Intelligent wireless remote security and protection monitoring system and method based on 5G

Abstract

The invention provides a 5G-based intelligent wireless remote security and protection monitoring system and method, relates to the field of intelligent security and protection, and adopts an artificial intelligence-based analysis technology to judge whether someone illegally intrudes or not by performing feature analysis on a video shot by an intelligent pan-tilt camera and an electric signal generated by an intelligent door magnet. Therefore, illegal intrusion behaviors can be judged more accurately in real time, and the security management quality is improved.

Description

Technical Field

[0001] This application relates to the field of intelligent security, and more specifically, to a 5G-based intelligent wireless remote security monitoring system and method. Background Art

[0002] The growing demand for home security has made wireless remote security monitoring systems an important part of the smart home field. This system can provide real-time monitoring and security protection, helping family members better manage household property and protect the safety of family members. Due to the bandwidth limitations of traditional wireless technologies, it may not be able to support the real-time transmission of high-definition video streams, resulting in a decline in the quality of monitoring images or a large delay. With the rapid development of 5G technology and network technology, intelligent security systems have the characteristics of efficient operation in a low-latency and high-bandwidth environment. Under the establishment of a systematic and systematic security management model, the overall security management level and control efficiency can be improved in all aspects.

[0003] Therefore, an optimized 5G-based intelligent wireless remote security monitoring solution is needed. Summary of the Invention

[0004] To solve the above technical problems, this application is proposed. Embodiments of this application provide a 5G-based intelligent wireless remote security monitoring system and method, which adopt an artificial intelligence-based analysis technology to determine whether someone breaks in illegally by analyzing the characteristics of the video captured by an intelligent pan-tilt camera and the electrical signals generated by an intelligent door magnetic sensor. In this way, illegal intrusion behaviors can be judged more real-time and accurately, improving the quality of security management.

[0005] According to one aspect of this application, a 5G-based intelligent wireless remote security monitoring system is provided, which includes:

[0006] A data acquisition module for acquiring the panoramic living room monitoring video captured by an intelligent pan-tilt camera for a predetermined period of time and the electrical signals generated by an intelligent door magnetic sensor;

[0007] A data feature encoding module for performing feature encoding on the panoramic living room monitoring video for the predetermined period of time and the electrical signals generated by the intelligent door magnetic sensor to obtain a behavior classification feature vector;

[0008] An abnormal behavior judgment module for judging whether someone breaks in illegally based on the behavior classification feature vector.

[0009] In the above 5G-based intelligent wireless remote security monitoring system, the data feature encoding module includes: a video data feature extraction unit for extracting human behavior features from the living room panoramic monitoring video in the predetermined time period to obtain a human behavior tracking feature vector; an electrical signal feature extraction unit for extracting features from the waveform diagram of the electrical signal generated by the intelligent door magnet to obtain an electrical signal waveform feature vector; and a feature fusion unit for performing hidden space projection calibration based on principal component regression on the human behavior tracking feature vector and the electrical signal waveform feature vector to obtain the behavior classification feature vector.

[0010] In the above 5G-based intelligent wireless remote security monitoring system, the video data feature extraction unit includes: a key frame sampling sub-unit for sampling the living room panoramic monitoring video in the predetermined time period to obtain a plurality of living room panoramic monitoring key frames; a human target detection sub-unit for passing the plurality of living room panoramic monitoring key frames through a human target detection network to obtain a plurality of human target monitoring region of interest images; and a human behavior tracking feature encoding sub-unit for passing the plurality of human target monitoring region of interest images through a behavior tracking feature encoder to obtain the human behavior tracking feature vector.

[0011] In the above 5G-based intelligent wireless remote security monitoring system, the electrical signal feature extraction unit is configured to: pass the waveform diagram of the electrical signal generated by the intelligent door magnet through a door magnet electrical signal waveform diagram feature encoder to obtain the electrical signal waveform feature vector.

[0012] In the above 5G-based intelligent wireless remote security monitoring system, the human target detection network is an anchor window-based target detection network, the behavior tracking feature encoder is a first convolutional neural network model using a temporal attention mechanism, and the door magnet electrical signal waveform diagram feature encoder is a second convolutional neural network model using a two-dimensional convolutional kernel.

[0013] In the above 5G-based intelligent wireless remote security monitoring system, the human target detection sub-unit is configured to: process each of the living room panoramic monitoring key frames in the plurality of living room panoramic monitoring key frames according to the following formula to obtain a human target monitoring region of interest image corresponding to the human target in each of the living room panoramic monitoring key frames;

[0014] wherein, the formula is:

[0015] ROI = H(ψ det , B) = (cls(ψ det , B), Regr(ψ det , B))

[0016] wherein, ψ detFor each panoramic monitoring key frame of the living room, B is the anchor box, ROI is the image of the region of interest for human target monitoring in each panoramic monitoring key frame of the living room, and cls(ψ det , B) represents classification, and Regr(ψ det , B) represents regression.

[0017] In the above 5G-based intelligent wireless remote security monitoring system, the feature fusion unit is used to: perform weighted fusion on the human behavior tracking feature vector and the electrical signal waveform feature vector to obtain the behavior electrical signal correlation feature vector, and construct a unitized correlation matrix of the behavior electrical signal correlation feature vector; perform implicit feature analysis on the unitized correlation matrix based on a convolutional layer to obtain a behavior electrical signal correlation hidden space dynamic activation matrix; perform spectral domain decomposition analysis on the unitized correlation matrix to obtain a set of behavior electrical signal correlation principal component coding vectors; input each behavior electrical signal correlation principal component coding vector in the set of behavior electrical signal correlation principal component coding vectors into a feature saliency modulation unit based on a self-attention mechanism to generate a set of behavior electrical signal correlation calibration coding vectors; map each behavior electrical signal correlation calibration coding vector in the set of behavior electrical signal correlation calibration coding vectors to the behavior electrical signal correlation hidden space dynamic activation matrix to obtain a set of behavior electrical signal correlation space mask coding vectors; perform cascade fusion on the set of behavior electrical signal correlation space mask coding vectors to obtain the behavior classification feature vector.

[0018] In the above 5G-based intelligent wireless remote security monitoring system, the corrected feature vector generation sub-unit includes: creating a first DAO interface, where the first DAO interface is used to define the methods required for the corrected feature vector; creating a first class that implements the first DAO, where the first class is used to implement the specific logic of the corrected feature vector; configuring the database in the Spring configuration file; configuring the first class as a Spring Bean so that the first class can be managed by the Spring container; injecting the first DAO interface and calling the methods required for the corrected feature vector in the first DAO interface; passing the first interference coefficient vector, the second interference coefficient vector, the human behavior tracking feature vector, and the electrical signal waveform feature vector into the first DAO interface to perform feature vector correction operations to obtain the corrected human behavior tracking feature vector and the corrected electrical signal waveform feature vector.

[0019] In the above 5G-based intelligent wireless remote security monitoring system, the abnormal behavior judgment module is used to: pass the behavior classification feature vector through a behavior classifier to obtain a classification result, and the classification result is used to indicate whether someone has illegally entered.

[0020] According to another aspect of the present application, there is provided a 5G-based intelligent wireless remote security monitoring method, which includes:

[0021] Obtain the panoramic monitoring video of the living room for a predetermined period taken by the intelligent pan-tilt camera and the electrical signals generated by the intelligent door magnet;

[0022] Perform feature encoding on the panoramic monitoring video of the living room for the predetermined period and the electrical signals generated by the intelligent door magnet to obtain a behavior classification feature vector;

[0023] Based on the behavior classification feature vector, determine whether someone has broken in illegally.

[0024] Compared with the prior art, the 5G-based intelligent wireless remote security monitoring system and method provided by the present application adopt an artificial intelligence-based analysis technology to determine whether someone has broken in illegally by analyzing the features of the video taken by the intelligent pan-tilt camera and the electrical signals generated by the intelligent door magnet. In this way, illegal intrusion behavior can be judged more real-time and accurately, improving the quality of security management. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, features and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0026] Figure 1 It is a system block diagram of a 5G-based intelligent wireless remote security monitoring system according to an embodiment of the present application.

[0027] Figure 2 It is a block diagram of a data feature encoding module in a 5G-based intelligent wireless remote security monitoring system according to an embodiment of the present application.

[0028] Figure 3 It is a block diagram of a video data feature extraction unit in a 5G-based intelligent wireless remote security monitoring system according to an embodiment of the present application.

[0029] Figure 4 It is a flowchart of a 5G-based intelligent wireless remote security monitoring method according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0031] As mentioned in the above background art, the growing demand for home security has made the wireless remote security monitoring system an important part of the smart home field. This system can provide real-time monitoring and security protection, helping family members better manage their household property and protect the safety of family members. Due to the bandwidth limitation of traditional wireless technologies, it may not be able to support the real-time transmission of high-definition video streams, resulting in a decline in the quality of the monitoring images or a significant delay. With the rapid development of 5G technology and network technology, the intelligent security system has the characteristics of efficient operation in an environment of low latency and high bandwidth. Under the establishment of a systematic and comprehensive security management mode, the overall security management level and control efficiency can be improved in all aspects. Therefore, an intelligent wireless remote security monitoring solution based on 5G is expected.

[0032] In recent years, deep learning and neural networks have been widely applied in fields such as computer vision, natural language processing, and text signal processing. In addition, deep learning and neural networks have also demonstrated levels close to or even exceeding those of humans in fields such as image classification, object detection, semantic segmentation, and text translation. The development of deep learning and neural networks provides new solutions and ideas for intelligent wireless remote security monitoring.

[0033] Figure 1 The system block diagram of the intelligent wireless remote security monitoring system based on 5G according to an embodiment of the present application. As Figure 1 shown, in the intelligent wireless remote security monitoring system 100 based on 5G, it includes: a data acquisition module 110, configured to obtain the panoramic monitoring video of the living room for a predetermined period of time captured by the intelligent pan-tilt camera and the electrical signals generated by the intelligent door magnet; a data feature encoding module 120, configured to perform feature encoding on the panoramic monitoring video of the living room for the predetermined period of time and the electrical signals generated by the intelligent door magnet to obtain a behavior classification feature vector; and an abnormal behavior determination module 130, configured to determine whether someone has broken in illegally based on the behavior classification feature vector.

[0034] Specifically, in the technical solution of the present application, first, the panoramic monitoring video of the living room for a predetermined period of time captured by the intelligent pan-tilt camera and the electrical signals generated by the intelligent door magnet are obtained. It should be well understood that the panoramic monitoring video of the living room for a predetermined period of time can provide real-time visual information of the living room, helping users remotely understand the situation at home. By analyzing the video data, human target detection can be performed, and by tracking the behaviors of the detected human targets, it can be determined whether someone has broken in illegally. The intelligent door magnet can be applied to positions such as doors and windows, and the electrical signals generated by the intelligent door magnet can reflect the opening and closing states of the positions where the intelligent door magnet is installed, providing information about whether the positions are opened or closed. By combining the two data sources of the door magnet signal and the video data, it can be determined whether someone has entered the house without permission, thereby realizing the detection of illegal intrusion.

[0035] Mobile communication follows the development law of one generation of technology every ten years and has gone through the development of 1G, 2G, 3G, and 4G. Each generational leap and each technological advancement have greatly promoted industrial upgrading and economic and social development. From 1G to 2G, the transition from analog communication to digital communication was achieved, and mobile communication entered thousands of households; from 2G to 3G and 4G, the transformation from voice services to data services was realized, and the transmission rate increased by hundreds of times, promoting the popularization and prosperity of mobile Internet applications. Currently, mobile networks have been integrated into all aspects of social life, profoundly changing people's communication, interaction, and even the entire lifestyle. The 4G network has created a prosperous Internet economy and solved the problem of people communicating with each other anytime and anywhere. With the rapid development of the mobile Internet, new services and new businesses have emerged continuously, and the mobile data service traffic has increased explosively. The 4G mobile communication system is difficult to meet the demand for the explosive growth of future mobile data traffic, and 5G has emerged as the times require. 5G is the abbreviation of the 5th Generation Mobile Communication Technology. It has the characteristics of high speed, low latency, and large connection. 5G communication facilities are the network infrastructure for realizing the interconnection of humans, machines, and things. It should be well understood that the monitoring system has relatively high requirements for the real-time nature of data. The low-latency characteristic of the 5G network can ensure the real-time transmission of monitoring data, thus achieving rapid response and accurate monitoring. On the other hand, the data volume of the living room panoramic monitoring video and the door magnetic electrical signal is large, requiring high-speed network transmission and processing. The 5G network can provide sufficient bandwidth and speed to process these large data volumes.

[0036] In an embodiment of the present application, an implementable way to obtain the living room panoramic monitoring video for a predetermined period captured by the intelligent pan-tilt camera and the electrical signal generated by the intelligent door magnetic can be as follows: 1. An intelligent pan-tilt camera (for example, supporting 340 degrees horizontally, 80 degrees vertically upward, and 10 degrees vertically downward rotation) can be selected to achieve panoramic view-free monitoring of the living room. The time point for starting the video recording work can be set artificially. For example, real-time video recording can be performed when the user is not at home, and the collected data is stored and uploaded to the system in real time; 2. The intelligent door magnetic consists of two parts: a wireless transmitter and a permanent magnet. After installing the intelligent door magnetic on the window or door and setting the time or scene (such as: 11 pm - 6 am, leaving home scene mode, etc.), the security function of the intelligent door magnetic is enabled. When the window or door is illegally opened or moved, the data is uploaded to the system in real time for analysis.

[0037] Figure 2 It is a block diagram of the data feature encoding module in the 5G-based intelligent wireless remote security monitoring system according to the embodiment of the present application. As Figure 2As shown, the data feature encoding module 120 includes: a video data feature extraction unit 121 for extracting human behavior feature vectors from the panoramic living room surveillance video in the predetermined time period; an electrical signal feature extraction unit 122 for extracting feature vectors of the waveform diagram of the electrical signal generated by the intelligent door magnet; and a feature fusion unit 123 for performing latent space projection calibration based on principal component regression on the human behavior tracking feature vector and the electrical signal waveform feature vector to obtain the behavior classification feature vector.

[0038] Figure 3 It is a block diagram of the video data feature extraction unit in the 5G-based intelligent wireless remote security monitoring system according to an embodiment of the present application. As Figure 3 shown, the video data feature extraction unit 121 includes: a key frame sampling sub-unit 1211 for sampling the panoramic living room surveillance video in the predetermined time period to obtain a plurality of panoramic living room surveillance key frames; a human target detection sub-unit 1212 for passing the plurality of panoramic living room surveillance key frames through a human target detection network to obtain a plurality of human target surveillance region of interest images; and a human behavior tracking feature encoding sub-unit 1213 for passing the plurality of human target surveillance region of interest images through a behavior tracking feature encoder to obtain the human behavior tracking feature vector.

[0039] The panoramic living room surveillance video in the predetermined time period works continuously, which means it contains a large amount of data, but not all of the data is very effective. That is, the panoramic living room surveillance video is composed of frame-by-frame video frames. The surveillance video data for a period of time contains a large number of video frames, and the difference between adjacent video frames is very small. If each video frame is processed, it will consume a large amount of computing resources and time. Therefore, in the technical solution of the present application, the panoramic living room surveillance video in the predetermined time period is sampled to obtain a plurality of panoramic living room surveillance key frames. The panoramic living room key frames usually contain important information. Sampling the video to obtain multiple key frames can not only retain the important content but also speed up the processing speed and improve the real-time performance.

[0040] In an embodiment of the present application, an implementable way to sample the panoramic living room surveillance video in the predetermined time period to obtain a plurality of panoramic living room surveillance key frames can be: sampling the panoramic living room surveillance video in the predetermined time period at a predetermined sampling frequency to obtain the plurality of panoramic living room surveillance key frames. It is worth mentioning that the sampling frequency here is not always fixed, but can be specifically determined according to the specific usage scenario.

[0041] Next, the multiple living room panoramic monitoring key frames are passed through a human target detection network to obtain multiple human target monitoring region of interest images. The target detection network is mainly divided into two types, AnchorBox and anchor-free. In the technical solution of this application, the human target detection network is an anchor-based target detection network. The purpose of the human target detection network is to find human targets in the living room panoramic monitoring key frames. By identifying the living room panoramic monitoring key frames, the system can quickly focus on the important areas where human targets are located, facilitating subsequent analysis of human target behaviors.

[0042] In an embodiment of this application, an implementable way to pass the multiple living room panoramic monitoring key frames through a human target detection network to obtain multiple human target monitoring region of interest images can be: processing each living room panoramic monitoring key frame among the multiple living room panoramic monitoring key frames according to the following formula to obtain a human target monitoring region of interest image corresponding to the human target in each living room panoramic monitoring key frame; where the formula is:

[0043] ROI = H(ψ det , B) = (cls(ψ det , B), Regr(ψ det , B))

[0044] where ψ det is each living room panoramic monitoring key frame, B is an anchor box, ROI is the human target monitoring region of interest image of the human target in each living room panoramic monitoring key frame, cls(ψ det , B) represents classification, and Regr(ψ det , B) represents regression.

[0045] The multiple human target monitoring region of interest images represent images containing human targets at multiple different time points. To help the system determine whether it is an illegal intrusion, illegal intruders generally have some suspicious behaviors, such as staying in a certain area for too long, covering the face, etc. It is necessary to extract the behavioral characteristics of the human target from these human target images, that is, passing the multiple human target monitoring region of interest images through a behavioral tracking feature encoder to obtain a human behavior tracking feature vector. In the technical solution of this application, the essence of the behavioral tracking feature encoder is a convolutional neural network model using a temporal attention mechanism. The temporal attention mechanism can focus on the importance of different time points in the multiple human target monitoring region of interest images, helping to capture the changes in human behaviors over time and improving the accuracy of behavior recognition.

[0046] In an embodiment of the present application, an implementable manner of obtaining a human body behavior tracking feature vector by passing the multiple human body target monitoring region-of-interest images through a behavior tracking feature encoder may be as follows: extracting adjacent first and second human body target monitoring region-of-interest images from the multiple human body target monitoring region-of-interest images; passing the first and second human body target monitoring region-of-interest images through the first and second convolutional layers of the behavior tracking feature encoder respectively to obtain a first human body target behavior convolutional feature map and a second human body target behavior convolutional feature map; calculating the element-wise multiplication between the first human body target behavior convolutional feature map and the second human body target behavior convolutional feature map to obtain a human body target behavior temporal attention map; inputting the human body target behavior temporal attention map into a Softmax activation function to obtain a human body target behavior temporal attention feature map; passing the second human body target monitoring region-of-interest image through the third convolutional layer of the behavior tracking feature encoder to obtain a third human body target behavior convolutional feature map; calculating the element-wise multiplication between the third human body target behavior convolutional feature map and the human body target behavior temporal attention feature map to obtain a human body target behavior temporal attention feature map corresponding to the second human body target monitoring region-of-interest image.

[0047] Meanwhile, passing the waveform diagram of the electrical signal generated by the intelligent door magnet through a door magnet electrical signal waveform feature encoder to obtain an electrical signal waveform feature vector. In the technical solution of the present application, the essence of the door magnet electrical signal waveform feature encoder is a convolutional neural network model using a two-dimensional convolutional kernel. By extracting features from the waveform diagram of the electrical signal, complex electrical signal information can be converted into a more representative feature representation. Extracting the waveform features of the electrical signal helps to accurately identify signal patterns in different situations, including signal changes caused by the presence of personnel intrusion.

[0048] In an embodiment of the present application, an implementable manner of passing the waveform diagram of the electrical signal generated by the intelligent door magnet through a door magnet electrical signal waveform feature encoder to obtain an electrical signal waveform feature vector may be as follows: each layer of the door magnet electrical signal waveform feature encoder performs the following operations on the input data during the forward pass of the layer: performing convolutional processing on the input data based on the two-dimensional convolutional kernel to obtain a convolutional feature map; performing mean pooling processing on each local feature matrix along the channel dimension of the convolutional feature map to obtain a pooled feature map; performing non-linear activation on the pooled feature map to obtain an activated feature map; wherein, the output of the last layer of the door magnet electrical signal waveform feature encoder is the electrical signal waveform feature vector, and the input of the first layer of the door magnet electrical signal waveform feature encoder is the waveform diagram of the electrical signal generated by the intelligent door magnet.

[0049] The human body behavior tracking feature vector is obtained by extracting features from the video data collected by the camera, and the electrical signal waveform feature vector is obtained by extracting features from the electrical signals generated by the intelligent door magnet. These two data source information provide information in different aspects. In order to comprehensively utilize the information of these two data sources, in the technical solution of this application, the human body behavior tracking feature vector and the electrical signal waveform feature vector are fused to obtain a behavior classification feature vector. Through fusion, a more comprehensive behavior feature description can be provided, and the complementary information can improve the accuracy and robustness of behavior classification.

[0050] Specifically, the feature fusion unit 123 is used for: First, the human body behavior tracking feature vector and the electrical signal waveform feature vector are weighted and fused to obtain the behavior electrical signal correlation feature vector, and a unitized correlation matrix of the behavior electrical signal correlation feature vector is constructed, which is expressed by the formula:

[0051]

[0052] where V represents the behavior electrical signal correlation feature vector, v i and v j respectively represent the feature values at the i-th and j-th positions of the behavior electrical signal correlation feature vector, d(v i , v j ) represents calculating the Euclidean distance, and D i,j represents the feature value at the (i, j) position of the unitized correlation matrix.

[0053] That is, by weighted fusing the human body behavior tracking feature vector and the electrical signal waveform feature vector, and constructing a unitized correlation matrix based on the fused behavior electrical signal correlation feature vector, a cross-modal interaction relationship of the behavior electrical signal correlation feature vector at the unitized granularity is established, and the discrete local features are mapped into an explicit structured correlation network, so as to more deeply explore the potential coupling law between behavior and electrical signal, thereby enhancing the robustness judgment of the system in noise interference and occlusion scenarios, and at the same time reducing the false alarm rate caused by single-modal misjudgment.

[0054] Secondly, based on the convolutional layer, the unitized correlation matrix is subjected to implicit feature analysis to obtain a behavior electrical signal correlation implicit space dynamic activation matrix, which is expressed by the formula:

[0055] M = Conv(D)

[0056] where D represents the unitized correlation matrix, Conv represents the convolutional layer, and M represents the behavior electrical signal correlation implicit space dynamic activation matrix.

[0057] That is, through the unique local perception and weight sharing mechanism of the convolutional layer, the latent non-linear topological structure in the unitized correlation matrix is dynamically mined to identify the temporal coupling pattern between specific location features or the hierarchical dependence relationship between spatial distribution features. Specifically, the discrete unitized correlation matrix is reconstructed into a fluid latent space representation, that is, the dynamic activation matrix of the behavior electrical signal correlation latent space is generated, which can adaptively generate a multi-dimensional context-aware field. For example, in an abnormal intrusion scenario, the system can not only identify the spatio-temporal matching relationship between the door magnetic trigger level and the video moving target, but also perceive the interference intensity of the environmental background noise on the correlation pattern through the gradient propagation mechanism in the latent space, thereby reducing the false alarm rate while enhancing the detection sensitivity of concealed intrusion behavior.

[0058] Then, perform spectral domain decomposition analysis on the unitized correlation matrix to obtain a set of behavior electrical signal correlation principal component coding vectors, which is expressed by the formula:

[0059]

[0060] where T represents the transpose of the vector, Λ represents the diagonal matrix, λ1 and λ m represent the first and the m-th eigenvalues on the diagonal of the diagonal matrix respectively, U represents the set of behavior electrical signal correlation principal component coding vectors, and x1, x2, x m represent the first, the second and the m-th behavior electrical signal correlation principal component coding vectors respectively.

[0061] That is, the unitized correlation matrix with mixed noise is decoupled into interpretable physical primitives, and the principal component coding vectors with stable physical meanings are extracted. Through this decomposition method, the system can focus on the key dimensions strongly related to security in the cross-modal data stream, rather than passively accepting all correlation noises. The generated set of behavior electrical signal correlation principal component coding vectors reconstructs the mixed time-frequency domain features in the unitized correlation matrix into independent event primitives with physical separability, avoiding misjudgment caused by the surface similarity of behavior patterns.

[0062] Next, input each behavior electrical signal correlation principal component coding vector in the set of behavior electrical signal correlation principal component coding vectors into the feature saliency modulation unit based on the self-attention mechanism to generate a set of behavior electrical signal correlation calibration coding vectors, which is expressed by the formula:

[0063] Y = Transformer{[x1, x2, …, x m} = [y1, y2, …, y m

[0064] ​Among them, Transformer represents a feature significant modulation unit based on the self-attention mechanism, Y represents a set of behavior electro-signal correlation calibration coding vectors, and y1, y2, y m represent the first, second, and m-th behavior electro-signal correlation calibration coding vectors respectively.

[0065] That is, by constructing a dynamic attention network among the behavior electro-signal correlation principal component coding vectors, re-evaluating the causal weights of each behavior electro-signal correlation principal component coding vector in the current overall behavior-electro-signal correlation topology, the feature expression is always anchored to the key nodes of the risk determination logic chain. The generated set of behavior electro-signal correlation calibration coding vectors can dynamically bind the physical principal components and behavior semantics, enabling the system to not only parse the static correlation between behavior and signal, but also establish a risk probability model under spatio-temporal constraints through context awareness, thereby achieving a dual optimization balance of false alarm rate and miss rate in a complex home environment.

[0066] Then, map each behavior electro-signal correlation calibration coding vector in the set of behavior electro-signal correlation calibration coding vectors to the behavior electro-signal correlation hidden space dynamic activation matrix to obtain a set of behavior electro-signal correlation space mask coding vectors, which is expressed by the formula:

[0067]

[0068] Among them, represents matrix multiplication, S represents the feature scale of the behavior electro-signal correlation hidden space dynamic activation matrix, y i represents the i-th behavior electro-signal correlation calibration coding vector, L represents the length of the behavior electro-signal correlation calibration coding vector, and z i represents the i-th behavior electro-signal correlation space mask coding vector.

[0069] That is, by constructing a dynamically reconfigurable cross-modal perception field, the feature expression can adaptively generate a spatial mask to focus on key risk events according to the real-time topology change of the hidden space activation matrix. The generated set of behavior electro-signal correlation space mask coding vectors can, through the hidden space gradient propagation mechanism, refine the cross-modal association pattern from coarse-grained event matching to a multi-scale risk probability distribution map, enabling the system to penetrate environmental interference and accurately capture the essential causal chain between cross-modal events, rather than staying at the level of correlation matching of surface features.

[0070] Finally, cascade and fuse the set of behavior electro-signal correlation space mask coding vectors to obtain the behavior classification feature vector, which is expressed by the formula:

[0071] V' = Concat{z1, z2, …, z m}

[0072] Among them, Concat represents the concatenation function, and z1, z2, and z m respectively represent the first, second, and m-th behavior electrical signal associated space mask coding vectors, and V′ represents the behavior classification feature vector.

[0073] That is, a high-dimensional feature space with physical interpretability is constructed, so that the generated behavior classification feature vector more completely retains the original action path of the multi-scale association pattern, enabling the model to perform dialectical analysis of "weak signal strong association" and "strong signal weak association" in complex environmental noise, avoiding the limitations of single-dimensional decision-making, and enabling the system to have the ability to extract essential differences from a large number of interference events.

[0074] Finally, the behavior classification feature vector is passed through a behavior classifier to obtain a classification result, and the classification result is used to indicate whether someone has illegally entered. The behavior classifier is a machine learning model, and its essence is to analyze and classify the input feature data to identify different behavior patterns. By quickly judging whether someone has illegally entered through the behavior classifier, the system can quickly respond to potential security risks, such as sending an alarm prompt message to the user, and the user can choose to call the police to strengthen security prevention and protect their own safety.

[0075] In an embodiment of the present application, an implementable manner of passing the behavior classification feature vector through a behavior classifier to obtain a classification result, where the classification result is used to indicate whether someone has illegally entered can be: using the behavior classifier to process the behavior classification feature vector with the following classification formula to generate the classification result; where the classification formula is:

[0076] O = softmax{(M c , B c )|V′}

[0077] where O is the classification result, V′ represents the behavior classification feature vector, M c is the weight matrix of the fully connected layer, B c represents the bias vector of the fully connected layer, and softmax is the normalized exponential function.

[0078] In summary, the 5G-based intelligent wireless remote security monitoring system 100 according to the embodiments of the present application is elucidated. It adopts artificial intelligence-based analysis technology to judge whether someone has illegally entered by analyzing the features of the video captured by the intelligent pan-tilt camera and the electrical signals generated by the intelligent door magnet. In this way, illegal entry behavior can be judged more real-time and accurately, improving the quality of security management.

[0079] As described above, the 5G-based intelligent wireless remote security monitoring system 100 according to the embodiments of the present application can be implemented in various terminal devices, such as servers for 5G-based intelligent wireless remote security monitoring. In one example, the 5G-based intelligent wireless remote security monitoring system 100 according to the embodiments of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the 5G-based intelligent wireless remote security monitoring system 100 can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the 5G-based intelligent wireless remote security monitoring system 100 can also be one of the many hardware modules of the terminal device.

[0080] Alternatively, in another example, the 5G-based intelligent wireless remote security monitoring system 100 and the terminal device can also be separate devices, and the 5G-based intelligent wireless remote security monitoring system 100 can be connected to the terminal device through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format.

[0081] Figure 4 FIG. is a flowchart of a 5G-based intelligent wireless remote security monitoring method according to an embodiment of the present application. As Figure 4 shown, in the 5G-based intelligent wireless remote security monitoring method, it includes: S110, obtaining a panoramic monitoring video of the living room for a predetermined period of time captured by an intelligent pan-tilt camera and an electrical signal generated by an intelligent door magnet; S120, performing feature encoding on the panoramic monitoring video of the living room for the predetermined period of time and the electrical signal generated by the intelligent door magnet to obtain a behavior classification feature vector; S130, based on the behavior classification feature vector, determining whether someone breaks in illegally.

[0082] Here, those skilled in the art can understand that the specific operations of each step in the above 5G-based intelligent wireless remote security monitoring method have been described in detail in the description of the 5G-based intelligent wireless remote security monitoring system above, and thus, the repeated description thereof will be omitted. Figures 1 to 3 of the 5G-based intelligent wireless remote security monitoring system, and thus, the repeated description thereof will be omitted.

[0083] In summary, the 5G-based intelligent wireless remote security monitoring method according to the embodiments of the present application is clarified. It adopts an artificial intelligence-based analysis technology to determine whether someone breaks in illegally by analyzing the features of the video captured by the intelligent pan-tilt camera and the electrical signal generated by the intelligent door magnet. In this way, it is possible to more real-time and accurately determine illegal intrusion behavior and improve the quality of security management.

[0084] In several embodiments provided by the present application, it should be understood that the disclosed method, system or device can be implemented in other ways. For example, the system embodiment described above is only illustrative. For example, the division of the modules is only a logical function division, and there can be other division methods in actual implementation.

[0085] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0086] In addition, in each embodiment of the present application, the functional modules can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware, or in the form of hardware plus software functional modules.

[0087] For those skilled in the art, it is obvious that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present application.

[0088] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present application is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any appended drawing reference signs in the claims should not be regarded as limiting the claimed rights.

[0089] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units or devices stated in the system claims can also be implemented by one unit or device through software or hardware. Words such as "second" are used to denote names and do not denote any particular order.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit them. Although the present application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit of the technical solutions of the present application.

Claims

1. An intelligent wireless remote security monitoring system based on 5G, characterized in that, Including: A data acquisition module, configured to obtain a panoramic living room surveillance video for a predetermined period captured by an intelligent PTZ camera and an electrical signal generated by an intelligent door magnetic sensor; A data feature encoding module, configured to perform feature encoding on the panoramic living room surveillance video for the predetermined period and the electrical signal generated by the intelligent door magnetic sensor to obtain a behavior classification feature vector; An abnormal behavior judgment module, configured to judge whether someone breaks in illegally based on the behavior classification feature vector.

2. The intelligent wireless remote security monitoring system based on 5G according to claim 1, characterized in that The data feature encoding module includes: A video data feature extraction unit, configured to extract human behavior features from the panoramic living room surveillance video for the predetermined period to obtain a human behavior tracking feature vector; An electrical signal feature extraction unit, configured to extract features from the waveform diagram of the electrical signal generated by the intelligent door magnetic sensor to obtain an electrical signal waveform feature vector; A feature fusion unit, configured to perform hidden space projection calibration based on principal component regression on the human behavior tracking feature vector and the electrical signal waveform feature vector to obtain the behavior classification feature vector.

3. The intelligent wireless remote security monitoring system based on 5G according to claim 2, characterized in that, The video data feature extraction unit includes: A key frame sampling sub-unit, configured to sample the panoramic living room surveillance video for the predetermined period to obtain a plurality of panoramic living room surveillance key frames; A human target detection sub-unit, configured to pass the plurality of panoramic living room surveillance key frames through a human target detection network to obtain a plurality of human target surveillance region of interest images; A human behavior tracking feature encoding sub-unit, configured to pass the plurality of human target surveillance region of interest images through a behavior tracking feature encoder to obtain the human behavior tracking feature vector.

4. The 5G-based intelligent wireless remote security monitoring system according to claim 3, characterized in that, The electrical signal feature extraction unit is configured to: pass the waveform diagram of the electrical signal generated by the intelligent door magnetic sensor through a door magnetic electrical signal waveform diagram feature encoder to obtain the electrical signal waveform feature vector.

5. The 5G-based intelligent wireless remote security monitoring system according to claim 4, characterized in that, The human target detection network is an anchor window-based target detection network, the behavior tracking feature encoder is a first convolutional neural network model using a temporal attention mechanism, and the door magnetic electrical signal waveform diagram feature encoder is a second convolutional neural network model using a two-dimensional convolutional kernel.

6. The 5G-based intelligent wireless remote security monitoring system according to claim 5, characterized in that The human target detection sub-unit is configured to: process each panoramic living room surveillance key frame in the plurality of panoramic living room surveillance key frames according to the following formula to obtain a human target surveillance region of interest image corresponding to the human target in each panoramic living room surveillance key frame; wherein, the formula is: ROI = H(ψ det , B) = (cls(ψ det , B), Regr(ψ det , B)) Among them, ψ det is the key frame of the panoramic monitoring of each living room, B is the anchor box, ROI is the image of the region of interest for human target monitoring in the key frame of the panoramic monitoring of each living room, cls(ψ det , B) represents classification, and Regr(ψ det , B) represents regression.

7. The 5G-based intelligent wireless remote security monitoring system according to claim 6, wherein, The feature fusion unit is configured to: perform weighted fusion on the human behavior tracking feature vector and the electrical signal waveform feature vector to obtain a behavior electrical signal correlation feature vector, and construct a unitized correlation matrix of the behavior electrical signal correlation feature vector; perform hidden feature analysis on the unitized correlation matrix based on a convolutional layer to obtain a behavior electrical signal correlation hidden space dynamic activation matrix; perform spectral domain decomposition analysis on the unitized correlation matrix to obtain a set of behavior electrical signal correlation principal component coding vectors; input each behavior electrical signal correlation principal component coding vector in the set of behavior electrical signal correlation principal component coding vectors into a feature significance modulation unit based on a self-attention mechanism to generate a set of behavior electrical signal correlation calibration coding vectors; Map each behavior electrical signal associated calibration coding vector in the set of behavior electrical signal associated calibration coding vectors to the behavior electrical signal associated hidden space dynamic activation matrix to obtain a set of behavior electrical signal associated space mask coding vectors; Perform cascading fusion on the set of behavior electrical signal associated space mask coding vectors to obtain the behavior classification feature vector.

8. The intelligent wireless remote security monitoring system based on 5G according to claim 7, characterized in that, The abnormal behavior judgment module is used to: pass the behavior classification feature vector through a behavior classifier to obtain a classification result, and the classification result is used to indicate whether someone breaks in illegally.

9. An intelligent wireless remote security monitoring method based on 5G, characterized in that, It includes: Obtain the living room panoramic surveillance video for a predetermined period of time captured by the intelligent pan-tilt camera and the electrical signals generated by the intelligent door magnetic; Perform feature encoding on the living room panoramic surveillance video for the predetermined period of time and the electrical signals generated by the intelligent door magnetic to obtain a behavior classification feature vector; Based on the behavior classification feature vector, determine whether someone breaks in illegally.

Images