Smart community management system based on big data

Abstract

The invention relates to the field, in particular to a smart community management system based on big data, and the system comprises a data collection module which is used for collecting video monitoring images and access control system records in a community in real time to form first data; the data fusion module is used for performing format unification and time synchronization processing on the first data, constructing corresponding community resident portraits and storing the portraits in a database; the abnormal behavior identification module is used for extracting residential attribute features from the community resident portraits and identifying dynamic information related to resident behaviors; the abnormal behavior recognition module outputs a first result or a second result based on the attribute characteristics and the dynamic information; the response processing module is used for executing a first operation when the first result is received; or executing a second operation when the second result is received. The system can respond in time when the elderly living alone falls down and is injured or possibly has an accident, so that the safety of the elderly is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of smart community management, and in particular to a smart community management system based on big data. Background Art

[0002] With the acceleration of urbanization, community management faces challenges such as an aging population, increased security needs, and diversified service quality.

[0003] Existing community security systems mostly rely on single video surveillance (i.e., facial recognition access control) or access control records. For example, the invention patent with publication number CN112700874A mentions: The present invention can monitor the health status and activity trajectory of the elderly in real time, improve the connection between the elderly and the smart community, and avoid the situation where elderly people living alone are unable to move or become ill and are not discovered in time, resulting in serious consequences. This will enable children and community managers to be more at ease with the lives of the elderly, and provide corresponding help and services based on the living conditions of the elderly, thereby reducing the problem of injuries to the elderly. However, the patent also lacks continuous analysis of residents' behavior, especially in scenarios where the elderly fall or are inactive for a long time, and cannot respond in time.

[0004] Therefore, it is necessary to provide a new technical solution to overcome the above-mentioned defects. Summary of the Invention

[0005] The purpose of the present invention is to provide a big data-based smart community management system that can effectively solve the above-mentioned technical problems.

[0006] In order to achieve the purpose of the present invention, the following technical solutions are adopted:

[0007] A data acquisition module, configured to collect video surveillance images and access control system records in the community in real time to form first data;

[0008] A data fusion module, configured to perform format unification and time synchronization processing on the first data, construct corresponding community resident portraits, and store the portraits in a database;

[0009] An abnormal behavior identification module is used to extract resident attribute characteristics from the community resident portrait and identify dynamic information related to the resident's behavior;

[0010] The abnormal behavior identification module outputs a first result or a second result based on the attribute characteristics and the dynamic information;

[0011] The response processing module is configured to execute a first operation upon receiving the first result, or execute a second operation upon receiving the second result.

[0012] Furthermore, when the resident is an elderly person and falls, the first result is output; when the resident is an elderly person and has not left home continuously, the second result is output.

[0013] Furthermore, the first operation includes: automatically turning on a voice call module of a related camera and attempting to interact with the resident to confirm the resident's status;

[0014] The second operation includes at least one of the following: sending an alarm message to a resident terminal; issuing a second-level alarm to a community service terminal, and recommending that staff be arranged for on-site verification.

[0015] Furthermore, the resident attribute characteristics include an age tag;

[0016] The dynamic information includes: time of going out, time of returning home, and electricity consumption; the duration of going out and the duration of consecutive absence from going out are calculated based on the time of going out and the time of returning home;

[0017] The dynamic behavior information is constructed as a behavior time series and input into a time series prediction model based on a long short-term memory network; if the degree of deviation between the current sequence and the historical pattern exceeds a set threshold and the model prediction score is lower than a preset threshold, it is determined to be abnormal behavior.

[0018] Furthermore, the community resident portrait includes the recorded health record data of the residents, family member information bound to the residents, and historical service request records.

[0019] Furthermore, the data fusion module also includes a face recognition unit;

[0020] The face recognition unit detects and captures the face area from the first data, pre-processes the face area, extracts the feature vector in the face area, and compares the similarity with the resident face feature template stored in the database; if the similarity is higher than a preset first threshold, the person is verified as a legal resident; if the similarity is lower than a preset second threshold, the person is judged as a suspicious person and an alarm is triggered; if the similarity is between the first threshold and the second threshold, a secondary verification process is initiated.

[0021] Furthermore, the triggering of the alarm includes: locking the access control system and reminding security personnel to go to the site for verification.

[0022] Furthermore, a user feedback module is used to receive service evaluations sent by resident terminals or community service terminals, add them to the data set, and optimize the model threshold;

[0023] storing the service evaluation information as training data, the service evaluation information at least including an evaluation of the on-site verification efficiency of staff and the recognition accuracy of the system;

[0024] Dynamically updating the model determination threshold based on the training data to improve the accuracy of abnormal behavior identification;

[0025] Provides a "false alarm" marking function, allowing residents or their families to mark triggered alarm events;

[0026] After receiving the "false alarm" mark, the corresponding alarm scene and related raw data are automatically included in the training set;

[0027] Retrain or fine-tune the time series prediction model based on the "false alarm" scenario data, and adjust the threshold setting based on the training results.

[0028] Furthermore, the preprocessing includes: gray-scaling the visible light image and extracting the edge contour of the face using the Canny edge detection algorithm;

[0029] Adopting contour tracking algorithm, discrete edge points are connected to form a closed facial contour curve, including the continuous contours of forehead, cheeks and jaw;

[0030] Smoothing the facial contour, removing noise points (such as interference from beards and hair), and outputting an accurate facial contour area;

[0031] Divide the collected visible light image into blocks according to resolution, mark the sub-blocks covered by the facial contour in the image as facial blocks, and mark the sub-blocks outside the facial contour as background blocks;

[0032] Contour features of the facial block and the background block are extracted respectively, and the contour complexity within each sub-block is calculated.

[0033] Furthermore, the face recognition module further includes: a liveness detection unit, which analyzes the dynamic continuity of the facial contour, calculates the displacement deviation of contour key points by comparing the facial contours of three consecutive frames of images, and outputs a contour dynamic consistency score;

[0034] If the mean facial contour complexity score is greater than 0.6 and the contour dynamic consistency score is greater than 0.8, the first liveness judgment is passed; otherwise, the test ends;

[0035] Calculate the mean texture uniformity of the background block: normalize the LBP texture features of all sub-blocks of the background block and calculate the average value as the "mean background texture uniformity";

[0036] Analyze the contour correlation between facial blocks and background blocks: calculate the degree of fusion between facial contour edges and background textures, and output the contour-background fusion score;

[0037] If the background texture uniformity mean is <0.5 and the outline-background fusion score is >0.7, it is determined to be a real face; otherwise, the detection ends.

[0038] Compared with the prior art, the present invention has the following beneficial effects: the big data-based smart community management system of the present invention collects first data in real time through a data acquisition module; the abnormal behavior recognition module is used to extract resident attribute characteristics from the community resident portrait, and identify dynamic information related to the resident behavior, and output a first result or a second result based on the attribute characteristics and the dynamic information; the response processing module is used to execute a first operation when receiving the first result; or execute a second operation when receiving the second result; thereby, a timely response can be made when an elderly person living alone falls and is injured or an accident may occur, thereby ensuring the safety of the elderly. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0040] Figure 1 This is a flow chart of the smart community management system based on big data of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments.

[0042] It should be understood that, although the various steps in the flow chart of each embodiment of the present invention are shown in sequence according to the indication of the arrows, these steps are not necessarily performed in sequence according to the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in order, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0043] like Figure 1 As shown, the present invention relates to a smart community management system based on big data, including:

[0044] The data acquisition module connects to the community's surveillance cameras and access control devices to collect video surveillance images and access control system records in the community in real time to form the first data; by connecting the community's surveillance cameras (video images) and access control devices (entry and exit records), it can obtain the original data of residents' behavior trajectories and activity status (such as time of going out and time of returning home) in real time, providing the most basic time anchor point and behavioral basis for subsequent abnormality analysis.

[0045] The data fusion module is used to unify the format and time synchronization of the first data, construct corresponding community resident portraits, and store the portraits in the database; it solves the problems of incompatible data formats and confusing timestamps of different devices (cameras, access control), ensuring that the data can be aligned and associated during subsequent analysis; the community resident portraits include the entered health record data of the residents, information on family members bound to the residents, and historical service request records; the database also stores video surveillance streams, access control system records, residents' historical behavior timelines (such as activity time / frequency), and resident information (such as advanced age / living alone / chronic diseases), which are used to specifically identify the potential risks of special groups (such as elderly people living alone) and achieve differentiated attention.

[0046] The abnormal behavior identification module is used to extract resident attribute characteristics from the community resident portrait and identify dynamic information related to the resident behavior; the resident attribute characteristics include age tags; the dynamic information includes: time of going out, time of returning home, and electricity consumption; based on the time of going out and the time of returning home, the duration of going out and the duration of continuous absence from going out are calculated; the dynamic behavior information is constructed as a behavior time series and input into a time series prediction model based on a long short-term memory network (LSTM); if the degree of deviation between the current sequence and the historical pattern exceeds a set threshold and the model prediction score is lower than the preset threshold, it is determined to be abnormal behavior.

[0047] The abnormal behavior recognition module outputs a first result based on the attribute characteristics and the dynamic information when the resident is an elderly person and falls; and outputs a second result when the resident is an elderly person and has not left home continuously.

[0048] The response processing module is configured to execute a first operation upon receiving the first result, or execute a second operation upon receiving the second result.

[0049] The first operation includes: automatically turning on the relevant camera voice call module and trying to interact with residents to confirm the residents' status; directly turning on the camera voice call confirmation status and quickly obtaining feedback from residents (such as "I'm fine" or "I need help"). If the elderly need help, community staff will rush to the scene as soon as possible to provide timely help.

[0050] The second operation includes at least one of the following: sending an alarm message to the resident terminal; issuing a second-level alarm to the community service terminal, and recommending that staff be arranged for on-site verification; the alarm message can be sent in sequence according to the priority order set in the system; for example, an alarm message is first sent to the resident terminal of the elderly person's family. If the elderly person's family responds in time, the second-level alarm will no longer be sent; otherwise, a second-level alarm will be issued to the community service terminal after a predetermined time (1-3 minutes), and it is recommended that staff be arranged for on-site verification.

[0051] The data fusion module also includes a face recognition unit;

[0052] The face recognition unit detects and captures the face area from the first data, pre-processes the face area, extracts the feature vector in the face area, and compares the similarity with the resident face feature template stored in the database; if the similarity is higher than a preset first threshold, the person is verified to be a legal resident; if the similarity is lower than a preset second threshold, the person is judged to be a suspicious person and an alarm is triggered; if the similarity is between the first threshold and the second threshold, a secondary verification process is started; the triggering alarm includes at least one of the following: reminding residents to re-scan the face for recognition; reminding residents to swipe the card to open the door; locking the access control system and reminding security personnel to go to the scene to verify whether the person is a community resident.

[0053] The preprocessing includes: grayscale processing of the visible light image to reduce color interference, extracting the edge contour of the face using the Canny edge detection algorithm, retaining strong edges and suppressing weak edges using the double threshold method;

[0054] Adopting contour tracking algorithm, discrete edge points are connected to form a closed facial contour curve, including the continuous contours of forehead, cheeks and jaw;

[0055] The facial contour is smoothed using B-spline curve fitting to remove noise points (such as interference from beards and hair) and output an accurate facial contour area;

[0056] Divide the captured visible light image into blocks according to resolution (for example, a 512×512 pixel image is divided into 8×8 sub-blocks, each 64×64 pixels). Mark the sub-blocks covered by the facial outline in the image as face blocks (at least 70% of the sub-block area must be covered). Mark the sub-blocks outside the facial outline as background blocks, giving priority to sub-blocks adjacent to the head block to reduce environmental interference.

[0057] Extract the contour features of the facial block and the background block respectively, and calculate the contour complexity of each sub-block; let the perimeter be C, the area be S, and the complexity be F;

[0058] Then F=C 2÷(4πS), calculate the average complexity of all sub-blocks based on the complexity F.

[0059] The face recognition module also includes a liveness detection unit that analyzes the dynamic continuity of the facial contour, calculates the displacement deviation of the contour key points (such as the position change of the highest point of the forehead and the lowest point of the mandible) by comparing the facial contours of three consecutive frames of images, and extracts N key points (coordinates expressed as Where t takes values ​​of 1, 2, or 3 to represent the frame number; i takes values ​​of 1, 2, 3, ..., N to represent the key point number. Formula 1 is used to calculate the Euclidean distance of each key point between two adjacent frames as the displacement. Formula 2 is used to normalize the displacements of all key points to eliminate the differences in the motion amplitudes of different key points. The average of all normalized displacements is taken and mapped to the interval [0, 1] using an exponential function. Formula 3 is used to calculate and output the contour dynamic consistency score.

[0060] Wherein, Formula 1 is as follows:

[0061] Where t = 1, 2;

[0062] Formula 2 is as follows:

[0063] in is the average motion amplitude of the key point preset in the system in the real face;

[0064] Formula 3 is as follows:

[0065]

[0066] , α is the adjustment parameter.

[0067] If the mean facial contour complexity score is > 0.6 and the contour dynamic consistency score is > 0.8, the first liveness judgment is passed; otherwise, the test is terminated and it is judged as a prosthesis, such as a mask or photo. The contour of a real person is more complex due to muscle and bone structure, while the contour of a prosthesis has no natural displacement.

[0068] Calculate the mean texture uniformity of the background block: normalize the LBP texture features of all sub-blocks of the background block and calculate the average value as the "mean background texture uniformity";

[0069] Analyze the contour correlation between facial blocks and background blocks: Calculate the degree of fusion between facial contour edges and background textures. For example, the contour edges of real people will naturally blend with the background (such as hair and collar), while the contour edges of prosthetics may appear discontinuous. Output the contour-background fusion score;

[0070] If the background texture uniformity mean is <0.5 and the contour-background fusion score is >0.7, the face is determined to be a real person; otherwise, the detection ends and the face is determined to be a fake. Real background textures are usually uneven, such as walls and green plants, with a natural transition between the contour and the background. The contour-background fusion score = the number of edge points with natural texture transition / the total number of edge points.

[0071] Through four-fold verification: dynamic contour consistency (displacement deviation of key points in three consecutive frames), contour complexity (real person muscles and bones are more complex), background texture uniformity (real backgrounds such as green plants / walls have uneven textures), and contour-background fusion (natural transition between real person contours and hair / collars), it can effectively identify prostheses such as masks and photos, and prevent malicious identity forgery from entering the community (such as thieves using photos to deceive access control).

[0072] By analyzing behavioral time series (such as time of going out / returning home, electricity consumption), the present invention can capture anomalies that deviate from historical patterns (such as an elderly person who usually goes out at 7 o'clock has not gone out for three consecutive days). Compared with traditional fixed threshold alarms, this invention is more in line with the personalized behavioral habits of residents and reduces false alarms. The present invention focuses on high-risk groups and designs recognition logic for the elderly (falls, long-term absences from home), accurately covering the groups in the community that "need the most attention" (such as elderly people living alone), and improving the efficiency of risk intervention.

[0073] In this embodiment, the data fusion module can directly reuse existing monitoring equipment in communities, access control and other scenarios, and the hardware cost is relatively low.

[0074] User feedback module, which is used to receive service evaluations sent by resident terminals or community service terminals, add them to the data set, and optimize the model threshold;

[0075] storing the service evaluation information as training data, the service evaluation information at least including an evaluation of the on-site verification efficiency of staff and the recognition accuracy of the system;

[0076] Dynamically updating the model determination threshold based on the training data to improve the accuracy of abnormal behavior identification;

[0077] Provides a "false alarm" marking function, allowing residents or their families to mark triggered alarm events;

[0078] After receiving the "false alarm" mark, the corresponding alarm scene and related raw data are automatically included in the training set;

[0079] Retrain or fine-tune the time series prediction model based on the "false alarm" scenario data, and adjust the threshold setting based on the training results;

[0080] By collecting service evaluations (such as staff efficiency and recognition accuracy) and "false alarm marking" data, the model threshold can be dynamically adjusted (such as reducing the sensitivity to "occasional late return") and the LSTM model can be retrained to make the system "more accurate with use"; allowing residents / family members to mark false alarms to enhance users' trust in the system ("my feedback can be heard and improved"), while using real-world scenario data to improve the model's generalization ability (avoiding the "water and soil incompatibility" caused by relying solely on laboratory data).

[0081] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0082] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0083] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims

1. A smart community management system based on big data, characterized by: include: A data acquisition module, configured to collect video surveillance images and access control system records in the community in real time to form first data; A data fusion module, configured to perform format unification and time synchronization processing on the first data, construct corresponding community resident portraits, and store the portraits in a database; An abnormal behavior identification module is used to extract resident attribute characteristics from the community resident portrait and identify dynamic information related to the resident's behavior; The abnormal behavior identification module outputs a first result or a second result based on the attribute characteristics and the dynamic information; The response processing module is configured to execute a first operation upon receiving the first result, or execute a second operation upon receiving the second result.

2. The big data-based smart community management system according to claim 1, characterized in that: When the resident is elderly and falls, the first result is output; when the resident is elderly and has not left home continuously, the second result is output.

3. The big data-based smart community management system according to claim 2, characterized in that: The first operation includes: automatically turning on a voice call module of a related camera and attempting to interact with the resident to confirm the resident's status; The second operation includes at least one of the following: sending an alarm message to a resident terminal; issuing a second-level alarm to a community service terminal, and recommending that staff be arranged for on-site verification.

4. The big data-based smart community management system according to claim 3, characterized in that: The resident attribute characteristics include age label; The dynamic information includes: time of going out, time of returning home, and electricity consumption; the duration of going out and the duration of consecutive absence from going out are calculated based on the time of going out and the time of returning home; The dynamic behavior information is constructed as a behavior time series and input into a time series prediction model based on a long short-term memory network; if the degree of deviation between the current sequence and the historical pattern exceeds a set threshold and the model prediction score is lower than a preset threshold, it is determined to be abnormal behavior.

5. The big data-based smart community management system according to claim 1, characterized in that: The community resident portrait includes the recorded health record data of the residents, family member information bound to the residents, and historical service request records.

6. The big data-based smart community management system according to claim 5, characterized in that: The data fusion module also includes a face recognition unit; The face recognition unit detects and captures a face region from the first data, pre-processes the face region, extracts a feature vector from the face region, and compares the feature vector with a resident face feature template stored in a database; if the similarity is higher than a preset first threshold, the person is verified as a legal resident; if the similarity is lower than a preset second threshold, the person is determined to be a suspicious person and an alarm is triggered; If the similarity is between the first threshold and the second threshold, the secondary verification process is started.

7. The big data-based smart community management system according to claim 6, characterized in that: The triggering of the alarm includes: locking the access control system and reminding security personnel to go to the scene for verification.

8. The big data-based smart community management system according to claim 1, characterized in that: User feedback module, which is used to receive service evaluations sent by resident terminals or community service terminals, add them to the data set, and optimize the model threshold; storing the service evaluation information as training data, the service evaluation information at least including an evaluation of the on-site verification efficiency of staff and the recognition accuracy of the system; Dynamically updating the model determination threshold based on the training data to improve the accuracy of abnormal behavior identification; Provides a "false alarm" marking function, allowing residents or their families to mark triggered alarm events; After receiving the "false alarm" mark, the corresponding alarm scene and related raw data are automatically included in the training set; Retrain or fine-tune the time series prediction model based on the "false alarm" scenario data, and adjust the threshold setting based on the training results.

9. The big data-based smart community management system according to claim 6, characterized in that: The preprocessing includes: grayscale processing of the visible light image and extracting the edge contour of the face using the Canny edge detection algorithm; Adopting contour tracking algorithm, discrete edge points are connected to form a closed facial contour curve, including the continuous contours of forehead, cheeks and jaw; Smoothing the facial contour, removing noise points (such as interference from beards and hair), and outputting an accurate facial contour area; Divide the collected visible light image into blocks according to resolution, mark the sub-blocks covered by the facial contour in the image as facial blocks, and mark the sub-blocks outside the facial contour as background blocks; Contour features of the facial block and the background block are extracted respectively, and the contour complexity within each sub-block is calculated.

10. The big data-based smart community management system according to claim 9, characterized in that: The face recognition module further includes: a liveness detection unit that analyzes the dynamic continuity of the facial contour, calculates the displacement deviation of contour key points by comparing the facial contours of three consecutive frames of images, and outputs a contour dynamic consistency score; If the mean facial contour complexity score is greater than 0.6 and the contour dynamic consistency score is greater than 0.8, the first liveness judgment is passed; otherwise, the test ends; Calculate the mean texture uniformity of the background block: Normalize the LBP texture features of all sub-blocks of the background block and calculate the average value as the "mean background texture uniformity"; Analyze the contour correlation between facial blocks and background blocks: calculate the degree of fusion between facial contour edges and background textures, and output the contour-background fusion score; If the background texture uniformity mean is <0.5 and the outline-background fusion score is >0.7, it is determined to be a real face; otherwise, the detection ends.

Images