the associated target through the extended Kalman filter state update and JPDA algorithm, and manages the target life cycle.

[0030] Furthermore, the effective management of dormitory personnel in step S5 is performed in the following manner:

[0031] Based on the analysis and processing of millimeter-wave radar point cloud data, the location of personnel can be detected and tracked. Through long-term monitoring of personnel, the dormitory residents' daily return time, on-site time, sleep time, etc. can be understood and supervised, forming regular experience and effectively managing student dormitories without contact.

[0032] The millimeter-wave radar dormitory personnel activity monitoring system for smart campuses provided by the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The above method is implemented when the processor executes the program.

[0033] The beneficial effects of the present invention are:

[0034] The present invention provides a millimeter-wave radar dormitory personnel activity monitoring method and system for smart campuses. The method uses millimeter-wave radar to achieve non-contact and effective monitoring of dormitory personnel activities. First, it has significant advantages in monitoring accuracy and reliability. By adopting a 60G millimeter-wave radar, the system can obtain high-precision real-time point cloud data. Combined with coordinate transformation and dynamic point cloud segmentation algorithms, the system can effectively filter out environmental noise and achieve stable and reliable personnel detection in complex dormitory environments. In particular, for special structures such as bunk beds, through precise area division, the system can accurately distinguish between activities in the seat area and the bed area, solving the problem of vertical space resolution of traditional monitoring methods. Secondly, the system realizes intelligent multi-target tracking and recognition. Using sliding window statistics and adaptive threshold triggering mechanism, the system can accurately detect the presence of people and avoid false alarms caused by environmental interference. Through the multi-target clustering tracking algorithm, it is possible to maintain a high tracking accuracy in scenarios where multiple people are active at the same time, greatly improving the level of dormitory safety protection.

[0035] This method not only realizes the real-time monitoring of the basic activities of dormitory personnel, but also generates data reports with timestamps, providing an objective basis for dormitory management. In addition, it also has unique advantages in terms of privacy protection. Unlike traditional video surveillance, the system only processes point cloud data and does not collect any biometric information. It not only meets management needs but also fully protects students' privacy. Compared with traditional solutions, the deployment is effectively reduced and there is no need to transform the dormitory infrastructure. Through 24-hour automated monitoring, the workload of manual inspections can be reduced by more than 80%. The monitoring data can be seamlessly connected to the campus smart management platform, supporting a variety of extended functions such as attendance statistics and behavior analysis, providing reliable technical support for creating a safe and intelligent campus environment.

[0036] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to make the purpose, technical solutions and beneficial effects of the present invention more clear, the present invention provides the following drawings for illustration.

[0038] Figure 1 This is the overall algorithm flow chart in this embodiment.

[0039] Figure 2 This is a diagram of the installation of the millimeter-wave radar in the dormitory in this embodiment.

[0040] Figure 3 This is a schematic diagram of the dormitory area layout in this embodiment.

[0041] Figure 4 This is a schematic diagram of the dormitory monitoring area division block diagram in this embodiment.

[0042] Figure 5 This is a flow chart of the millimeter wave radar personnel detection algorithm in this embodiment.

[0043] Figure 6 This is a flow chart of the personnel tracking algorithm in this embodiment.

[0044] Figure 7 This is a real-time monitoring diagram of the two-dimensional and three-dimensional data of the millimeter wave point cloud in this embodiment.

[0045] Figure 8 This is a schematic diagram of the millimeter-wave radar personnel detection effect in this embodiment.

[0046] Figure 9 This is a waveform diagram of dormitory personnel activities within 24 hours in this embodiment. DETAILED DESCRIPTION

[0047] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0048] Example 1

[0049] like Figure 1 The figure shows the overall algorithm flow chart of the present invention. First, the raw ADC data of the 60G millimeter-wave radar is collected, FFT and preliminary processing are performed, and then the monitoring area is extracted to implement algorithms such as personnel detection and target tracking. Finally, based on the above data, effective 24-hour contactless management of dormitory personnel activities can be achieved. The millimeter-wave radar dormitory personnel activity monitoring method for smart campuses provided in this embodiment mainly includes the following steps:

[0050] S1. Real-time point cloud data within the dormitory area is acquired through a 60G millimeter-wave radar, and a series of preliminary processing, i.e., preprocessing, is performed in the following manner:

[0051] S11 coordinate conversion: convert the original radar polar coordinates into Cartesian coordinate system (x, y, z);

[0052] S12 Boundary filtering: Use boundary filtering function to eliminate invalid points in non-monitoring areas of the dormitory;

[0053] S13 dynamic point cloud segmentation: Separate static background and dynamic targets based on velocity threshold (vx,vy,v).

[0054] The details are as follows:

[0055] Install the millimeter wave radar above the left side of the dormitory door, 2.55m from the ground. Figure 2 As shown in the figure. During the signal processing process, the intermediate frequency signal is processed using fast Fourier transform technology, converting the signal from the time domain to the frequency domain, thereby accurately extracting the signal's individual frequency components. A two-dimensional fast Fourier transform is then performed to expand the signal from the one-dimensional frequency domain to the range-Doppler domain. During the initial data processing phase, the millimeter-wave radar's detection range is set, and all collected point clouds are screened for validity. Coordinate conversion and dynamic point cloud segmentation are then performed.

[0056] Since millimeter-wave radar measurement is based on the radar sphere coordinate system, and personnel information detection is based on the ground coordinate system, the point cloud coordinates need to be transformed. Specifically:

[0057] Millimeter-wave radar usually uses a spherical coordinate system to describe the position of the target. The coordinates in the spherical coordinate system are usually expressed as Where r is the radial distance from the target to the radar, θ is the azimuth, which represents the angle of the target in the horizontal plane relative to the front of the radar. is the pitch angle, which indicates the angle of the target in the vertical plane relative to the horizontal plane.

[0058] From the radar sphere coordinate system The conversion formula to the ground Cartesian coordinate system (x, y, z) is as follows:

[0059] x=r·cos(φ)·sin(θ), y=r·cos(φ)·cos(θ), z=r·sin(φ)

[0060] S2. Extract the monitoring area in the dormitory based on the obtained point cloud dataset and divide the key monitoring areas into regional blocks. The dormitory monitoring area is divided into two core functional areas: the bed area and the seat area based on the characteristics of human activity. The division of the monitoring area in step S2 is performed as follows:

[0061] S21 determines the dormitory space measurement information and the installation location information of the millimeter wave radar in the dormitory, including height and angle.

[0062] S22 combines the collected point cloud data to determine the seat monitoring area and the bed monitoring area to form multiple areas. In this embodiment, the dormitory area is divided into four areas: bed No. 1, table No. 1, bed No. 2, and table No. 2.

[0063] like Figure 4 As shown, Figure 4 This is a schematic diagram of the dormitory monitoring area division block diagram in this embodiment, specifically:

[0064] According to the establishment of the radar measurement coordinate system, only the y and z planes need to be divided into regions in a two-dimensional perspective. Combined with the spatial density distribution and height characteristics of the millimeter-wave radar point cloud, the z-axis height characteristics are used to distinguish the high-altitude bed and the low-altitude seat. For horizontal positioning, the left and right berths and adjacent seat areas are divided by the y-axis coordinates.

[0065] The dormitory monitoring area is divided as shown in Table 1:

[0066] Unit (m) Bed No. 1 Table No. 1 Bed No. 2 Table number two Y [-1.5,0.5] [-1.5,0.5] [0,5,2.5] [0.5,2.5] Z [1.5,3.5] [0,1.5] [1.5,3.5] [0,1.5]

[0067] S3. For the processed point cloud data, human presence detection is performed through sliding window statistics and threshold triggering. First, the collected continuous frame point cloud data is pre-processed and spatially divided. Then, a fixed-length sliding window is used to perform time series statistical analysis of the number of point clouds in each monitoring area for a specific number of frames. The dynamic change characteristics of the point cloud density in the window are calculated, and the two are compared with the threshold to achieve the detection of human presence.

[0068] S31 According to the division of the dormitory area in step S2, the number of point clouds in each area and the point cloud density in the last N frames are counted in real time, and the instantaneous noise is smoothed through historical data;

[0069] S32 Set the related threshold and calibrate the experimental data, set the static clutter point number frame when there is no one and the dynamic point number frame when there is someone, count the cumulative point number and point cloud density in the last several frames through the sliding window, and judge according to the experimental data comprehensive average, if it exceeds the threshold, it is detected as someone; In this embodiment, the following method is used: the static clutter point number when there is no one is usually <10 / frame, and the dynamic point number when there is someone can reach 50+ / frame. Count the cumulative point number and point cloud density in the last 100 frames through the sliding window, and judge according to the experimental data comprehensive average, if it exceeds the threshold, it is detected as someone, which balances the sensitivity and noise resistance.

[0070] The algorithm flow is as shown in Figure 5 , which is the personnel detection algorithm flowchart of the millimeter wave radar in this embodiment, and specifically: Figure 5

[0071] S301, initialize the monitoring area, divide the monitoring space into uniform cubic grids, each grid unit as a detection area ri, each area ri associated with a sliding window Wi and a threshold Ti.

[0072] S302, initialize the monitoring area, divide the monitoring space into uniform cubic grids, each grid unit as a detection area ri, each area ri associated with a sliding window Wi and a threshold Ti.

[0073] S303, for the t-th frame point cloud P t , count the number of points in each area ri The formula is:

[0074]

[0075] Among them, is an indicator function, which returns 1 when the coordinates of the point p satisfy min(r i )≤p≤max(r i ), otherwise 0.

[0076] S304, after completing the single-frame area point number statistics, it is necessary to accumulate the point number change of each area in the last N frames. The specific implementation includes:

[0077] Assign a queue Wi with a length of N to each monitoring area ri:

[0078]

[0079] Among them, ​Represents the number of points in region ri in the t-th frame.

[0080] Then accumulate the points within the specified number of frames, which is achieved through the following formula:

[0081]

[0082] S305, calculate the point cloud density of each monitoring area within the specified frame range, the current density of area ri is the mean of the points in the window:

[0083]

[0084] S306: After completing the sliding window statistics for each monitoring area, the accumulated point cloud points in the specified frame are calculated. and point cloud density Comparison is made with the preset dual thresholds: first, a check is made to see whether the cumulative number of points exceeds the basic threshold T_count, and at the same time, the point cloud density is verified to see whether it reaches the density threshold T_density. Only when these two conditions are met will the system determine that there are people in the area; otherwise, it will be marked as unmanned. This dual verification mechanism effectively avoids false detections caused by sparse distribution of point clouds or transient noise. At the same time, density normalization ensures detection fairness in areas of different sizes, significantly improving the accuracy of the system.

[0085] S4: Multi-target tracking is performed on the point cloud data after the presence of personnel is detected. The state of the existing target is predicted using Kalman filtering, and the Mahalanobis distance between each predicted target and the current frame point cloud cluster is calculated to evaluate the possibility of association. The observation data with successful matching is used to update the target state using a probability weighting method. At the same time, the life cycle of the target is managed, thereby achieving continuous tracking of multiple targets in the scene. The multi-target cluster tracking of personnel in step S4 is performed as follows:

[0086] S41 Kalman filter prediction: Predict the target's state in the next frame, such as position and speed, through a linear dynamic model and update the covariance matrix.

[0087] S42 Data Association: Matches radar point clouds to target trajectories through a two-stage data association process: global matching followed by local compensation. First, existing targets are associated, and then unmatched points are reassigned based on distance thresholds.

[0088] S43 Data Management: Use the DBSCAN clustering algorithm to spatially cluster unassociated radar point clouds, initializing point clusters that meet density conditions and are far away from existing targets as new tracking targets, achieving new detection and lifecycle management of dynamic targets;

[0089] S44 target update: the update module updates the motion state of the associated target by extending the Kalman filter state update and JPDA algorithm, and manages the target life cycle.

[0090] The algorithm flow is shown in Figure 6 Figure 6 The personnel tracking algorithm flowchart in the embodiment is shown in

[0091] S401, input the point cloud data Pt={p1, p2,..., pn} of the current frame, for the first frame data, initialize the clustering result Cj as a new target Tk, and have the target state vector xk=[x, y, z, vx, vy, vz] T , covariance matrix Pk, Kalman filter prediction, the formula is:

[0092] State prediction:

[0093] Where F k is the state transition matrix, and Q k is the process noise covariance.

[0094] S402, for each target predicted state and point cloud cluster C j , the association of the target is carried out, the Mahalanobis distance is calculated, if the distance D M is less than the threshold, it is considered that the association is effective, and the formula is as follows:

[0095]

[0096] Where z j is the observation value of the cluster C j , H is the observation matrix, is the innovation covariance, and R is the observation noise.

[0097] S403, calculate the association probability of each measurement point and the target, specifically as follows:

[0098] Calculate the residual of the measurement point z i and the target j:

[0099]

[0100] Thus, the covariance matrix of the residual is:

[0101]

[0102] Calculate the association probability β ij of the measurement point z i and the target j and the normalized association probability and use the association probability to weight the residual: ​

[0103]

[0104] wherein β0 is the clutter probability.

[0105] S404, then the calculation of the weighted residual error is carried out:

[0106]

[0107] S405, finally the Kalman gain is calculated for updating the target state and the covariance matrix by the weighted residual error:

[0108]

[0109] wherein P w is the covariance matrix of the weighted residual error.

[0110] S5, through the above steps, the entering, leaving, getting into bed and other activity states of the dormitory personnel are monitored in real time, data reports are formed, and monitoring data information combined with real-time time is sent to effectively manage the dormitory personnel.

[0111] The effective management of the dormitory personnel in the step S5 is carried out in the following manner:

[0112] Based on the analysis and processing of the millimeter wave radar point cloud data, the position of the personnel can be detected and tracked, through the long period monitoring of the personnel, the back-to-dormitory time, the in-place time, the sleep time and the like of the dormitory personnel in a day can be understood and supervised, regular experience is formed, and the student dormitory is effectively managed in a non-contact manner.

[0113] Embodiment 2

[0114] In this example, a 60G millimeter wave radar is used, which can penetrate common dormitory fabrics such as bed curtains and bedding, while ensuring sufficient resolution to identify human posture changes. It is installed in a standard two-person student dormitory in a certain university, and long-term uninterrupted data collection is carried out. During the experiment, the system continuously records various activity data including daily entry and exit, normal living, learning activities, etc. As shown in Figure 2 The radar device is installed on the left upper side of the dormitory door, 2.55 meters from the ground, to ensure that the angle error in the vertical direction is controlled within a reasonable range. This height can ensure that the monitoring range covers the entire room and obtain the best point cloud data resolution. The key monitoring areas in the experimental scene include two standard student beds: bed 1 and bed 2, and two matching study desks: desk 1 and desk 2, as shown in Figure 3 These areas are the core areas of personnel activities in the dormitory.

[0115] This example uses data from 18:00 on April 1 to 18:00 on April 2 for a detailed description of the 24 hours. During the monitoring period, the system collects data at a fixed frame rate. Each frame contains sufficient point cloud data and records the timestamp synchronously. The characteristics of human activities at different time periods are intuitively presented through two-dimensional yz plane and three-dimensional xyz space visualization. Figure 7 As shown, Figure 7 This is a real-time monitoring diagram of millimeter wave point cloud two-dimensional and three-dimensional data in this embodiment, where: Figure 7 (a) means there is no one in the dormitory. Figure 7 (b) means one person is sitting at a seat. Figure 7 (c) means one person is in bed. Figure 7 (d) in the figure means two people are resting in bed.

[0116] During the initial monitoring phase (18:00-22:00), the point cloud data collected by the system displayed typical characteristics of an unoccupied state: the point cloud was sparsely distributed and fixed in position, primarily concentrated on the metal structures of static furniture such as bed frames, where the reflected signal was strong. The point cloud density remained low, and the spatial distribution pattern remained stable, with no noticeable movement or density changes. This continued until 22:00, when the system first detected a significant change. In the area near the second desk, the point cloud density suddenly increased, forming a new high-density point cloud cluster. The center of this cluster was approximately 0.7-1.2 meters above the ground, which corresponds to the height range of a person in a seat. This point cloud cluster also exhibited a specific spatial distribution pattern: a narrow vertical distribution range and a nearly elliptical expansion in the horizontal plane, corresponding to a person returning to the dormitory and sitting down.

[0117] At 22:27, the system detected the emergence of a second dynamic point cloud cluster, located in the area of ​​desk number one, approximately 0.5 meters away from the first point cloud cluster. Both point cloud clusters exhibited similar spatial characteristics: highly concentrated vertical distribution and relatively expanded horizontal distribution. Three-dimensional visualization clearly showed that the two active point cloud clusters were completely independent in space, each maintaining a stable relative position and both located within the preset desk coordinate area. This dual-source point cloud distribution pattern lasted for nearly two hours, during which time both point cloud clusters maintained regular, small position fluctuations, reflecting the slight activity characteristics of the personnel.

[0118] During the subsequent sleep period (00:25-8:45), the point cloud distribution of the bed area remained relatively stable, but high-precision analysis still detected periodic small changes, with a fluctuation amplitude of about 15%-20% of the basic density. This fluctuation may correspond to the human body's turning over and reflecting the natural body position adjustment during sleep.

[0119] The system also fully recorded the following morning's activities: Around 9:05 a.m., the point cloud cluster in the area surrounding bed number one slowly moved along a pre-set trajectory toward the door, eventually disappearing at the entrance. Thirteen minutes later, the point cloud cluster in the area surrounding bed number two exhibited similar changes, accurately corresponding to the two students' subsequent movements as they got up and left the dormitory. The dormitory remained unoccupied until 6:00 p.m., with no unusual incidents, such as falls, occurring during the entire monitoring period.

[0120] While recording personnel activities, the system also selected data from any time period on April 9 for personnel detection, such as Figure 8 shown. Figure 8 This is a schematic diagram of the millimeter wave radar personnel detection effect in this embodiment, where: Figure 8 (a) means there is no one in the dormitory. Figure 8 (b) means one person is sitting at a seat. Figure 8 (c) means one person is in bed. Figure 8 (d) in the figure means two people are resting in bed.

[0121] Finally, based on the data collation, the activities of dormitory personnel can be recorded and analyzed, such as Figure 9 As shown, Figure 9 This waveform shows dormitory occupant activity over a 24-hour period in this example. Throughout the monitoring process, the system demonstrated strong environmental adaptability and behavioral recognition accuracy, particularly in complete darkness after lights out at night. The system remained stable and continuously output high-quality point cloud data. By analyzing the spatial distribution characteristics of point cloud clusters, the system was able to accurately identify key information such as occupant presence and location, demonstrating the unique advantages of millimeter-wave radar in monitoring everyday life scenarios: non-contact measurement, all-weather operation, and excellent privacy protection.

[0122] The above embodiments are merely preferred embodiments for the purpose of fully illustrating the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are within the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the claims.

Claims

1. A millimeter-wave radar dormitory personnel activity monitoring method for smart campuses, characterized by: The following steps are involved: S1, obtains 3D point cloud data stream through the millimeter wave radar deployed on the roof of the dormitory; Perform preliminary processing on real-time point cloud data, including coordinate conversion, boundary filtering, and dynamic point cloud segmentation; S2. Extract and simulate the monitoring areas in the dormitory, divide the key monitoring areas into regional blocks, and combine the obtained point cloud dataset to divide them into seating areas and bed areas; S3. Performing human presence detection on the processed point cloud data through sliding window statistics and threshold triggering; S4. Perform multi-target tracking on the point cloud data after personnel presence detection.

2. The millimeter-wave radar dormitory personnel activity monitoring method for smart campuses according to claim 1 is characterized in that: The following steps are also included: S5. The real-time monitoring data obtained through the above steps is used to continuously monitor and analyze the daily behavior of dormitory residents, such as entering and exiting the dormitory, getting in and out of bed, etc., and the analysis results are combined with timestamp information accurate to milliseconds to generate a structured monitoring data report.

3. The millimeter-wave radar dormitory personnel activity monitoring method for smart campuses according to claim 1 is characterized in that: The preliminary processing in step S1 is performed as follows: S11 coordinate conversion: uses a transformation algorithm to convert the original radar polar coordinates into a Cartesian coordinate system (x, y, z); S12 boundary filtering: Based on the pre-set dormitory three-dimensional space boundary parameters, a spatial filtering model is established to automatically eliminate invalid point cloud data outside the monitoring area; S13 dynamic point cloud segmentation: By calculating the velocity components (vx, vy, vz) of each point cloud in the three coordinate axis directions, a reasonable velocity threshold is set to distinguish between static background points and moving dynamic target points.

4. The millimeter-wave radar dormitory personnel activity monitoring method for smart campuses according to claim 1 is characterized in that: The division of the monitoring area in step S2 is performed in the following manner: S21 determines dormitory space measurement information and the installation location information of the millimeter-wave radar in the dormitory, including height and angle; S22 performs spatial distribution analysis on the collected point cloud data and divides the dormitory area into multiple functional sub-areas, such as the seating area for study monitoring and the bed area for rest monitoring.

5. The millimeter-wave radar dormitory personnel activity monitoring method for smart campuses according to claim 1 is characterized in that: The person presence detection in step S3 is performed in the following manner: S31 uses a sliding time window technique to count the number of point clouds and spatial distribution density characteristics of the most recent frames in each area in real time based on the division of the dormitory area in step S2, and uses historical monitoring data to perform data smoothing to eliminate instantaneous noise interference; S32 sets relevant thresholds and performs experimental data calibration, setting static clutter point count frames when no one is present and dynamic point count frames when someone is present. The cumulative number of points and point cloud density in the most recent consecutive frames are counted through a sliding window. Based on the comprehensive average judgment of the experimental data, if the threshold is exceeded, it is detected as a person.

6. The millimeter-wave radar dormitory personnel activity monitoring method for smart campuses according to claim 1, characterized in that: In step S4, multi-target tracking is performed on the point cloud data after the presence of personnel is detected. First, the next state of the existing tracked target is predicted and estimated using the Kalman filter algorithm. Then, the Mahalanobis distance between the predicted target and each point cloud cluster in the current frame is calculated to evaluate their association possibility. Finally, a probability-weighted data fusion method is used to update the target state information using the successfully matched observation data.

7. The millimeter-wave radar dormitory personnel activity monitoring method for smart campuses according to claim 1, characterized in that: The multi-target cluster tracking of personnel in step S4 is performed in the following manner: S41 Kalman filter prediction: Establish linear motion state equations, predict the position coordinates, motion speed and other state parameters of each tracking target at the next moment, and update the uncertainty covariance matrix of the state estimation; S42 Data Association: Matches radar point clouds to target trajectories through a two-stage data association process: global matching followed by local compensation. First, existing targets are associated, and then unmatched points are reassigned based on a distance threshold. S43 Data Management: Use the DBSCAN clustering algorithm to spatially cluster unassociated radar point clouds, initializing point clusters that meet density conditions and are far away from existing targets as new tracking targets, achieving new detection and lifecycle management of dynamic targets; S44 Target Update: This update module updates the motion state of the associated target through the extended Kalman filter state update and JPDA algorithm, and manages the target life cycle.

8. The millimeter-wave radar dormitory personnel activity monitoring method for smart campuses according to claim 1, characterized in that: The effective management of dormitory personnel in step S5 is performed in the following manner: Based on the analysis and processing of millimeter-wave radar point cloud data, the location of personnel can be detected and tracked. Through long-term monitoring of personnel, the dormitory occupants' daily return time, on-site time, sleep time, etc. can be understood and supervised, forming regular experience and effectively managing student dormitories without contact.

9. A millimeter-wave radar dormitory personnel activity monitoring system for smart campuses, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

Images