Internet of Things Security Management System and Method

Abstract

The present application discloses an Internet of Things security management system and method, which uses a camera to collect face detection images of objects to be identified, and introduces image processing and analysis algorithms at the back end to perform feature comparison and analysis on the face detection images and face reference images with predetermined permissions, so as to verify the identity of the staff. This can prevent unauthorized personnel from entering dangerous areas and reduce illegal intrusions and thefts. At the same time, the system can also generate early warning prompts based on the face recognition results, detect abnormal situations in time and take corresponding measures. In addition, according to personnel with different predetermined permissions, the system can provide them with different path planning to ensure that they can reach the destination along the specified path. This can prevent staff from running around and passing through dangerous areas, reducing the occurrence of accidents. In addition, the path planning function can also improve work efficiency and reduce time waste.

Description

Technical Field

[0001] The present application relates to the field of Internet of Things, and more specifically, to an Internet of Things security management system and method. Background Art

[0002] With the development of society, information technology has made rapid progress. As an important part of the new generation of information technology, the Internet of Things plays an important role in the information age. In coal mining, since coal mines are a high-risk working environment, there are various safety hazards and risks, such as gas explosions, mine collapses, fires, etc. Therefore, ensuring the safety of coal mine workers has become an urgent and important task.

[0003] However, traditional IoT security management systems usually rely on manual identity verification and path planning. Manual operations may involve risks of subjectivity and negligence, leading to inaccurate identity verification and path planning. At the same time, manual operations also increase workload and time costs.

[0004] In addition, there are some existing identity verification management systems that rely on machines to compare facial images. However, such management systems often have certain delays in identity verification and early warning, and are difficult to cope with complex coal mining scenarios. In other words, there may be a large number of workers and complex working areas in coal mines. The current system is difficult to quickly and accurately identify and distinguish different types of work and permissions, which may lead to misjudgment and incorrect path planning, increasing the risk of workers.

[0005] Therefore, an optimized IoT security management system is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an Internet of Things security management system and method, which collects face detection images of objects to be identified through cameras, and introduces image processing and analysis algorithms at the back end to perform feature comparison and analysis on the face detection images and face reference images with predetermined permissions, so as to verify the identity of the staff. This can prevent unauthorized personnel from entering dangerous areas and reduce illegal intrusions and thefts. At the same time, the system can also generate early warning prompts based on the face recognition results, detect abnormal situations in time and take corresponding measures. In addition, according to personnel with different predetermined permissions, the system can provide them with different path planning to ensure that they can reach the destination along the specified path. This can prevent staff from running around and passing through dangerous areas, reducing the occurrence of accidents. In addition, the path planning function can also improve work efficiency and reduce time waste.

[0007] According to one aspect of the present application, there is provided an Internet of Things security management system, which includes:

[0008] A face detection image acquisition module is used to obtain a face detection image of an object to be identified acquired by a camera;

[0009] A predetermined authority reference image acquisition module is used to acquire a face reference image with predetermined authority from a database;

[0010] A face image feature interaction analysis module, used for performing image feature interaction correlation analysis on the face detection image and the face reference image to obtain interaction features between face images;

[0011] The warning and path planning module is used to generate warning prompts and / or path planning diagrams based on the interaction features between the facial images.

[0012] According to another aspect of the present application, there is provided an Internet of Things security management method, which includes:

[0013] Obtaining a face detection image of an object to be identified captured by a camera;

[0014] Acquire a face reference image with predetermined authority from a database;

[0015] Performing image feature interactive correlation analysis on the face detection image and the face reference image to obtain interactive features between face images;

[0016] Based on the interaction features between the facial images, an early warning prompt and / or a path planning diagram is generated.

[0017] Compared with the prior art, the present application provides an Internet of Things security management system and method, which collects face detection images of objects to be identified through cameras, and introduces image processing and analysis algorithms at the back end to perform feature comparison and analysis on the face detection images and face reference images with predetermined permissions, so as to verify the identity of the staff. This can prevent unauthorized personnel from entering dangerous areas and reduce illegal intrusions and thefts. At the same time, the system can also generate early warning prompts based on the face recognition results, detect abnormal situations in time and take corresponding measures. In addition, according to personnel with different predetermined permissions, the system can provide them with different path planning to ensure that they can reach the destination along the specified path. This can prevent staff from running around and passing through dangerous areas, reducing the occurrence of accidents. In addition, the path planning function can also improve work efficiency and reduce time waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 is a block diagram of an Internet of Things security management system according to an embodiment of the present application;

[0020] Figure 2 A system architecture diagram of an Internet of Things security management system according to an embodiment of the present application;

[0021] Figure 3 A block diagram of a facial image feature interactive analysis module in an Internet of Things security management system according to an embodiment of the present application;

[0022] Figure 4 A block diagram of an early warning and path planning module in an Internet of Things security management system according to an embodiment of the present application;

[0023] Figure 5 Flow chart of the Internet of Things security management method according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0026] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0027] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0029] Traditional IoT security management systems usually rely on manual identity verification and path planning. Manual operations may be subjective and negligent, resulting in inaccurate identity verification and path planning. At the same time, manual operations will also increase workload and time costs. In addition, there are some identity verification management systems that rely on machines for facial image comparison. However, such management systems often have certain delays in identity verification and early warning, and are difficult to cope with complex coal mining scenarios. In other words, there may be a large number of staff and complex work areas in coal mines. The current system is difficult to quickly and accurately identify and distinguish different types of work and permissions, which may lead to misjudgments and incorrect path planning, increasing the risk of staff. Therefore, an optimized IoT security management system is desired.

[0030] In the technical solution of the present application, an Internet of Things security management system is proposed. Figure 1 is a block diagram of an Internet of Things security management system according to an embodiment of the present application. Figure 2 FIG. 1 is a system architecture diagram of the Internet of Things security management system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the Internet of Things security management system 300 includes: a face detection image acquisition module 310, which is used to obtain a face detection image of an object to be identified collected by a camera; a predetermined authority reference image acquisition module 320, which is used to obtain a face reference image with predetermined authority from a database; a face image feature interaction analysis module 330, which is used to perform image feature interaction correlation analysis on the face detection image and the face reference image to obtain interaction features between face images; an early warning and path planning module 340, which is used to generate early warning prompts and / or path planning maps based on the interaction features between face images.

[0031] In particular, the face detection image acquisition module 310 and the predetermined authority reference image acquisition module 320 are used to acquire the face detection image of the object to be identified acquired by the camera; and to acquire the face reference image with predetermined authority from the database. It should be understood that in the technical solution of the present application, the face detection image of the object to be identified is acquired by the camera, and the image processing and analysis algorithm is introduced in the back end to perform feature comparison and analysis on the face detection image and the face reference image with predetermined authority, so as to realize the verification of the identity of the staff. Therefore, first, the face detection image of the object to be identified is acquired by the camera; and the face reference image with predetermined authority is acquired from the database.

[0032] In particular, the face image feature interaction analysis module 330 is used to perform image feature interaction correlation analysis on the face detection image and the face reference image to obtain interaction features between face images. In particular, in a specific example of the present application, Figure 3 As shown, the face image feature interaction analysis module 330 includes: a face detection image enhancement unit 331, used to perform bilateral filtering on the face detection image to obtain an enhanced face detection image; a face image feature extraction unit 332, used to pass the enhanced face detection image and the face reference image through a dual detection network model to obtain a face detection feature map and a face reference feature map; a global feature interaction analysis unit 333 between face images, used to perform full-perceptual interaction association coding on the face detection feature map and the face reference feature map respectively to obtain the interaction features between the face images.

[0033] Specifically, the face detection image enhancement unit 331 is used to perform bilateral filtering on the face detection image to obtain an enhanced face detection image. Considering that various types of noise may exist in the face detection image during the face detection process, such as Gaussian noise, salt and pepper noise, etc. These noises will interfere with the accuracy of the face detection algorithm, resulting in false detection or missed detection. It should be understood that bilateral filtering is a nonlinear filtering method, which can remove noise while maintaining edge information, and effectively improve image quality. Therefore, in the technical solution of the present application, the face detection image is further subjected to bilateral filtering to obtain an enhanced face detection image. By performing bilateral filtering on the original image, image noise can be removed and image details can be enhanced. And bilateral filtering has the characteristic of edge preservation, which can retain edge information in the image and enhance edge contrast. In face detection, the face usually has obvious edge features. By enhancing edge contrast, the face can be made clearer and more prominent, which helps to improve the performance of the face detection algorithm.

[0034] It is worth noting that bilateral filtering is an image filtering technique that smoothes images while maintaining edge clarity. Unlike traditional linear filtering methods (such as mean filtering and Gaussian filtering), bilateral filtering considers two factors: the spatial distance between pixels and the similarity of pixel values, thereby retaining the detailed information of the image more accurately. In bilateral filtering, the new value of each pixel is obtained by weighted averaging the pixels around it. The weight here is determined by two factors: spatial distance weight: represents the spatial distance between pixels. The closer the distance, the greater the weight. This weight function usually uses a Gaussian function to attenuate the influence of distance to ensure that pixels closer to the target pixel contribute more to the smoothing result; pixel value similarity weight: represents the similarity between pixel values. If the pixel values ​​are similar, the greater the weight. This weight function usually uses a Gaussian function of grayscale value difference to measure the similarity of pixel values ​​to ensure that similar pixels contribute more to the smoothing result. The final weight of each pixel can be obtained by multiplying the spatial distance weight and the pixel value similarity weight. Then the value of each pixel is multiplied by the corresponding weight, and all pixels are summed up. Finally, the result is normalized to obtain the smoothed image.

[0035] Accordingly, in a possible implementation, the face detection image can be subjected to bilateral filtering to obtain an enhanced face detection image by the following steps, for example: obtaining the face detection image to be processed; determining the parameters of bilateral filtering, including the size of the filter, the standard deviation of the spatial distance weight, the standard deviation of the pixel value similarity weight, etc.; and performing bilateral filtering on the face detection image. The specific steps are as follows: traversing each pixel in the image; for the current pixel, determining the range of the filter, that is, determining the size of the surrounding pixels; calculating the spatial distance weight and the pixel value similarity weight between the current pixel and the surrounding pixels; multiplying the spatial distance weight and the pixel value similarity weight to obtain the final weight of the current pixel; multiplying the value of the current pixel by the final weight and summing all pixels; normalizing the result to obtain the enhanced value of the current pixel; repeating steps a to f until all pixels are traversed; outputting the enhanced face detection image: outputting the image composed of pixel values ​​after bilateral filtering as the enhanced face detection image.

[0036] Specifically, the face image feature extraction unit 332 is used to pass the enhanced face detection image and the face reference image through a dual detection network model to obtain a face detection feature map and a face reference feature map. That is, in the technical solution of the present application, a convolutional neural network model with excellent performance in extracting implicit features of an image is used to perform feature mining of the enhanced face detection image and the face reference image respectively. In particular, considering that when performing face image feature comparison and identity recognition, the faces of miners may have a high degree of similarity, which makes it difficult to distinguish and recognize. Therefore, in order to further improve the accuracy of identity verification of the object to be identified, so as to perform more accurate early warning and path planning, in the technical solution of the present application, the enhanced face detection image and the face reference image are further passed through a dual detection network model including a first image encoder and a second image encoder to obtain a face detection feature map and a face reference feature map. In particular, here, the first image encoder and the second image encoder have the same network structure. It should be understood that using the dual detection network model of the image encoder with the same network structure to respectively perform feature extraction of the enhanced face detection image and the face reference image can mine feature information of the two images that are not obviously different at the image source domain end, so as to perform face recognition and identity verification of the object to be identified, thereby improving the accuracy of face recognition and identity verification of the object to be identified. Specifically, using each layer of the first image encoder to perform convolution processing, pooling processing based on the local feature matrix and nonlinear activation processing on the enhanced face detection image in the forward pass of the layer to output an initial face detection feature map by the last layer of the first image encoder; inputting the initial face detection feature map into the spatial attention layer of the first image encoder to obtain the face detection feature map; and, using each layer of the second image encoder to perform convolution processing, pooling processing based on the local feature matrix and nonlinear activation processing on the face reference image in the forward pass of the layer to output an initial face reference feature map by the last layer of the second image encoder; and, inputting the initial face reference feature map into the spatial attention layer of the second image encoder to obtain the face reference feature map.

[0037] Specifically, the global feature interaction analysis unit 333 between face images is used to perform full-perception interaction correlation coding on the face detection feature map and the face reference feature map respectively to obtain the interaction feature between face images. In particular, in a specific example of the present application, the global feature interaction analysis unit 333 between face images includes: a face feature full-perception subunit, used to pass the face detection feature map and the face reference feature map through a full-perception module respectively to obtain a face detection full-perception feature vector and a face reference full-perception feature vector; a face full-perception feature interaction subunit, used to use the inter-feature attention layer to perform feature interaction based on the attention mechanism on the face detection full-perception feature vector and the face reference full-perception feature vector to obtain an interaction feature vector between face images as the interaction feature between face images.

[0038] More specifically, the face feature full perception subunit is used to pass the face detection feature map and the face reference feature map through the full perception module to obtain a face detection full perception feature vector and a face reference full perception feature vector. It should be understood that, considering that when the features of the face detection image and the face reference image are actually compared to verify the identity of the object to be identified, different areas of the face may have different degrees of correlation features, which plays an important role in face recognition. However, due to the inherent limitations of convolution operations, it is difficult for pure CNN methods to learn clear global and long-range semantic information interactions. Therefore, in the technical solution of the present application, the face detection feature map and the face reference feature map are further passed through the full perception module to obtain a face detection full perception feature vector and a face reference full perception feature vector. In particular, here, the full perception module can globally perceive and integrate the face detection feature map and the face reference feature map respectively to extract higher-level semantic information. Therefore, the obtained inter-feature attention layer can better represent the features and attributes of the face to be identified for the full perception feature vector of the face detection, and provide more accurate and useful feature representation for subsequent identity verification and early warning functions.

[0039] Accordingly, in a possible implementation, the face detection feature map and the face reference feature map can be respectively passed through the full perception module to obtain the face detection full perception feature vector and the face reference full perception feature vector through the following steps, for example: obtaining the face detection feature map and the face reference feature map to be processed; determining the parameters of the full perception module; constructing a module for extracting the face detection full perception feature. The specific steps are as follows: performing multi-layer convolution operations on the face detection feature map to extract features; performing global average pooling on the extracted features to obtain the global average value of the feature map; using the global average value as the face detection full perception feature vector; constructing a module for extracting the face reference full perception feature. The specific steps are as follows: performing multi-layer convolution operations on the face reference feature map to extract features; performing global average pooling on the extracted features to obtain the global average value of the feature map; using the global average value as the face reference full perception feature vector; outputting the face detection full perception feature vector and the face reference full perception feature vector: outputting the face detection full perception feature vector and the face reference full perception feature vector as the results for subsequent interactive feature extraction between face images or other related tasks.

[0040] More specifically, the face fully perceptual feature interaction subunit is used to use the inter-feature attention layer to perform feature interaction based on the attention mechanism on the face detection fully perceptual feature vector and the face reference fully perceptual feature vector to obtain the face image interaction feature vector as the face image interaction feature. That is, in the technical solution of the present application, the inter-feature attention layer is used to perform feature interaction based on the attention mechanism on the face detection fully perceptual feature vector and the face reference fully perceptual feature vector to obtain the face image interaction feature vector, so as to capture the association and mutual influence between the fully perceptual features of the face detection image and the fully perceptual features of the face reference image. It should be understood that since the goal of the traditional attention mechanism is to learn an attention weight matrix, larger weights are assigned to important features and smaller weights are assigned to secondary features, thereby selecting information that is more critical to the current task goal. This approach focuses more on weighting the importance of each feature and ignores the dependencies between features. The inter-feature attention layer can capture the correlation and mutual influence between the full-perceptual features of the face detection image and the full-perceptual features of the face reference image through feature interaction based on the attention mechanism, can learn the dependencies between different features, and interact and integrate features according to these dependencies, so as to obtain the interaction feature vector between face images.

[0041] It is worth mentioning that in other specific examples of the present application, the face detection feature map and the face reference feature map can also be fully perceptually interactively associated with encoding in other ways to obtain the interactive features between the face images, for example: obtaining the face detection feature map and the face reference feature map to be processed; determining the parameters of the fully perceptually interactively associated encoding, including the size of the feature map, the number of channels, the number of layers of the encoder, the size of the convolution kernel, etc.; constructing a fully perceptually interactively associated encoder. The specific steps are as follows: perform multi-layer convolution operations on the face detection feature map and the face reference feature map to extract features; splice the two feature maps together to obtain a joint feature map; perform further convolution operations on the joint feature map to extract higher-level features; and use the feature map extracted by the encoder to extract interactive features. The specific steps are as follows: perform global average pooling on the feature map output by the encoder to obtain the global average of the feature map; multiply the global average by the feature map to obtain the interactive features; output the interactive features as the result for subsequent interaction analysis between face images or other related tasks.

[0042] It is worth mentioning that in other specific examples of the present application, the face detection image and the face reference image can also be subjected to image feature interactive correlation analysis in other ways to obtain interactive features between face images, for example: preprocessing the face detection image and the face reference image, including adjusting the image size, grayscale or color conversion, etc., for subsequent processing; using a face detection algorithm to perform face detection on the face detection image to determine the face position and bounding box in the image; using a face feature extraction algorithm (such as a face feature extractor based on deep learning) to extract features from the face area in the face detection image and the face reference image to obtain a feature vector for each face; matching each face feature vector in the face detection image with each face feature vector in the face reference image to calculate the similarity or distance between them. Common matching methods include cosine similarity, Euclidean distance, etc.; according to the result of feature matching, performing interactive correlation analysis on each face feature vector in the face detection image and each face feature vector in the face reference image. Some statistical methods or machine learning models can be used to analyze the relationship between them, such as calculating the mean, variance, correlation coefficient, etc.; based on the results of feature interaction analysis, the interaction features between face images can be extracted. These features can include similarity, correlation, overlap, etc. between faces; based on the obtained interaction features, various applications can be performed, such as face recognition, face comparison, face clustering, etc.

[0043] In particular, the warning and path planning module 340 is used to generate a warning prompt and / or a path planning diagram based on the interaction features between the face images. In particular, in a specific example of the present application, Figure 4As shown, the early warning and path planning module 340 includes: a facial feature distribution optimization unit 341, which is used to perform Hilbert orthogonal space domain representation decoupling on the interaction feature vector between facial images to obtain an optimized interaction feature vector between facial images; an image matching unit 342, which is used to pass the optimized interaction feature vector between facial images through a classifier to obtain a classification result, and the classification result is used to indicate whether the facial detection image of the object to be identified matches the facial reference image of the predetermined authority; a result generation unit 343, which is used to generate an early warning prompt and / or a path planning map based on the classification result.

[0044] Specifically, the facial feature distribution optimization unit 341 is used to perform Hilbert orthogonal space domain representation decoupling on the interaction feature vector between the facial images to obtain an optimized interaction feature vector between the facial images. In the technical solution of the present application, the face detection full-perception feature vector and the face reference full-perception feature vector respectively express the image semantic features of the face detection image and the face reference image, thereby extracting the dependency features between the image semantic features of different source images through feature interaction based on the attention mechanism. Here, it is considered that there is an uneven distribution of dependency between the face detection image and the face reference image in the local distribution of features, that is, there are local areas where the semantics of the source images are the same and the degree of difference between the face detection image and the face reference image deviates greatly, so that the semantic similarity of the source images in some local areas is much greater than the semantic similarity of the source images in other local areas, so that the dependency characteristics of the image semantic features expressed by the interactive feature vector between face images have diversified local feature expressions between the distributions of each local feature. In this way, when the interactive feature vector between face images passes through the classifier, it will affect the generalization effect of the interactive feature vector between face images as a whole in the classification domain, that is, it will affect the accuracy of the classification result. Based on this, when the applicant of the present application classifies the interactive feature vector between face images, it is preferred to decouple the interactive feature vector between face images, for example, denoted as V, by Hilbert orthogonal space domain representation as follows:

[0045]

[0046]

[0047]

[0048] Where V is the interaction feature vector between the face images, is the global feature mean of the interaction feature vector between the face images, ||V||2 is the bi-norm of the interaction feature vector between the face images, L is the length of the interaction feature vector between the face images, and I is a unit vector, Represents vector subtraction, Cov1D (·) represents the covariance matrix, and V′ is the optimized face image interaction feature vector. Here, the Hilbert orthogonal space domain representation decoupling is used to emphasize the essential domain-specific information in the diversified feature expression of the face image interaction feature vector V, that is, by using the Hilbert space metric based on the vector self-space metric and the vector self-inner product representation, to perform domain-invariant characterization from the orthogonal space domain decoupling in the overall domain representation of the face image interaction feature vector V, so as to improve the domain adaptive generalization performance of the face image interaction feature vector V in the classification domain, thereby improving the accuracy of the classification result of the face image interaction feature vector. In this way, the identity verification and early warning of the staff can be realized to detect abnormal situations and take corresponding measures to prevent unauthorized personnel from entering dangerous areas and reduce illegal intrusion and theft. In addition, the security management system can also provide different path planning according to personnel with different predetermined permissions to ensure that staff of different types of work can reach the destination according to the designated path, avoid staff running around and passing through dangerous areas, reduce accidents, and improve work efficiency and reduce time waste.

[0049] Specifically, the image matching unit 342 is used to pass the optimized interaction feature vector between face images through a classifier to obtain a classification result, and the classification result is used to indicate whether the face detection image of the object to be identified matches the face reference image of the predetermined authority. That is, the interactive correlation feature information between the full-perceptual features of the face detection image and the full-perceptual features of the face reference image is used to perform classification processing, so as to evaluate the correlation and consistency of the features of the two face images, so as to realize the identity verification of the object to be identified. More specifically, the optimized interaction feature vector between face images is fully connected encoded using multiple fully connected layers of the classifier to obtain an encoded classification feature vector; and the encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain the classification result.

[0050] A classifier is a machine learning model or algorithm that is used to classify input data into different categories or labels. Classifiers are part of supervised learning and perform classification tasks by learning the mapping relationship from input data to output categories.

[0051] A fully connected layer is a common type of layer in neural networks. In a fully connected layer, each neuron is connected to all neurons in the previous layer, and each connection has a weight. This means that each neuron in the fully connected layer receives inputs from all neurons in the previous layer, and performs a weighted sum of these inputs through the weights, and then passes the result to the next layer.

[0052] The Softmax classification function is a commonly used activation function for multi-classification problems. It converts each element of the input vector into a probability value between 0 and 1, and the sum of these probability values ​​is equal to 1. The Softmax function is often used in the output layer of a neural network, especially for multi-classification problems, because it can map the network output to the probability distribution of each category. During the training process, the output of the Softmax function can be used to calculate the loss function and update the network parameters through the back propagation algorithm. It is worth noting that the output of the Softmax function does not change the relative size relationship between the elements, but only normalizes them. Therefore, the Softmax function does not change the characteristics of the input vector, but only converts it into a probability distribution form.

[0053] Specifically, the result generation unit 343 is used to generate a warning prompt and / or a path planning map based on the classification result. That is, based on the verification result, a warning prompt and / or a path planning map is generated. It is worth mentioning that in the technical solution of the present application, when performing face recognition and authority detection on the object to be identified, it is necessary to perform feature analysis and comparison based on the face detection image and each face reference image with predetermined authority in the database, so as to complete the authority matching and path planning. In this way, unauthorized persons can be prevented from entering dangerous areas, and illegal intrusions and thefts can be reduced. At the same time, warning prompts are generated for unauthorized persons to be identified, so that abnormal situations can be discovered in time and corresponding measures can be taken. Different path planning is provided for persons with different predetermined authorities to ensure that they can reach the destination along the designated path, avoid staff running around and passing through dangerous areas, and reduce the occurrence of accidents.

[0054] It is worth mentioning that in other specific examples of the present application, warning prompts and / or path planning maps can be generated based on the interaction features between facial images in other ways, for example: using the aforementioned method to obtain interaction features between facial images, these features capture the correlation and interaction information between facial images; formulating warning rules according to specific application scenarios and needs, for example, detecting a specific interaction feature pattern between facial images indicates potential danger or abnormality; analyzing the extracted interaction features between facial images, and judging whether there is potential danger or abnormality according to the warning rules; if there is potential danger or abnormality, generating corresponding warning prompts, which can be text messages, sound alarms or other forms of warning signals; determining the goal of path planning according to specific application scenarios and needs, such as finding a specific face in a crowd or performing a specific task; using a path planning algorithm to determine the best path or action strategy based on the interaction features between facial images to achieve the predetermined goal; presenting the results of path planning in a graphical manner, which can be path markings on a map, navigation instructions or other forms of path planning maps; outputting the generated warning prompts and / or path planning maps as results for applications such as real-time warning and path navigation.

[0055] As described above, the Internet of Things security management system 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with an Internet of Things security management algorithm. In a possible implementation, the Internet of Things security management system 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the Internet of Things security management system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the Internet of Things security management system 300 can also be one of the many hardware modules of the wireless terminal.

[0056] Alternatively, in another example, the IoT security management system 300 and the wireless terminal may also be separate devices, and the IoT security management system 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0057] Furthermore, a method for security management of the Internet of Things is also provided.

[0058] Figure 5 FIG. 1 is a flow chart of the IoT security management method according to an embodiment of the present application. Figure 5As shown, the Internet of Things security management method according to the embodiment of the present application includes the following steps: S1, obtaining a face detection image of an object to be identified captured by a camera; S2, obtaining a face reference image with predetermined authority from a database; S3, performing image feature interaction correlation analysis on the face detection image and the face reference image to obtain interaction features between face images; S4, generating an early warning prompt and / or a path planning map based on the interaction features between face images.

[0059] In summary, the IoT security management method according to the embodiment of the present application is explained, which collects the face detection image of the object to be identified through the camera, and introduces the image processing and analysis algorithm at the back end to perform feature comparison and analysis on the face detection image and the face reference image with predetermined authority, so as to verify the identity of the staff. This can prevent unauthorized personnel from entering dangerous areas and reduce illegal intrusion and theft. At the same time, the system can also generate early warning prompts based on the face recognition results, detect abnormal situations in time and take corresponding measures. In addition, according to the personnel with different predetermined authorities, the system can provide them with different path planning to ensure that they can reach the destination along the specified path. This can prevent staff from running around and passing through dangerous areas, reducing the occurrence of accidents. In addition, the path planning function can also improve work efficiency and reduce time waste.

[0060] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An Internet of Things security management system, characterized in that: include: A face detection image acquisition module is used to obtain a face detection image of an object to be identified acquired by a camera; A predetermined authority reference image acquisition module is used to acquire a face reference image with predetermined authority from a database; A face image feature interaction analysis module, used for performing image feature interaction correlation analysis on the face detection image and the face reference image to obtain interaction features between face images; An early warning and path planning module, used to generate early warning prompts and / or path planning diagrams based on the interaction features between the facial images; The face image feature interactive analysis module comprises: A face detection image enhancement unit, used for performing bilateral filtering on the face detection image to obtain an enhanced face detection image; A face image feature extraction unit, used for passing the enhanced face detection image and the face reference image through a dual detection network model to obtain a face detection feature map and a face reference feature map; A global feature interaction analysis unit between face images, used for performing full perceptual interaction correlation coding on the face detection feature map and the face reference feature map respectively to obtain the interaction features between the face images; The global feature interaction analysis unit between face images includes: A face feature full perception subunit, used for passing the face detection feature map and the face reference feature map through a full perception module to obtain a face detection full perception feature vector and a face reference full perception feature vector; a face full perception feature interaction subunit, configured to use the inter-feature attention layer to perform feature interaction based on an attention mechanism on the face detection full perception feature vector and the face reference full perception feature vector to obtain an inter-face image interaction feature vector as the inter-face image interaction feature; The warning and path planning module includes: A facial feature distribution optimization unit, used for performing Hilbert orthogonal space domain representation decoupling on the interaction feature vector between the facial images to obtain an optimized interaction feature vector between the facial images; An image matching unit, used for passing the optimized interactive feature vector between face images through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the face detection image of the object to be identified matches the face reference image of the predetermined authority; A result generating unit, used for generating a warning prompt and / or a path planning diagram based on the classification result; The facial feature distribution optimization unit is used to: perform Hilbert orthogonal space domain representation decoupling on the interaction feature vector between facial images using the following optimization formula to obtain the optimized interaction feature vector between facial images; Wherein, the optimization formula is: ; in is the interaction feature vector between the face images, is the global feature mean of the interaction feature vectors between the face images, is the binary norm of the interaction feature vector between the face images, is the length of the interaction feature vector between the face images, and is a unit vector, represents vector subtraction, represents the covariance matrix, is the optimized interaction feature vector between face images; The Hilbert orthogonal space domain representation is decoupled to emphasize the interaction feature vectors between the face images The essential domain-specific information in the diverse feature expressions of the face images is obtained by using the Hilbert space metric based on the vector self-space metric and the vector self-inner product representation. The orthogonal spatial domain decoupling of domain-invariant representation is performed within the overall domain representation of to enhance the interaction feature vector between the face images. Generalization performance of domain adaptation in the classification domain.

2. The Internet of Things security management system according to claim 1, characterized in that: The dual detection network model is a dual detection network model including a first image encoder and a second image encoder, and the first image encoder and the second image encoder have the same network structure.

3. A method for Internet of Things security management, characterized in that: include: Obtaining a face detection image of an object to be identified captured by a camera; Acquire a face reference image with predetermined authority from a database; Performing image feature interactive correlation analysis on the face detection image and the face reference image to obtain interactive features between face images; Based on the interaction features between the facial images, generate an early warning prompt and / or a path planning diagram; The performing image feature interactive correlation analysis on the face detection image and the face reference image to obtain interactive features between face images includes: Performing bilateral filtering on the face detection image to obtain an enhanced face detection image; Passing the enhanced face detection image and the face reference image through a dual detection network model to obtain a face detection feature map and a face reference feature map; Performing full perceptual interactive correlation coding on the face detection feature map and the face reference feature map respectively to obtain the interactive features between the face images; The performing full perceptual interactive correlation coding on the face detection feature map and the face reference feature map respectively to obtain the interactive features between the face images includes: Passing the face detection feature map and the face reference feature map through a full perception module to obtain a face detection full perception feature vector and a face reference full perception feature vector respectively; Using the inter-feature attention layer, perform feature interaction based on the attention mechanism on the face detection full-perception feature vector and the face reference full-perception feature vector to obtain an inter-face image interaction feature vector as the inter-face image interaction feature; The generating of an early warning prompt and / or a path planning diagram based on the interaction features between the face images includes: Performing Hilbert orthogonal space domain representation decoupling on the interaction feature vector between face images to obtain an optimized interaction feature vector between face images; Passing the optimized interaction feature vector between face images through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the face detection image of the object to be identified matches the face reference image of the predetermined authority; Based on the classification results, generate warning prompts and / or path planning maps; Decoupling the interaction feature vector between face images by Hilbert orthogonal space domain representation using the following optimization formula to obtain the optimized interaction feature vector between face images; Wherein, the optimization formula is: ; in is the interaction feature vector between the face images, is the global feature mean of the interaction feature vectors between the face images, is the binary norm of the interaction feature vector between the face images, is the length of the interaction feature vector between the face images, and is a unit vector, represents vector subtraction, represents the covariance matrix, is the optimized interaction feature vector between face images; The Hilbert orthogonal space domain representation is decoupled to emphasize the interaction feature vectors between the face images The essential domain-specific information in the diverse feature expressions of the face images is obtained by using the Hilbert space metric based on the vector self-space metric and the vector self-inner product representation. The orthogonal spatial domain decoupling of domain-invariant representation is performed within the overall domain representation of to enhance the interaction feature vector between the face images. Generalization performance of domain adaptation in the classification domain.

Images