Community data management system for intelligent old-age care community

Abstract

The invention relates to the field of data management, in particular to a community data management system for an intelligent old-age care community, and the system comprises a video monitoring unit which comprises a plurality of community monitoring devices and is used for collecting monitoring videos of all monitoring areas; the video analysis unit is used for periodically determining the video frame category of each basic video frame according to the personnel change reference value and the personnel influence reference value of the basic video frame corresponding to the community monitoring equipment; the video optimization unit is used for determining the state of the monitoring equipment according to the first-class video frame proportion and the first-class video frame density corresponding to the community monitoring equipment, and determining a corresponding video frame optimization mode as compensation optimization or activity feature optimization according to the state of the monitoring equipment; the key frame optimization unit is used for determining a processing mode to supplement or generate image description for the associated frame according to the effective state; and the monitoring data processing efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of data management, and particularly to a community data management system for an intelligent elderly care community. Background Art

[0002] With the development of intelligent technology, elderly care communities provide efficient and comprehensive monitoring and security services for the elderly through monitoring devices. Existing monitoring devices integrate advanced technologies such as big data, cloud computing, and artificial intelligence in the data processing stage, and identify the dangerous states of the elderly through deep learning models and data analysis, providing more convenient and efficient intelligent elderly care services for the elderly. However, during the use of monitoring devices, problems such as wide distribution of monitoring areas and large amounts of monitoring data caused by continuous collection of monitoring data are usually faced. Therefore, how to achieve intelligent monitoring data management to improve the safety monitoring efficiency of elderly care communities is a technical problem that needs to be solved urgently by those skilled in the art.

[0003] Chinese Patent Publication No. CN114343595B discloses a community intelligent elderly care service system, including: a behavior acquisition module for collecting the behavior data of the elderly through a camera combined with a posture acquisition belt; a vital sign acquisition module for collecting the vital sign data of the elderly and real-time feedback to the vital sign analysis module; a behavior analysis module for analyzing the behavior data of the elderly to obtain their abnormal behaviors and psychological needs; a vital sign analysis module for analyzing the vital sign data to obtain their abnormal vital signs; an abnormal real-time feedback module for generating corresponding abnormal investigation tasks to the corresponding management personnel terminals in a preset template, and the management personnel need to complete and upload the abnormal investigation report in the established template to complete the corresponding abnormal investigation tasks. It can be seen that the above technical solution has the following problems: it does not consider the processing difficulty of a large amount of monitoring data generated by long-term and multi-region monitoring modes, resulting in poor monitoring data processing efficiency, and further causing poor community safety monitoring effects. Summary of the Invention

[0004] Therefore, the present invention provides a community data management system for an intelligent elderly care community to overcome the problems in the prior art that the processing difficulty of a large amount of monitoring data generated by long-term and multi-region monitoring modes in the monitoring of elderly care communities is not considered, resulting in poor monitoring data processing efficiency and further causing poor community safety monitoring effects.

[0005] To achieve the above object, the present invention provides a community data management system for an intelligent elderly care community, including:

[0006] A video monitoring unit, which includes a number of community monitoring devices for collecting monitoring videos of each monitoring area;

[0007] A video analysis unit, which is connected to the video monitoring unit and is used to periodically determine the video frame category of each basic video frame according to the personnel change reference value and the personnel influence reference value corresponding to the community monitoring device;

[0008] A video optimization unit, which is respectively connected to the video monitoring unit and the video analysis unit, and is used to determine the monitoring device status according to the proportion of a certain type of video frame and the density of a certain type of video frame corresponding to the community monitoring device, and determine the corresponding video frame optimization method as compensation optimization or activity feature optimization according to the monitoring device status;

[0009] A key frame optimization unit, which is respectively connected to the video monitoring unit, the video analysis unit and the video optimization unit, and is used to determine the processing method as associated frame supplementation or image description generation according to the effective status.

[0010] Further, the video analysis unit determines the video frame category according to the personnel change reference value and the personnel influence reference value corresponding to the basic video frame of the community monitoring device. The video frame categories include:

[0011] A certain type of video frame in which the personnel change reference value is greater than the preset personnel change reference value or the personnel influence reference value is greater than the preset personnel influence reference value;

[0012] A second type of video frame in which the personnel change reference value is less than or equal to the preset personnel change reference value and the personnel influence reference value is less than or equal to the preset personnel influence reference value.

[0013] Further, the video optimization unit determines the monitoring device status of each community monitoring device according to the proportion of a certain type of video frame and the density of a certain type of video frame corresponding to each community monitoring device, and determines the corresponding video frame optimization method according to the monitoring device status of the community monitoring device;

[0014] If the monitoring device status is that the proportion of a certain type of video frame is greater than the preset proportion of a certain type of video frame and the density of a certain type of video frame is greater than the preset density of a certain type of video frame, the video frame optimization method is compensation optimization;

[0015] If the monitoring device status is that the proportion of a certain type of video frame is less than or equal to the preset proportion of a certain type of video frame or the density of a certain type of video frame is less than or equal to the preset density of a certain type of video frame, the video frame optimization method is activity feature optimization.

[0016] Further, when the video optimization unit performs compensation optimization on a single community monitoring device, the community monitoring device is recorded as the target analysis device;

[0017] Detect the radiation monitoring device whose associated influence coefficient corresponding to the target analysis device is greater than the preset associated influence coefficient;

[0018] Perform video frame enlargement extraction on the surveillance video for the dense video period corresponding to the target analysis device;

[0019] Perform video frame enlargement extraction on the surveillance video for the sparse video period corresponding to the radiation monitoring device.

[0020] Furthermore, during the process of enlarging video frame extraction, the video optimization unit respectively detects the personnel flow coefficients of the target analysis device and the radiation monitoring device;

[0021] Determine the amount of video frame enlargement according to the difference between the personnel flow coefficient and the preset personnel flow coefficient.

[0022] Furthermore, when the video optimization unit performs activity feature optimization, it includes:

[0023] Determine the risk evaluation value of each basic video frame according to the activity amplitude coefficient and the position anomaly coefficient of the basic video frame;

[0024] If the risk evaluation value corresponding to the basic video frame is greater than the preset risk evaluation value, the video frame is an activity warning video frame, and extract a preset number of video frames for the warning range of each activity warning video frame;

[0025] The preset number has a positive correlation with the risk evaluation value.

[0026] Furthermore, the key frame optimization unit respectively performs effective analysis on each extracted video frame;

[0027] The effective analysis includes determining the effective state of the extracted video frame according to the dynamic-static imbalance coefficient and the image quality coefficient corresponding to the extracted video frame, and determining the corresponding processing method according to the effective state.

[0028] Furthermore, in response to the effective state that the image quality coefficient is less than or equal to the preset image quality coefficient or the dynamic-static imbalance coefficient is greater than the preset dynamic-static imbalance coefficient, the key frame optimization unit determines that the processing method is to supplement associated frames for the extracted video frame;

[0029] The number of supplemented associated frames has a positive correlation with the effective difference.

[0030] Furthermore, in response to the effective state that the dynamic-static imbalance coefficient is less than or equal to the preset dynamic-static imbalance coefficient and the dynamic-static imbalance coefficient is greater than the preset dynamic-static imbalance coefficient, the key frame optimization unit determines that the processing method is to generate an image description and generate keywords for the extracted video frame and the associated frames.

[0031] Furthermore, the video analysis unit determines the video frame category of the basic video frame once per monitoring cycle.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows. In the technical solution of the present invention, the video analysis unit determines the video frame category according to the personnel change reference value and the personnel influence reference value corresponding to the basic video frame of the community monitoring device. The personnel change reference value and the personnel influence reference value reflect the analysis difficulty of the personnel image in the video frame and the actual state of the personnel, and are correspondingly divided into different categories. And according to the proportion of one type of video frame and the density of one type of video frame, the monitoring device state is determined, and the corresponding video frame optimization method is determined according to the monitoring device state, avoiding the problem that the single video frame selection method in the prior art cannot effectively meet the analysis requirements of the actual complex monitoring video, making the selection of video frames more in line with the actual working scenario, and thus improving the management efficiency of the monitoring video data.

[0033] Further, in the technical solution of the present invention, during the process of increasing the video frame extraction, the amount of video frame increase is determined according to the difference between the personnel flow coefficient and the preset personnel flow coefficient. The personnel flow coefficient reflects the personnel flow activity degree between the camera device and the surrounding camera devices, and different increasing methods are correspondingly selected, making the selection of the increasing method more beneficial to the subsequent analysis of the video frame, avoiding the problem that the extracted video frames are not conducive to analysis caused by a single increasing method, and thus improving the monitoring analysis efficiency.

[0034] Further, when the video optimization unit performs activity feature optimization in the technical solution of the present invention, the warning range of the video frame is determined according to the activity amplitude coefficient and the position anomaly coefficient of each video frame. The activity characteristics of people are quantitatively analyzed through the activity amplitude and position change, so as to judge whether the personnel are in a normal activity state, and thus reflect whether the personnel in the video are in danger, which helps to improve the extraction efficiency of dangerous video frames and facilitates the community to take corresponding rescue measures.

[0035] Further, in the technical solution of the present invention, the effective state of the extracted video frame is determined according to the dynamic-static imbalance coefficient and the dynamic-static imbalance coefficient, and whether to supplement the associated frame is determined according to the effective state, reducing the low accuracy of the subsequent image recognition caused by the poor image acquisition quality due to occlusion and movement, etc., improving the image quality, and thus improving the subsequent video analysis efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a unit connection diagram of the community data management system for the intelligent elderly care community of the present invention;

[0037] Figure 2 It is a flowchart for the present invention to determine the video frame category;

[0038] Figure 3 It is a flowchart for the present invention to determine the monitoring device state of the community monitoring device. DETAILED DESCRIPTION OF THE INVENTION

[0039] In order to make the objectives and advantages of the present invention more clear and understandable, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0040] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0041] It should be noted that in the description of the present invention, the terms indicating the direction or positional relationship such as "upper", "lower", "left", "right", "inner", "outer", etc. are based on the direction or positional relationship shown in the drawings. This is only for convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0042] In addition, it should also be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installation", "connection", and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; 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. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0043] Please refer to Figures 1 to 3 As shown, the present invention provides a community data management system for an intelligent elderly care community, including:

[0044] A video monitoring unit, which includes a number of community monitoring devices for collecting monitoring videos of each monitoring area;

[0045] A video analysis unit, which is connected to the video monitoring unit and is used to periodically determine the video frame category of each basic video frame according to the personnel change reference value and the personnel influence reference value corresponding to the community monitoring device;

[0046] A video optimization unit, which is respectively connected to the video monitoring unit and the video analysis unit, and is used to determine the monitoring device status according to the proportion of a certain type of video frame and the density of a certain type of video frame corresponding to the community monitoring device, and determine the corresponding video frame optimization method as compensation optimization or activity feature optimization according to the monitoring device status;

[0047] A key frame optimization unit, which is respectively connected to the video monitoring unit, the video analysis unit, and the video optimization unit, and is used to determine the processing method as associated frame supplementation or generation of an image description according to the effective status.

[0048] The application scenario of the present invention is the management of surveillance video data in a pension community. The community surveillance device is a surveillance device with video acquisition and video transmission functions. The specific installation location of the surveillance device is not limited, and the area of the surveillance video captured by the surveillance device is recorded as its corresponding surveillance area.

[0049] The present invention has historical records. A single historical record records a number of detected values detected during a historical surveillance process, including but not limited to a personnel change reference value, a personnel influence reference value, a proportion of a type of video frame, a density of a type of video frame, a correlation influence coefficient, a personnel quantity difference, a basic value of a bone rotation angle, a basic value of a translation displacement, a risk evaluation value, an image quality coefficient, and a dynamic and static imbalance coefficient. And the historical record is attached with a qualified mark, and the qualified mark records whether the historical record meets the user's requirements. Among them, whether it meets the user's requirements is determined by the user according to the surveillance effect. For example, the user can set the missed acquisition rate as a judgment index. If the missed acquisition rate is within the user's acceptable range, the surveillance effect is qualified, that is, the historical record meets the user's requirements. The missed acquisition rate is a dangerous event that has not been captured in the extracted video frames and associated frames. The dangerous events include but are not limited to personnel falling and vehicle collisions, etc.

[0050] Specifically, the video analysis unit determines the video frame category according to the personnel change reference value and the personnel influence reference value of the basic video frame corresponding to the community surveillance device. The video frame categories include:

[0051] A type of video frame in which the personnel change reference value is greater than the preset personnel change reference value or the personnel influence reference value is greater than the preset personnel influence reference value;

[0052] A type of video frame in which the personnel change reference value is less than or equal to the preset personnel change reference value and the personnel influence reference value is less than or equal to the preset personnel influence reference value.

[0053] At the end of each monitoring period, the video analysis unit makes a determination of the video frame category, extracts the surveillance videos corresponding to each community surveillance device in the most recent monitoring period. For a single surveillance video, a number of video frames are evenly extracted, and each video frame is recorded as a basic video frame. The number of the extracted video frames is set by the user himself. The user can set the number of the extracted video frames according to the video surveillance accuracy requirement. The greater the video surveillance accuracy requirement, the greater the number of the extracted basic video frames. A value of the number of the basic video frames is provided. For a single surveillance video, the number of its corresponding basic video frames is 30% of its total number of video frames;

[0054] For a single basic video frame, the method for confirming the corresponding personnel change reference value is as follows: extract the relevant video segment of this basic video frame, extract the movement trajectories of the personnel with a presence duration greater than the preset presence duration in the relevant video segment through the object detection algorithm, detect the trajectory conflict degree, and record the trajectory conflict degree as the personnel change reference value. The trajectory conflict degree = the number of intersection points + the average value of the trajectory complexity. An intersection point is a point where there is an intersection between two movement trajectories. The trajectory complexity of a single movement trajectory is the number of inflection points of this movement trajectory, and an inflection point is a point where the movement direction changes on this movement trajectory. The presence duration is the maximum duration for which a person continuously appears in the surveillance video. The relevant video segment is a surveillance video paragraph with the basic video frame as the midpoint. The method for confirming the duration of the relevant video segment is to extract the average value of the durations of the relevant video segments corresponding to the historical records that meet the user's requirements and record it as the duration of the current relevant video segment; the object detection algorithm can be YOLO, SSD, or Faster R-CNN, which is easily understood by those skilled in the art and will not be elaborated here.

[0055] For a single basic video frame, the method for confirming the corresponding personnel influence reference value is as follows: detect the minimum distance between each person in this basic video frame, record the average value of the minimum distances as A, and record 1 / A as the personnel influence reference value. For a single person, the corresponding minimum distance is the distance between this person and the nearest other person. It should be noted that if the number of people in the basic video frame is less than or equal to 1, the personnel influence reference value is 0.

[0056] For the values of the preset personnel change reference value and the preset personnel influence reference value, the user can set them according to the actual application scenario. The personnel change reference value reflects the complexity of the personnel movement trajectories and the movement conflict degree between personnel in the surveillance area, and the personnel influence reference value reflects the distance distribution between personnel in the surveillance area. The greater the user's monitoring requirement for the monitored personnel, the smaller the values of the preset personnel change reference value and the preset influence reference value. A value-taking method is provided. Statistically analyze the personnel change reference value and the personnel influence reference value in the historical records that meet the user's requirements, eliminate the outliers among them, and record the average values of the personnel change reference value and the personnel influence reference value after eliminating the outliers as the current preset personnel change reference value and the preset personnel influence reference value respectively.

[0057] Specifically, the video optimization unit determines the monitoring device status of each community monitoring device according to the proportion of a type of video frame and the density of a type of video frame corresponding to each community monitoring device, and determines the corresponding video frame optimization method according to the monitoring device status of the community monitoring device;

[0058] If the status of the monitoring device is that the proportion of Class I video frames is greater than the preset proportion of Class I video frames and the density of Class I video frames is greater than the preset density of Class I video frames, the video frame optimization method is compensation optimization;

[0059] If the status of the monitoring device is that the proportion of Class I video frames is less than or equal to the preset proportion of Class I video frames or the density of Class I video frames is less than or equal to the preset density of Class I video frames, the video frame optimization method is activity feature optimization.

[0060] For a single community monitoring device, the proportion of Class I video frames corresponding to its single monitoring period = the number of Class I video frames / (the number of Class I video frames + the number of Class II video frames). For the monitoring video corresponding to this monitoring period, detect the first Class I video frame and the last Class I video frame in the time sequence, and record the ratio of the number of Class I video frames / the number of Class II video frames between the two Class I video frames as the density of Class I video frames;

[0061] The value of the preset proportion of Class I video frames can be set by the user according to the actual application scenario. In the present invention, the proportion of Class I video frames can effectively reflect the personnel activity status of the corresponding monitoring video. The greater the user's need for monitoring risk identification, the smaller the preset proportion of Class I video frames; provide a value-taking method, detect the average value of the proportion of Class I video frames in the historical records that meet the user's needs, and record it as the current preset proportion of Class I video frames; provide a preset proportion of Class I video frames, and the preset proportion of Class I video frames is 30%. The value of the preset density of Class I video frames can be set by the user according to the actual application scenario. In the present invention, the preset density of Class I video frames can effectively reflect the personnel activity degree of the corresponding monitoring video. Provide a value, and the preset density of Class I video frames is 0.6.

[0062] Specifically, when the video optimization unit performs compensation optimization for a single community monitoring device, this community monitoring device is recorded as the target analysis device;

[0063] Detect the radiation monitoring devices whose associated influence coefficients corresponding to the target analysis device are greater than the preset associated influence coefficient;

[0064] Perform video frame enlargement extraction on the monitoring video of the dense video period corresponding to the target analysis device;

[0065] Perform video frame enlargement extraction on the monitoring video of the sparse video period corresponding to the radiation monitoring device.

[0066] The radiation monitoring device is a community monitoring device whose associated influence coefficient corresponding to the target analysis device is greater than the preset associated influence coefficient, and there is at least one intercommunication path between the monitoring area corresponding to the radiation monitoring device and the target analysis device. The intercommunication path is a path for personnel to move, and the path length should be less than the preset area distance. Whether there is an intercommunication path, the number of intercommunication paths, and the path length are determined by the user in advance. The preset area distance is set according to the user's needs. It can be understood that when people are flowing, the people in the monitoring area corresponding to the target analysis device are likely to flow to the surrounding areas, thus increasing the monitoring requirements for the surrounding areas. In the implementation of the present invention, a value of the preset area distance is preferably provided, and the preset area distance is 200m.

[0067] The associated influence coefficient is equal to the number of intercommunication paths between the target analysis device and the radiation monitoring device. The value of the preset associated influence coefficient can be set by the user according to the actual application scenario. In the present invention, the associated influence coefficient reflects the degree of influence of the target analysis device on the personnel flow between the radiation monitoring device. The greater the user's requirement for monitoring timeliness, the smaller the value of the preset associated influence coefficient. A method for obtaining the value of the preset associated influence coefficient is provided. The associated influence coefficients in the historical records that meet the user's needs are statistically analyzed, and the outliers are removed. The average value of the associated influence coefficients after removing the outliers is recorded as the current preset associated influence coefficient.

[0068] The method for confirming the intensive video period is as follows: extract the first type-I video frame and the last type-I video frame in the time sequence in the monitoring video corresponding to the current detection period, and record the time period between the video times corresponding to the two type-I video frames as the intensive video period; record the time period other than the intensive video period as the sparse video period; for example, when the current monitoring period is from 12:00 to 1:00 (including 1:00, not including 12:00), at the end of the monitoring period from 12:00 to 1:00, if the first type-I video frame appears at 12:10 and the last type-I video frame appears at 12:30, then the time period from 12:10 to 12:30 (including 12:10 and 12:30) is recorded as the intensive video period, and the time period from 12:30 to 1:00 (including 12:30 and 1:00) is recorded as the sparse video period. For the monitoring video segment from 12:10 to 12:30 corresponding to the target analysis device, video frame extraction is increased, and for the monitoring video segment from 12:30 to 1:00 corresponding to the radiation monitoring device, video frame extraction is increased.

[0069] Specifically, during the process of increasing video frame extraction, the video optimization unit respectively detects the personnel flow coefficients of the target analysis device and the radiation monitoring device;

[0070] The amount of increased video frames is determined according to the difference between the personnel flow coefficient and the preset personnel flow coefficient.

[0071] Among them, the difference between the personnel flow coefficient and the preset personnel flow coefficient is denoted as V, and V = personnel flow coefficient - preset personnel flow coefficient. When performing video frame enlargement extraction for the dense video period corresponding to the target analysis device, the method for confirming the personnel flow coefficient is to detect the number of personnel in the basic video frame of the dense video period corresponding to the monitoring video of the target analysis device, and record the largest number of personnel as the personnel flow coefficient; when performing video frame enlargement extraction for the sparse video period corresponding to the radiation monitoring device, the method for confirming the personnel flow coefficient is to detect the number of personnel in the basic video frame of the sparse video period corresponding to the monitoring video of the radiation monitoring device, and calculate the difference in the number of personnel between each basic video frame and the nearest basic video frame before it in chronological order in turn, and record the average value of the differences in the number of personnel as the personnel flow coefficient; when the personnel flow coefficient is less than or equal to the preset personnel flow coefficient, the video frame enlargement amount is denoted as the minimum video frame enlargement amount; when the personnel flow coefficient is greater than the preset personnel flow coefficient, the difference between the personnel flow coefficient and the preset personnel flow coefficient has a positive correlation with the video frame enlargement amount; the value-taking methods of the preset personnel flow coefficient and the minimum video frame enlargement amount are that users can set according to the actual application scenario. The embodiments of the present invention preferably provide a value-taking of the preset personnel flow coefficient and the minimum video frame enlargement amount. The preset personnel flow coefficient is taken as 5, and the minimum video frame enlargement amount is 10 frames. During the enlargement extraction process, the time intervals between the extracted adjacent video frames are the same.

[0072] Specifically, when the video optimization unit performs activity feature optimization, it includes:

[0073] Determine the risk evaluation value of each basic video frame according to the activity amplitude coefficient and the position anomaly coefficient of the basic video frame;

[0074] If the risk evaluation value corresponding to the basic video frame is greater than the preset risk evaluation value, then the video frame is an activity warning video frame, and a preset number of video frames are extracted for the warning range of each activity warning video frame;

[0075] The preset number has a positive correlation with the risk evaluation value.

[0076] The confirmation method of the activity amplitude coefficient is as follows: By using YOLOv8-pose to detect the positions and movement trajectories of the human body's skeletal key points, analyze the rotation angle v and translation displacement d of the bones, preset the basic value v0 of the bone rotation angle and the basic value d0 of the translation displacement, and the activity amplitude coefficient = (v / v0) + (d / d0). Provide a method for obtaining the values of the preset basic value v0 of the bone rotation angle and the preset basic value d0 of the translation displacement, detect the basic values of the bone rotation angle and the translation displacement that meet the user's requirements in the historical records that meet the user's requirements, and record the respective average values of the basic value of the bone rotation angle and the basic value of the translation displacement as the preset basic value of the bone rotation angle and the preset basic value of the translation displacement.

[0077] The confirmation method of the position anomaly coefficient is as follows: Obtain the target positions of each person within the basic video frame, where the target position is the position where the person's head is located, detect the minimum distance between each person's target position and the reference plane, calculate the average value of all the minimum distances, calculate the difference between each person's minimum distance and this average value, and record the average value of the differences between each person's minimum distance and this average value as the position anomaly coefficient.

[0078] The calculation formula for the risk evaluation value is:

[0079] Risk evaluation value = K1×α1 + K2×α2;

[0080] Among them, K1 is the activity amplitude coefficient, K2 is the position anomaly coefficient, α1 is the first weight coefficient, α2 is the second weight coefficient. The values of α1 and α2 can be obtained by the user through the deep learning network and historical records. How to obtain training samples through historical records and learn the weight coefficients is content that those skilled in the art can easily understand and will not be elaborated here. Provide a preferred value in the implementation of the present invention, α1 = 0.6, α2 = 0.4.

[0081] For the value of the preset risk evaluation value, the user can set it according to the actual activity situation of the person. Provide a method for obtaining the value, detect the average value of the risk evaluation values in the historical records that meet the user's requirements, and record it as the preset risk evaluation value.

[0082] For a single activity warning video frame, its corresponding warning range is a video segment with this activity warning video frame as the midpoint, and the duration of the video segment is one-tenth of the monitoring period. The user can set the duration of the video segment according to the actual duration of the monitoring period and the monitoring accuracy requirements. The longer the duration of the monitoring period, the smaller the duration of the video segment; the smaller the monitoring accuracy requirements, the smaller the duration of the video segment.

[0083] Specifically, the key frame optimization unit performs effective analysis on each extracted video frame respectively;

[0084] The effective analysis includes determining the effective state of the extracted video frames according to the dynamic and static imbalance coefficients and the image quality coefficients corresponding to the extracted video frames, and determining the corresponding processing method according to the effective state.

[0085] The extracted video frames include one or more of the basic video frames, the video frames obtained by enlarged extraction, and the preset number of video frames extracted within the warning range of each activity warning video frame.

[0086] The method for confirming the dynamic and static imbalance coefficient is to perform person detection on each extracted video frame. A pre-trained object detection model can be used to detect the people in the video. Then, the object tracking algorithm is used to track the detected people to ensure that the same person can be recognized in different frames. For each detected person, the acceleration is detected, and the average acceleration of all people is calculated;

[0087] The calculation formula for the dynamic imbalance coefficient Cp is:

[0088]

[0089] where Ci is the ith acceleration detected, C0 is the average acceleration, i = 1, 2, 3,..., n, and n is the number of all accelerations detected.

[0090] The method for confirming the image quality coefficient is to detect the blurred areas of each extracted video frame through a pre-trained deep learning model, and record the average value of the areas of the blurred areas as the image quality coefficient.

[0091] The identification method of the blurred area can also be, but not limited to, gradient blur detection and blur detection based on the Laplacian operator.

[0092] For the values of the preset image quality coefficient and the preset dynamic and static imbalance coefficient, the user can set them according to the actual application scenario. The user can extract the image quality coefficient and the dynamic and static imbalance coefficient in the historical records that meet the requirements, and eliminate the abnormal values among them. The average values corresponding to the image quality coefficient and the dynamic and static imbalance coefficient after elimination are respectively recorded as the preset image quality coefficient and the preset dynamic and static imbalance coefficient.

[0093] Specifically, in response to the effective state that the image quality coefficient is less than or equal to the preset image quality coefficient or the dynamic and static imbalance coefficient is greater than the preset dynamic and static imbalance coefficient, the key frame optimization unit determines that the processing method is to supplement the associated frames for the extracted video frames;

[0094] The number of supplemented associated frames has a positive correlation with the effective difference.

[0095] The associated frames are supplemented by extracting the same number of video frames before and after the time sequence of the extracted video frames respectively, which are denoted as associated frames. Each associated frame has at least one associated frame with a time interval smaller than that of any other video frame.

[0096] Specifically, the key frame optimization unit determines the processing method as generating an image description and generating keywords for the extracted video frames and the associated frames in response to the effective state where the dynamic-static imbalance coefficient is less than or equal to the preset dynamic-static imbalance coefficient and greater than the preset dynamic-static imbalance coefficient.

[0097] In the present invention, keyword generation generates a text description by using the trained model for the extracted video frames and the associated frames. Model construction is required for keyword generation, including:

[0098] Obtain an image-text pair dataset, preprocess the images, tokenize the text, and construct a vocabulary.

[0099] Use a pre-trained CNN to extract image features, use an RNN / LSTM or Transformer decoder to generate text, input the image features into the decoder to generate text, and optimize the model using the cross-entropy loss function.

[0100] Specifically, the video analysis unit determines the video frame category of the basic video frames once per monitoring cycle.

[0101] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0102] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A community data management system for a smart elderly care community, characterized in that: include: A video surveillance unit, which includes several community surveillance devices for collecting surveillance videos of each surveillance area; A video analysis unit connected to the video monitoring unit, for periodically determining a video frame category of each basic video frame according to a personnel change reference value and a personnel impact reference value of the basic video frame corresponding to the community monitoring device; A video optimization unit, which is connected to the video monitoring unit and the video analysis unit respectively, and is used to determine the state of the monitoring device according to the proportion of a type of video frame and a type of video frame density corresponding to the community monitoring device, and determine the corresponding video frame optimization method as compensation optimization or activity feature optimization according to the state of the monitoring device; The key frame optimization unit is connected to the video monitoring unit, the video analysis unit and the video optimization unit respectively, and is used to determine the processing method according to the effective state to supplement or generate image description for the associated frame.

2. The community data management system for a smart elderly care community according to claim 1, characterized in that: The video analysis unit determines the video frame category according to the personnel change reference value and the personnel impact reference value of the basic video frame corresponding to the community monitoring device. The video frame categories include: A type of video frame whose personnel change reference value is greater than a preset personnel change reference value or whose personnel impact reference value is greater than a preset personnel impact reference value; The second type of video frame has a personnel change reference value that is less than or equal to a preset personnel change reference value and a personnel impact reference value that is less than or equal to a preset personnel impact reference value.

3. The community data management system for a smart elderly care community according to claim 2, characterized in that: The video optimization unit determines the monitoring device status of each community monitoring device according to the proportion of a type of video frame and the density of a type of video frame corresponding to each community monitoring device, and determines the corresponding video frame optimization method according to the monitoring device status of the community monitoring device; If the monitoring device status is that the proportion of a type of video frame is greater than the preset proportion of a type of video frame and the density of a type of video frame is greater than the preset density of a type of video frame, the video frame optimization method is compensation optimization; If the status of the monitoring device is that the proportion of a type of video frame is less than or equal to the preset proportion of a type of video frame or the density of a type of video frame is less than or equal to the preset density of a type of video frame, the video frame optimization method is activity feature optimization.

4. The community data management system for a smart elderly care community according to claim 3, characterized in that: When the video optimization unit performs compensation optimization on a single community monitoring device, the community monitoring device is recorded as a target analysis device; Detecting radiation monitoring equipment whose correlation influence coefficient corresponding to the target analysis equipment is greater than a preset correlation influence coefficient; Perform video frame enlargement and extraction on the surveillance video of the intensive video period corresponding to the target analysis device; Video frame enlargement and extraction are performed on the monitoring video of the sparse video period corresponding to the radiation monitoring equipment.

5. The community data management system for a smart elderly care community according to claim 4, characterized in that: The video optimization unit detects the personnel flow coefficients of the target analysis equipment and the radiation monitoring equipment respectively during the process of increasing the video frame extraction; The video frame increase amount is determined according to the difference between the personnel flow coefficient and the preset personnel flow coefficient.

6. The community data management system for a smart elderly care community according to claim 5, characterized in that: When the video optimization unit optimizes activity features, it includes: Determine the risk assessment value of each basic video frame according to the activity amplitude coefficient and the position abnormality coefficient of the basic video frame; If the risk assessment value corresponding to the basic video frame is greater than the preset risk assessment value, the video frame is an activity warning video frame, and a preset number of video frames are extracted for the warning range of each activity warning video frame; The preset quantity is positively correlated with the risk assessment value.

7. The community data management system for a smart elderly care community according to claim 6, characterized in that: The key frame optimization unit performs effective analysis on each extracted video frame; The effective analysis includes determining the effective state of the extracted video frame according to the dynamic and static imbalance coefficients and the image quality coefficients corresponding to the extracted video frame, and determining the corresponding processing method according to the effective state.

8. The community data management system for a smart elderly care community according to claim 7, characterized in that: The key frame optimization unit responds to the valid state that the image quality coefficient is less than or equal to the preset image quality coefficient or the dynamic and static imbalance coefficient is greater than the preset dynamic and static imbalance coefficient, and determines the processing method to supplement the associated frame for the extracted video frame; The number of supplemented associated frames is positively correlated with the effective difference.

9. The community data management system for a smart elderly care community according to claim 8, characterized in that: The key frame optimization unit responds to the valid state that the dynamic-static imbalance coefficient is less than or equal to the preset dynamic-static imbalance coefficient and the dynamic-static imbalance coefficient is greater than the preset dynamic-static imbalance coefficient, and determines the processing method to generate image description and generate keywords for the extracted video frames and related frames.

10. The community data management system for a smart elderly care community according to claim 9, characterized in that: The video analysis unit determines the video frame category of the basic video frame once in each monitoring cycle.

Images