Data acquisition method and system for smart building and storage medium

Abstract

The invention relates to the field of data collection and processing, in particular to a data collection method and system for a smart building and a storage medium, and the method comprises the steps: periodically collecting monitoring videos of corresponding regions through a plurality of camera units, and recognizing the recognition information of each target object in each monitoring video; obtaining a current behavior probability value of each target object, and extracting a key image frame and an early warning video of the monitoring video based on the current behavior probability value of each target object; marking the key image frame and the early warning video, and mapping the key image frame and the early warning video to a three-dimensional map of the smart building according to the position of the camera unit; obtaining a complete moving path diagram of each target object; and sending the complete monitoring path diagram corresponding to the target object to a data receiving end. The problems that the monitoring data volume is large and the early warning information cannot be obtained in time are solved, the monitoring data volume is reduced, and the early warning information can be obtained in time.

Description

Technical Field

[0001] The present application relates to the field of data acquisition and processing, and in particular to a data acquisition method, system and storage medium for smart buildings. Background Art

[0002] Security monitoring is a vital link in modern smart buildings. With the development of technology, video surveillance systems have become a core part of building security management. However, existing video surveillance systems face some challenges and problems. Existing video surveillance systems usually generate a large amount of monitoring data. This data needs to be stored, processed and analyzed to identify and respond to potential security threats. However, due to the huge amount of data, traditional monitoring systems often find it difficult to process and analyze all data in a timely manner, resulting in delays in early warning information and even possible missing of key security events. Similar prior art includes a Chinese patent with publication number CN113310468A, which proposes a spatiotemporal information acquisition system, method and storage medium for monitoring targets. The system includes a video acquisition module for acquiring video image information of a monitoring area; a positioning module for locating the position information of a monitoring device; a distance perception module for calculating the distance information between the monitoring target and the monitoring device; a direction perception module for obtaining the direction information when the video acquisition module acquires the monitoring video; a time acquisition module for acquiring the current time information; a central processing module for processing the acquired video image information, position information, distance information, direction information and time information to obtain the spatiotemporal information of the monitoring target, and displaying the spatiotemporal information and video image information in real time. The spatiotemporal information acquisition system for the monitoring target can quickly acquire the four-dimensional spatiotemporal information of the monitoring target, quickly locate the geographical location of the monitoring target in the map, and quickly find the target object. In addition, similar prior art includes a US patent with publication number US11727580B2, which is used to collect information about objects moving in an area of ​​interest by using artificial intelligence-based video processing and analysis, object detection, and tracking moving objects from video frame to frame. The system, method, and computer can be implemented as a platform including an analysis dashboard and a backend dashboard. The analysis dashboard can display the camera sources being monitored and the analysis results of these sources over a period of time to the user through a UI. The backend dashboard allows the user to set the camera input by himself and post-process the pre-captured video files using the algorithm provided by the system, and the user can choose from them. Both of the above patent documents solve the video data collection method, but with the increase in the number of surveillance cameras, how to effectively manage and integrate data from multiple cameras becomes a problem. Existing monitoring systems often lack intelligent processing capabilities, and the huge amount of data cannot automatically identify and extract key information from video data, which results in security personnel being unable to obtain early warning information in a timely manner. Summary of the invention

[0003] The present application provides a data collection method for a smart building, the method comprising: Step S1: periodically collecting surveillance videos of corresponding areas through multiple camera units, identifying multiple target objects in each surveillance video through a recognition model, and obtaining identification information of each target object; Step S2: obtaining a current behavior probability value of each target object according to the identification information of each target object, the surveillance video and the behavior probability distribution of each target object, and extracting a key image frame and a warning video of the surveillance video based on the current behavior probability value of each target object and a first similarity between the behavior feature of the target object and a preset behavior feature; Step S3: marking the identification information of the target object, the acquisition time and the position information of the camera unit corresponding to the key image frame and the warning video, and mapping the position of the camera unit to the three-dimensional map, marking the key image frame and the warning video in the three-dimensional map, and storing them in the storage node corresponding to the camera unit; Step S4: Display the key image frames and the warning videos corresponding to each camera unit in the three-dimensional map in real time on the time coordinate axis according to the acquisition time and the corresponding time period, obtain the monitoring movement path of each target object, and mark the danger level of the target object in the three-dimensional map, and supplement the monitoring movement path of each target object in the blind spot of each camera unit, so as to obtain a complete monitoring path map.

[0004] As a preferred technical solution of the present invention, step S1 includes: The monitoring video of the corresponding area is periodically collected by multiple camera units in the smart building, and each video frame of the target object is separated from the monitoring video, and the video frame corresponding to each target object is input into the recognition model to obtain the recognition information of the target object, wherein the recognition model is a learning model trained with target data, and the target data is the physiological characteristics and motion characteristics of any target object that has appeared in the camera unit.

[0005] As a preferred technical solution of the present invention, step S2 includes: Step S21: extracting behavior features of each of the target objects through the surveillance video, wherein the behavior features include facial behavior features and body behavior features, obtaining the behavior type of the target object based on the behavior features of the target object and the behavior data corresponding to the target object stored in the storage unit through the identification information, and obtaining the current behavior probability value of the target object based on the behavior probability distribution of the target object and the behavior type; Step S22: When the current behavior probability value of each target object is greater than or equal to the first threshold, a random number generation unit randomly generates N time points within the corresponding time period of the surveillance video, and uses N surveillance images corresponding to the N time points as the key image frames; When the current behavior probability value of any of the target objects is less than the first threshold, the target object is taken as a suspected object, and the behavior characteristics of the suspected object are compared with the characteristics of the preset behavior, and a first similarity is obtained. When the first similarity is less than the set threshold, M surveillance images are extracted from the surveillance video on average as the key image frames, and the value of M is increased or decreased by increasing or decreasing the first similarity. When the first similarity is greater than or equal to the set threshold, the surveillance video is saved, and an early warning message is sent to other camera units and a receiving terminal, wherein the early warning message includes the identification information of the suspected object, the current image of the suspected object and an early warning mark, and the surveillance video including the target object captured by the camera unit and the other camera units is taken as the early warning video, and the suspected object is rendered in the early warning video, wherein N is less than M.

[0006] As a preferred technical solution of the present invention, when the recognition model cannot obtain the recognition information of the target object, the feature information of the target object is extracted through the surveillance video including the target object, and the feature information includes facial features, body features and behavioral features. The feature information of the target object is also added to the recognition model, and the recognition information is set for the target object, and the recognition information and the feature information of the target object are sent to other camera units.

[0007] As a preferred technical solution of the present invention, step S3 includes: Step S31: marking the identification information of each target object and the position information of the corresponding camera unit in each key image frame and the warning video, and marking the acquisition time and the corresponding time of each video frame in the key image frame and the warning video respectively; Step S32: constructing a three-dimensional map of the smart building based on the design structure of the smart building, and mapping the position of each camera unit into the three-dimensional map; Step S33: updating the key image frame and the warning video corresponding to each camera unit to the three-dimensional map in real time, and storing them at the storage node position corresponding to each camera unit.

[0008] As a preferred technical solution of the present invention, step S4 includes: Step S41: mapping the key image frames acquired by each camera unit in the three-dimensional map on the time coordinate axis according to the acquisition time, and also mapping the warning video acquired by each camera unit on the time coordinate axis; Step S42: connecting the key image frames corresponding to the same target object or the position information of the camera unit corresponding to the warning video in the three-dimensional map in chronological order, obtaining the monitored moving path of the target object, and marking the target object with a dangerous degree in the three-dimensional map, wherein the higher the dangerous degree, the more eye-catching the color of the monitored moving path; Step S43: The monitoring movement path of each target object is supplemented and completed for the blind spot of each camera unit according to the design structure of the smart building, so as to obtain the complete monitoring path map including the monitoring movement paths of all the target objects.

[0009] As a preferred technical solution of the present invention, the step S4 further includes a step S5: When the communication load is less than or equal to a second threshold, the three-dimensional monitoring map composed of the complete monitoring path map of all the target objects is sent to the data collection end in real time. When the communication load is greater than the second threshold, the sending priority of the corresponding complete monitoring path map is set according to the danger level of each target object, wherein the higher the danger level, the higher the sending priority.

[0010] As a preferred technical solution of the present invention, when the current behavior probability value of the target object is greater than or equal to a first threshold, the corresponding danger level of the target object is level three danger; when the current behavior probability value of the target object is less than the first threshold, when the first similarity is less than a set threshold, the danger level of the target object is level two danger; when the first similarity is greater than or equal to the set threshold, the danger level of the target object is level one danger.

[0011] The present invention also provides a data acquisition system for smart buildings, the system is used to implement the above method, the system comprises: Multiple camera units are used to periodically collect surveillance videos of corresponding areas; An identification unit, used to identify multiple target objects in each of the surveillance videos through a recognition model, and obtain identification information of each of the target objects; A collection unit, used for obtaining a current behavior probability value of each target object according to the identification information of each target object, the monitoring video and the behavior probability distribution of each target object, and extracting a key image frame and a warning video of the monitoring video based on the current behavior probability value of each target object and a first similarity between the behavior feature of the target object and a preset behavior feature; A mapping unit, used to mark the identification information of the target object, the acquisition time and the position information of the camera unit in the key image frame and the warning video, and a three-dimensional map of the smart building, and map the position of the camera unit to the three-dimensional map, mark the key image frame and the warning video in the three-dimensional map, and store them in the storage node corresponding to the camera unit; A generating unit is used to display the key image frames and the warning videos corresponding to each of the camera units in the three-dimensional map in real time on the time coordinate axis according to the acquisition time and the corresponding time period, and obtain the monitoring movement path of each of the target objects, and mark the degree of danger of the target objects in the three-dimensional map, and also supplement the monitoring movement path of each of the target objects in the blind area of ​​each of the camera units, so as to obtain a complete monitoring path map.

[0012] The present invention also provides a computer-readable storage medium, on which instructions are stored, and the above method is implemented when the instructions are executed by a processor.

[0013] Technical Effects The present invention periodically collects monitoring videos through multiple camera units, and uses a recognition model to identify the target object in the video, obtains its recognition information, extracts key image frames and warning videos by analyzing the behavior probability value of the target object, and annotates these key image frames and maps them to the three-dimensional map of the smart building, thereby obtaining a complete moving path diagram of the target object. This method effectively solves the problem that the amount of monitoring data is large and the warning information cannot be obtained in time, reduces the amount of monitoring data and can provide warning information in time, and intelligently screens out key image frames through similarity analysis of behavior probability values ​​and behavior characteristics, reduces the pressure of data storage and processing, and improves monitoring efficiency at the same time. When the recognition model cannot obtain the target object information, the system can extract the characteristic information of the target object, and update the recognition model to enhance the recognition ability of the system. In addition, by annotating the key image frames and warning videos in the three-dimensional map, it is convenient to update and query in real time, improves the availability and intuitiveness of the monitoring data, and dynamically adjusts the data sending strategy according to the communication load and the danger level of the target object, ensures that the information of the target object with high danger level is transmitted preferentially when the communication pressure is high, and improves the efficiency of emergency response. Through the mutual coordination of the above-mentioned technical solutions, through intelligent processing and analysis of monitoring video data, the amount of data collection and transmission is reduced, and key data is retained to ensure timely acquisition of early warning information, realize effective monitoring and management of target objects in smart buildings, improve the intelligence level and response speed of the monitoring system, and provide strong support for the safe operation of smart buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0015] Figure 1 This is a flow chart of a data collection method for smart buildings in an embodiment of the present application; Figure 2 This is a structural diagram of a data acquisition system for smart buildings in an embodiment of the present application. DETAILED DESCRIPTION

[0016] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 As shown, the data collection method for smart buildings in the embodiment of the present application includes: Step S1: periodically collecting surveillance videos of corresponding areas through multiple camera units, identifying multiple target objects in each surveillance video through a recognition model, and obtaining recognition information of the multiple target objects; Specifically, the plurality of camera units arranged in the smart building periodically collect the monitoring video in the corresponding area, separate the video frames of each target object from the monitoring video, input each video frame into the recognition model, and obtain the recognition information of the target object. Through the technical solution, the recognition information of each target object can be recognized based on the monitoring video, thus laying a foundation for further obtaining the behavior of each target object.

[0018] Step S2: obtaining a current behavior probability value of each target object according to the identification information of each target object, the surveillance video and the behavior probability distribution of each target object, and extracting a key image frame and a warning video of the surveillance video based on the current behavior probability value of each target object and a first similarity between the behavior feature of the target object and a preset behavior feature; Specifically, behavioral features of each of the target objects are extracted through the surveillance video, the behavioral features include facial behavioral features and limb behavioral features, the behavioral features are compared with the behavioral data of the target object to obtain the behavior type of the target object, and according to the behavior type of the target object and the behavior probability distribution of the target object, a current behavior probability value corresponding to the current behavior type of the target object is obtained. When the current behavior probability values ​​are greater than or equal to the first threshold, it is considered that each of the target objects is in a normal state. Therefore, N surveillance images corresponding to N randomly generated time points are used as the key image frames. When the current behavior probability value of any of the target objects is less than When the first threshold is reached, the target object whose current behavior probability value is less than the first threshold is taken as the suspected object, and the behavior characteristics of the suspected object are compared with the characteristics of the preset behavior, and the first similarity is obtained, and M monitoring images are extracted according to the first similarity as the key image frames. When the first similarity is greater than the set threshold, it is considered that the current behavior of the suspected object is more harmful. At this time, the monitoring video is saved and the warning information is sent to other monitoring units. Through the technical solution, not only can the corresponding key image frames be obtained according to the current behavior probability value of the target object and the first similarity with the preset behavior, but also the amount of data can be reduced, and warnings can be issued for suspected objects with higher risks.

[0019] Step S3: marking the identification information of the target object, the acquisition time and the position information of the camera unit corresponding to the key image frame and the warning video, and mapping the position of the camera unit to the three-dimensional map, marking the key image frame and the warning video in the three-dimensional map, and storing them in the storage node corresponding to the camera unit; Specifically, by marking the identification information of each target object and the location information of the corresponding camera unit in each of the above-mentioned key image frames and warning videos, the acquisition time of the key image frames and the corresponding time of each video frame in the warning video are also marked, and a three-dimensional map of the above-mentioned smart building is constructed based on the design structure of the above-mentioned smart building, and each of the above-mentioned camera units is updated to the above-mentioned three-dimensional map, and the key image frames and warning videos corresponding to each of the above-mentioned camera units are also updated to the above-mentioned three-dimensional map, and the key image frames and the above-mentioned warning videos are stored in the storage nodes of the corresponding camera units, so as to facilitate the analysis of the movement path of each target object. Through the above-mentioned technical solution, a foundation is laid for further analysis and acquisition of the movement path map of each target object.

[0020] Step S4: Displaying the key image frames and the warning video corresponding to each camera unit in the three-dimensional map in real time on the time coordinate axis according to the acquisition time and the corresponding time period, and obtaining the monitoring movement path of each target object, and marking the danger level of the target object in the three-dimensional map, and supplementing the monitoring movement path of each target object with the blind spot of each camera unit, so as to obtain a complete monitoring path map; Specifically, the key image frames corresponding to each camera unit in the three-dimensional map are mapped onto the time coordinate axis according to the acquisition time, and the warning videos corresponding to each camera unit are also mapped onto the time coordinate axis. The position information of the target objects is connected in chronological order to obtain the monitored moving path of the target objects. The danger level of each target object is marked, and its monitored moving path is displayed according to its danger level. The integrity of the monitored moving path of each target object is supplemented through the design structure of the smart building. Through the technical solution, a complete monitoring path map including all target objects can be obtained, which is convenient for analyzing the moving path of the suspected object when warning information appears.

[0021] Furthermore, the step S1 comprises: The monitoring video of the corresponding area is periodically collected by multiple camera units in the smart building, and each video frame of the target object is separated from the monitoring video, and the video frame corresponding to each target object is input into the recognition model to obtain the recognition information of the target object, wherein the recognition model is a learning model trained with target data, and the target data is the physiological characteristics and motion characteristics of any target object that has appeared in the camera unit.

[0022] Specifically, the monitoring video in the corresponding area is periodically collected by the multiple camera units arranged in the smart building, and the video frames of each target object are separated from the monitoring video, and each video frame is input into the recognition model to obtain the recognition information of the target object, wherein the recognition model is a learning model trained by the target object, that is, when the target object appears in the first monitoring video for the first time, the physiological characteristics and behavioral characteristics of the target object are obtained through the first monitoring video, and the physiological characteristics include facial features, height and limb proportions, and movement characteristics such as gait characteristics and habitual limb movement characteristics, and the facial features and movement characteristics of the target object are used as target data to train the recognition model. The recognition model can also generate multi-angle recognition images of the target object through the target data of the target object, and generate recognition information for the target object. When the video frame is input into the recognition model, the target object can be accurately identified even if the clarity of the video frame is not enough or it is partially blocked. Through the technical solution, the recognition information of each target object can be identified based on the monitoring video, laying a foundation for further obtaining the behavior of each target object.

[0023] Furthermore, the step S2 comprises: Step S21: extracting behavior features of each of the target objects through the surveillance video, wherein the behavior features include facial behavior features and body behavior features, obtaining the behavior type of the target object based on the behavior features of the target object and the behavior data corresponding to the target object stored in the storage unit through the identification information, and obtaining the current behavior probability value of the target object based on the behavior probability distribution of the target object and the behavior type; Specifically, behavioral characteristics of each of the target objects are extracted through the surveillance video, the behavioral characteristics include facial behavioral characteristics and limb behavioral characteristics, the facial behavioral characteristics include at least: tension, relaxation, anxiety, anger and happiness, the limb characteristics include at least gestures, speed of limb movements and repetitiveness of limb movements, and also include the distance between the target object and other people, and the behavioral characteristics are compared with the behavioral data of the target object to obtain the behavior type of the target object, and according to the behavior type of the target object and the behavior probability distribution of the target object, the probability value corresponding to the current behavior type of the target object, that is, the current behavior probability value, is obtained. Through the technical solution, the current behavior probability value of each of the target objects can be accurately obtained, thereby laying a foundation for determining the number of key image frames obtained in the corresponding time period of the surveillance video according to the current behavior probability value.

[0024] Step S22: When the current behavior probability value of each target object is greater than or equal to the first threshold, a random number generation unit randomly generates N time points within the corresponding time period of the surveillance video, and uses N surveillance images corresponding to the N time points as the key image frames; When the current behavior probability value of any of the target objects is less than the first threshold, the target object is taken as a suspected object, and the behavior characteristics of the suspected object are compared with the characteristics of the preset behavior, and a first similarity is obtained. When the first similarity is less than the set threshold, M surveillance images are extracted on average from the surveillance video as the key image frames, and the value of M is increased or decreased by increasing or decreasing the first similarity. When the first similarity is greater than or equal to the set threshold, the surveillance video is saved, and an early warning message is sent to other camera units, wherein the early warning message includes the identification information of the suspected object, the current image of the suspected object and the early warning mark, and the surveillance video including the target object captured by the camera unit and the other camera units is taken as the early warning video, and the suspected object is rendered in the early warning video, and the rendered early warning video is sent to the receiving end, wherein N is less than M.

[0025] Specifically, when the current behavior probability value of each of the target objects in the surveillance video is greater than or equal to the first threshold, that is, the current behavior probability value of each of the target objects is consistent with the usual behavior, then it is considered that each of the target objects is in a normal state. Therefore, the random number generation unit randomly generates N time points within the corresponding time period of the surveillance video, and uses the N surveillance images corresponding to the N time points as the key image frames. Since the N surveillance images are randomly acquired, they can better reflect the behavior process of each target object within the time period, which is convenient for playback and viewing, and also reduces the amount of data. The value of N is proportional to the length of the corresponding time period of the surveillance video. In the surveillance video, when the current behavior probability value of any of the target objects is less than the first threshold, the target object whose current behavior probability value is less than the first threshold is regarded as the suspected object, that is, the behavior of the suspected object is greatly different from the usual behavior, but at this time it cannot be said that the current behavior of the suspected object is suspicious. Therefore, the behavior characteristics of the suspected object are also compared with the characteristics of the preset behavior, wherein the preset behavior, for example, sabotage, forced entry into unauthorized areas, attempts to conceal the true identity, etc., and a first similarity is obtained. When the first similarity is less than the set threshold, it is considered that although the suspected object is greatly different from the usual behavior, the harm is not particularly large, and because of the above The greater the first similarity, the greater the harmfulness of the behavior of the suspected object. Therefore, when the first similarity changes, the value of M also changes. The greater the first similarity, the greater the value of M. Conversely, the smaller the first similarity, the smaller the value of M. In addition, M surveillance images are extracted at the same time interval as the key image frames, which can not only reflect the behavior process of the target object according to the current behavior dangerousness of the target object according to the key image frames, but also reduce the amount of data. When the first similarity is greater than the set threshold, it is considered that the current behavior of the suspected object is more harmful. At this time, the surveillance video is saved, and the warning information is sent to other monitoring units, so that other monitoring units can quickly Identify the above-mentioned suspected object, timely monitor the moving path of the above-mentioned target object, and issue a warning in time. Render the suspected object in the warning video, and send the rendered warning video to the above-mentioned receiving end, so that the user can take corresponding measures and collect evidence according to the above-mentioned warning video. When there are multiple suspected objects, adjust the value of M according to the maximum value of the first similarities corresponding to the multiple suspected objects, wherein the values ​​of N and M are both positive integers greater than or equal to 1. Through the above-mentioned technical solution, not only can the corresponding key image frame be obtained according to the current behavior probability value of the target object and the first similarity with the preset behavior to record and reflect the behavior process of the target object, but also the amount of data can be reduced, and warnings can be issued for suspected objects with higher risks.

[0026] Furthermore, when the recognition model cannot obtain the recognition information of the target object, the feature information of the target object is extracted through the surveillance video including the target object, and the feature information includes physiological features and behavioral features. Based on the feature information, the angle discrimination information of the target object is generated, and the feature information and the discrimination information of the target object are added to the recognition model for learning, and the recognition information of the target object is generated, and the recognition information and the feature information of the target object are sent to other camera units.

[0027] Specifically, when the above-mentioned recognition model cannot obtain the recognition information of the above-mentioned target object, that is, the above-mentioned target object may be entering the monitoring video corresponding to the above-mentioned camera unit for the first time, and the above-mentioned recognition model does not have the recognition features of the above-mentioned target object, therefore, the physiological characteristics and behavioral characteristics of the above-mentioned target object are extracted through the above-mentioned monitoring video, wherein the above-mentioned physiological characteristics include facial features, height and limb proportions, and movement characteristics such as: gait characteristics and habitual limb movement characteristics, and multi-angle identification information of the above-mentioned target object is generated through the above-mentioned physiological characteristics and behavioral characteristics. Through the above-mentioned identification information, even if the above-mentioned target object is not very clear in the monitoring video or the dress has changed, it can be accurately identified, and the identification information of the above-mentioned target object is also generated, and the above-mentioned identification information and feature information of the above-mentioned target object are sent to other camera units, so that other camera units have unified identification information for the same target object and can identify the above-mentioned target object in time. Through the above-mentioned technical solution, the above-mentioned feature information and identification information of the above-mentioned target object can be obtained when the above-mentioned target object appears for the first time, thereby laying the foundation for accurate and timely identification of the above-mentioned target object.

[0028] Furthermore, the step S3 comprises: Step S31: marking the identification information of each target object and the position information of the corresponding camera unit in each key image frame and the warning video, and marking the acquisition time and the corresponding time of each video frame in the key image frame and the warning video respectively; Step S32: constructing a three-dimensional map of the smart building based on the design structure of the smart building, and mapping the position of each camera unit into the three-dimensional map; Step S33: updating the key image frame and the warning video corresponding to each camera unit to the three-dimensional map in real time, and storing them at the storage node position corresponding to each camera unit.

[0029] Specifically, by marking the identification information of each target object and the location information of the corresponding camera unit in each of the above-mentioned key image frames and warning videos, the acquisition time of the key image frames and the corresponding time of each video frame in the warning video are also marked, and a three-dimensional map of the above-mentioned smart building is constructed based on the design structure of the above-mentioned smart building, and each of the above-mentioned camera units is updated to the above-mentioned three-dimensional map, and the key image frames and warning videos corresponding to each of the above-mentioned camera units are also updated to the above-mentioned three-dimensional map, and the key image frames and the above-mentioned warning videos are stored in the storage nodes of the corresponding camera units, so as to facilitate the analysis of the movement path of each target object. Through the above-mentioned technical solution, a foundation is laid for further analysis and acquisition of the movement path map of each target object.

[0030] Furthermore, the step S4 comprises: Step S41: mapping the key image frames acquired by each camera unit in the three-dimensional map on the time coordinate axis according to the acquisition time, and also mapping the warning video acquired by each camera unit on the time coordinate axis; Step S42: connecting the key image frames corresponding to the same target object or the position information of the camera unit corresponding to the warning video in the three-dimensional map in chronological order, obtaining the monitored moving path of the target object, and marking the target object with a dangerous degree in the three-dimensional map, wherein the higher the dangerous degree, the more eye-catching the color of the monitored moving path; Step S43: The monitoring movement path of each target object is supplemented and completed for the blind spot of each camera unit according to the design structure of the smart building, so as to obtain the complete monitoring path map including the monitoring movement paths of all the target objects.

[0031] Specifically, the key image frames corresponding to each camera unit in the three-dimensional map are mapped on the time coordinate axis according to the acquisition time, and the warning video corresponding to each camera unit is also mapped on the time coordinate axis, and the position information of each target object is extracted in time sequence through the key image frames and the warning video, and the position information of the target object is connected in time sequence to obtain the monitoring movement path of the target object, and the danger level of each target object is marked, and its monitoring movement path is displayed according to its danger level, and the more serious the danger level, the more eye-catching its monitoring movement path, which is convenient for the data receiving end user to view and take corresponding measures in time. However, since the camera unit may have a monitoring blind spot, it is impossible to obtain a complete monitoring movement path map of the target object. Therefore, each channel of the blind spot is obtained through the design structure of the smart building, so as to supplement the integrity of the monitoring movement path of each target object, and the supplementary path part is displayed by a dotted line, wherein there may be multiple supplementary paths. Through the above technical solution, a complete monitoring path map including all target objects can be obtained, which is convenient for analyzing the movement path of the suspected object when an early warning occurs.

[0032] Furthermore, the step S4 further includes a step S5: When the communication load is less than or equal to a second threshold, the three-dimensional monitoring map consisting of the complete monitoring path map of all the target objects is sent to the data collection end in real time. When the communication load is greater than the second threshold, the sending priority of the corresponding complete monitoring path map is set according to the danger level of each target object, wherein the higher the danger level, the higher the sending priority.

[0033] Specifically, when the communication load is less than or equal to the second threshold, that is, when the communication load is small, that is, when the communication pressure is small, a three-dimensional monitoring map composed of a complete monitoring path map of all the target objects is sent to the data collection end in real time; when the communication load is greater than the second threshold, that is, when the communication pressure is large, the complete monitoring path map corresponding to the target object is sent according to the sending priority of the target object; wherein, the higher the danger level of the target object, the higher the corresponding sending priority; through the above technical solution, the data collection unit can timely obtain the complete monitoring path map of the target object with a higher danger level, thereby facilitating the adoption of corresponding measures.

[0034] Furthermore, when the current behavior probability value of the target object is greater than or equal to the first threshold, the corresponding danger level of the target object is level three danger; when the current behavior probability value of the target object is less than the first threshold, when the first similarity is less than the set threshold, the danger level of the target object is level two danger; when the first similarity is greater than or equal to the set threshold, the danger level of the target object is level one danger.

[0035] The present invention also provides a data acquisition system for smart buildings, the system is used to implement the above method, such as Figure 2 As shown, the system comprises: Multiple camera units are used to periodically collect surveillance videos of corresponding areas; An identification unit, used to identify multiple target objects in each of the surveillance videos through a recognition model, and obtain identification information of each of the target objects; A collection unit, used for obtaining a current behavior probability value of each target object according to the identification information of each target object, the monitoring video and the behavior probability distribution of each target object, and extracting a key image frame and a warning video of the monitoring video based on the current behavior probability value of each target object and a first similarity between the behavior feature of the target object and a preset behavior feature; A mapping unit, used to mark the identification information of the target object, the acquisition time and the position information of the camera unit in the key image frame and the warning video, and a three-dimensional map of the smart building, and map the position of the camera unit to the three-dimensional map, mark the key image frame and the warning video in the three-dimensional map, and store them in the storage node corresponding to the camera unit; A generating unit is used to display the key image frames and the warning videos corresponding to each of the camera units in the three-dimensional map in real time on the time coordinate axis according to the acquisition time and the corresponding time period, and obtain the monitoring movement path of each of the target objects, and mark the degree of danger of the target objects in the three-dimensional map, and also supplement the monitoring movement path of each of the target objects in the blind area of ​​each of the camera units, so as to obtain a complete monitoring path map.

[0036] The present invention also provides a computer-readable storage medium, on which instructions are stored, and the above method is implemented when the instructions are executed by a processor.

[0037] In summary, the present invention periodically collects monitoring videos through multiple camera units, and uses the recognition model to identify the target object in the video, obtains its recognition information, extracts key image frames and warning videos by analyzing the behavior probability value of the target object, and annotates these key image frames and maps them to the three-dimensional map of the smart building, thereby obtaining a complete moving path diagram of the target object. This method effectively solves the problem that the amount of monitoring data is large and the warning information cannot be obtained in time, reduces the amount of monitoring data and can provide warning information in time, and intelligently screens out key image frames through similarity analysis of behavior probability values ​​and behavior characteristics, reduces the pressure of data storage and processing, and improves monitoring efficiency at the same time. When the recognition model cannot obtain the target object information, the system can extract the characteristic information of the target object, and update the recognition model to enhance the recognition ability of the system. In addition, by annotating the key image frames and warning videos in the three-dimensional map, it is convenient to update and query in real time, improves the availability and intuitiveness of the monitoring data, and dynamically adjusts the data sending strategy according to the communication load and the danger level of the target object, ensuring that the information of the target object with high danger level is transmitted first when the communication pressure is high, and improves the efficiency of emergency response. Through the mutual coordination of the above-mentioned technical solutions, through intelligent processing and analysis of monitoring video data, the amount of data collection and transmission is reduced, and key data is retained to ensure timely acquisition of early warning information, realize effective monitoring and management of target objects in smart buildings, improve the intelligence level and response speed of the monitoring system, and provide strong support for the safe operation of smart buildings.

[0038] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0039] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0040] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A data collection method for smart buildings, characterized in that: The method comprises: Step S1: periodically collecting surveillance videos of corresponding areas through multiple camera units, identifying multiple target objects in each surveillance video through a recognition model, and obtaining identification information of each target object; Step S2: obtaining a current behavior probability value of each target object according to the identification information of each target object, the surveillance video and the behavior probability distribution of each target object, and extracting a key image frame and a warning video of the surveillance video based on the current behavior probability value of each target object and a first similarity between the behavior feature of the target object and a preset behavior feature; Step S3: marking the identification information of the target object, the acquisition time and the position information of the camera unit corresponding to the key image frame and the warning video, and mapping the position of the camera unit to the three-dimensional map, marking the key image frame and the warning video in the three-dimensional map, and storing them in the storage node corresponding to the camera unit; Step S4: Display the key image frames and the warning videos corresponding to each camera unit in the three-dimensional map in real time on the time coordinate axis according to the acquisition time and the corresponding time period, obtain the monitoring movement path of each target object, and mark the danger level of the target object in the three-dimensional map, and supplement the monitoring movement path of each target object in the blind spot of each camera unit, so as to obtain a complete monitoring path map.

2. The method according to claim 1, characterized in that The step S1 comprises: The monitoring video of the corresponding area is periodically collected by multiple camera units in the smart building, and each video frame of the target object is separated from the monitoring video, and the video frame corresponding to each target object is input into the recognition model to obtain the recognition information of the target object, wherein the recognition model is a learning model trained with target data, and the target data is the physiological characteristics and motion characteristics of any target object that has appeared in the camera unit.

3. The method according to claim 1, characterized in that , the step S2 comprises: Step S21: extracting behavior features of each of the target objects through the surveillance video, wherein the behavior features include facial behavior features and body behavior features, obtaining the behavior type of the target object based on the behavior features of the target object and the behavior data corresponding to the target object stored in the storage unit through the identification information, and obtaining the current behavior probability value of the target object based on the behavior probability distribution of the target object and the behavior type; Step S22: When the current behavior probability value of each target object is greater than or equal to the first threshold, a random number generation unit randomly generates N time points within the corresponding time period of the surveillance video, and uses N surveillance images corresponding to the N time points as the key image frames; When the current behavior probability value of any of the target objects is less than the first threshold, the target object is taken as a suspected object, and the behavior characteristics of the suspected object are compared with the characteristics of the preset behavior, and a first similarity is obtained. When the first similarity is less than the set threshold, M surveillance images are extracted from the surveillance video on average as the key image frames, and the value of M is increased or decreased by increasing or decreasing the first similarity. When the first similarity is greater than or equal to the set threshold, the surveillance video is saved, and an early warning message is sent to other camera units and a receiving terminal, wherein the early warning message includes the identification information of the suspected object, the current image of the suspected object and an early warning mark, and the surveillance video including the target object captured by the camera unit and the other camera units is taken as the early warning video, and the suspected object is rendered in the early warning video, wherein N is less than M.

4. The method according to claim 1, characterized in that When the recognition model cannot obtain the recognition information of the target object, the feature information of the target object is extracted through the surveillance video including the target object, and the feature information includes facial features, body features and behavioral features. The feature information of the target object is also added to the recognition model, and the recognition information is set for the target object, and the recognition information and the feature information of the target object are sent to other camera units.

5. The method according to claim 1, characterized in that , the step S3 comprises: Step S31: marking the identification information of each target object and the position information of the corresponding camera unit in each key image frame and the warning video, and marking the acquisition time and the corresponding time of each video frame in the key image frame and the warning video respectively; Step S32: constructing a three-dimensional map of the smart building based on the design structure of the smart building, and mapping the position of each camera unit into the three-dimensional map; Step S33: updating the key image frame and the warning video corresponding to each camera unit to the three-dimensional map in real time, and storing them at the storage node position corresponding to each camera unit.

6. The method according to claim 1, characterized in that , the step S4 comprises: Step S41: mapping the key image frames acquired by each camera unit in the three-dimensional map on the time coordinate axis according to the acquisition time, and also mapping the warning video acquired by each camera unit on the time coordinate axis; Step S42: connecting the key image frames corresponding to the same target object or the position information of the camera unit corresponding to the warning video in the three-dimensional map in chronological order, obtaining the monitored moving path of the target object, and marking the target object with a dangerous degree in the three-dimensional map, wherein the higher the dangerous degree, the more eye-catching the color of the monitored moving path; Step S43: The monitoring movement path of each target object is supplemented and completed for the blind spot of each camera unit according to the design structure of the smart building, so as to obtain the complete monitoring path map including the monitoring movement paths of all the target objects.

7. The method according to claim 1, characterized in that , the step S4 further includes a step S5: When the communication load is less than or equal to a second threshold, the three-dimensional monitoring map composed of the complete monitoring path map of all the target objects is sent to the data collection end in real time. When the communication load is greater than the second threshold, the sending priority of the corresponding complete monitoring path map is set according to the danger level of each target object, wherein the higher the danger level, the higher the sending priority.

8. The method according to claim 1, characterized in that When the current behavior probability value of the target object is greater than or equal to a first threshold, the corresponding danger level of the target object is level three danger; when the current behavior probability value of the target object is less than the first threshold, when the first similarity is less than a set threshold, the danger level of the target object is level two danger; when the first similarity is greater than or equal to the set threshold, the danger level of the target object is level one danger.

9. A data acquisition system for smart buildings, the system being used to implement the method according to any one of claims 1 to 8, characterized in that: The system comprises: Multiple camera units are used to periodically collect surveillance videos of corresponding areas; An identification unit, used to identify multiple target objects in each of the surveillance videos through a recognition model, and obtain identification information of each of the target objects; A collection unit, used for obtaining a current behavior probability value of each target object according to the identification information of each target object, the monitoring video and the behavior probability distribution of each target object, and extracting a key image frame and a warning video of the monitoring video based on the current behavior probability value of each target object and a first similarity between the behavior feature of the target object and a preset behavior feature; A mapping unit, used to mark the identification information of the target object, the acquisition time and the position information of the camera unit in the key image frame and the warning video, and a three-dimensional map of the smart building, and map the position of the camera unit to the three-dimensional map, mark the key image frame and the warning video in the three-dimensional map, and store them in the storage node corresponding to the camera unit; A generating unit is used to display the key image frames and the warning videos corresponding to each of the camera units in the three-dimensional map in real time on the time coordinate axis according to the acquisition time and the corresponding time period, and obtain the monitoring movement path of each of the target objects, and mark the degree of danger of the target objects in the three-dimensional map, and also supplement the monitoring movement path of each of the target objects in the blind area of ​​each of the camera units, so as to obtain a complete monitoring path map.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images