End-cloud collaborative face image processing method and system

Abstract

The invention belongs to the technical field of face recognition, and particularly relates to an end-cloud collaborative face image processing method and system. Comprising the steps that a terminal device collects a face image, preprocesses the face image, extracts feature values of the face image, constructs the feature values into a data packet, encrypts the data packet, and transmits the encrypted data packet to a cloud server; and the cloud server receives the encrypted data packet, decrypts the encrypted data packet, acquires a feature value from the decrypted data packet, clusters and archives the feature value, performs cloud data aggregation according to a clustering and archiving result, and pushes the clustering and archiving result to the terminal equipment. According to the method, the face image feature values are extracted and uploaded through the terminal device, the cloud server performs clustering archiving and data aggregation, and the method has the advantages that on the premise of fully protecting user privacy and biological feature data security, the network transmission load and cloud processing pressure are remarkably reduced, and the efficiency and consistency of cross-device face data management are improved.

Description

A cloud-edge collaborative face image processing method and system Technical Field

[0001] This invention belongs to the field of face recognition technology, and in particular relates to a face image processing method and system with edge-cloud collaboration. Background Technology

[0002] Facial recognition technology, as an important branch of biometric identification, has been widely used in many fields such as identity verification, intelligent security, and personalized services. With the popularization of mobile internet and smart terminals, users' demands for facial image processing are increasing daily. They not only require accurate recognition and rapid response, but also expect cross-device data synchronization and unified management. Currently, common facial image processing solutions are mainly divided into two categories: one is a purely localized processing solution that completes image acquisition, feature extraction, and comparison analysis entirely on the terminal device; the other is a purely cloud-based processing solution that directly uploads the original facial image to the cloud and relies on cloud servers to complete all processing.

[0003] However, purely localized processing solutions are limited by the computing resources and storage capacity of terminal devices, making it difficult to handle complex and computationally intensive deep learning models. This is especially true when facing tasks such as large-scale face clustering and cross-device archiving, where local algorithms are often inefficient and can lead to increased power consumption and significant overheating. Furthermore, the processing results generated by this solution are typically stored locally and cannot be synchronized or shared across different devices, limiting the consistency and convenience of use. On the other hand, while purely cloud-based processing solutions can leverage the powerful computing capabilities of the cloud for efficient analysis and archiving, they require terminals to upload raw images containing sensitive biometric information. This not only consumes significant network bandwidth and increases transmission latency but also introduces significant privacy risks, failing to meet users' increasingly stringent requirements for data security.

[0004] Therefore, there is an urgent need for a face image processing method and system that can take into account both terminal privacy and security and the advantages of cloud computing, and achieve end-to-cloud collaborative processing. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a cloud-edge collaborative face image processing method and system. This method and system have the advantages of significantly reducing network transmission load and cloud processing pressure while fully protecting user privacy and biometric data security, and improving the efficiency and consistency of cross-device face data management.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a face image processing method with end-to-end cloud collaboration, comprising: a terminal device acquiring a face image, preprocessing the face image, extracting feature values ​​from the face image, constructing a data packet from the feature values, encrypting the data packet, and transmitting the encrypted data packet to a cloud server; the cloud server receiving the encrypted data packet, decrypting the encrypted data packet, obtaining feature values ​​from the decrypted data packet, clustering and archiving the feature values, performing cloud data aggregation based on the clustering and archiving results, and pushing the clustering and archiving results to the terminal device.

[0007] Furthermore, the preprocessing of the face image includes: calling an image signal processor to perform one or more of the following on the face image: multi-frame noise reduction, HDR synthesis, and face beautification.

[0008] Furthermore, after preprocessing the face image, the method also includes: evaluating the quality of the face image's blurriness, brightness, and face pose.

[0009] Furthermore, the extraction of feature values ​​from the face image includes: running a deep convolutional neural network to map the face region image in the face image into a fixed-length floating-point vector.

[0010] Furthermore, in the process of constructing the feature values ​​into a data packet, the data packet includes a unique image identifier, facial feature values, image capture time, and device identifier.

[0011] Furthermore, the step of clustering and archiving the feature values ​​includes: performing a nearest neighbor search on the feature values, determining similarity based on the nearest neighbor search results, and clustering the feature values ​​according to the determination results.

[0012] Furthermore, the nearest neighbor search for the feature value includes: using the HNSW algorithm to quickly search for several nearest neighbor features most similar to the feature value in the global feature map.

[0013] Furthermore, the similarity determination based on the nearest neighbor search results includes: calculating the cosine similarity between the feature value and each of the nearest neighbor features, and determining whether each cosine similarity exceeds a preset threshold.

[0014] Furthermore, the clustering and assignment of feature values ​​based on the determination result includes: if there are neighboring features with a cosine similarity exceeding a preset threshold, then the feature value is added to the cluster to which the neighboring feature with the maximum cosine similarity belongs; if there are no neighboring features with a cosine similarity exceeding the preset threshold, then the feature value is added to a new cluster.

[0015] This invention also provides an edge-cloud collaborative face image processing system, including a terminal device and a cloud server, wherein: the terminal device includes: a face acquisition module for acquiring face images; a preprocessing module for preprocessing the face images; a feature value extraction module for extracting feature values ​​from the face images; a data packet construction module for constructing data packets from the feature values; an encryption module for encrypting the data packets; and a sending module for transmitting the encrypted data packets to the cloud server; the cloud server includes: a receiving module for receiving the encrypted data packets; a decryption module for decrypting the encrypted data packets and obtaining feature values ​​from the decrypted data packets; a clustering and archiving module for clustering and archiving the feature values; a data aggregation module for performing cloud data aggregation based on the clustering and archiving results; and a push module for pushing the clustering and archiving results to the terminal device.

[0016] The beneficial effects of this invention are as follows: This invention extracts facial image feature values ​​from terminal devices and encrypts and uploads them. After decryption by the cloud server, clustering, archiving, and data aggregation are performed. This achieves both protection of users' original biometric data privacy and efficient and accurate unified management of cross-device facial data using cloud computing power. By adding an image quality assessment step after preprocessing, unqualified facial images such as blurry, overly dark, or poorly posed images are filtered out, saving unnecessary feature extraction and transmission overhead and improving the overall processing efficiency of the system. This invention enables preprocessing and feature extraction of facial images to be completed on the terminal side, transmitting only the encrypted feature data to the cloud. The cloud performs efficient clustering analysis and global archiving, and synchronizes the results to the terminal. It has the advantages of significantly reducing network transmission load and cloud processing pressure while fully protecting user privacy and biometric data security, and improving the efficiency and consistency of cross-device facial data management. Attached Figure Description

[0017] Figure 1 is a flowchart of the face image processing method of the present invention; Figure 2 is a flowchart of the face image processing method of the present invention on a terminal device; Figure 3 is a flowchart of the face image processing method of the present invention on a cloud server; Figure 4 is a flowchart of the face image processing method of the present invention on a cloud server; Figure 5 is a schematic diagram of the structure of the face image processing system of the present invention. Detailed Implementation

[0018] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0019] In the description of this invention, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0020] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] In the embodiments of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0022] Referring to Figures 1 to 5, the figures illustrate a specific embodiment of the edge-cloud collaborative face image processing method and system provided by the present invention.

[0023] Referring to Figure 1, the edge-cloud collaborative face image processing method includes: the terminal device acquiring a face image, preprocessing the face image, extracting feature values ​​from the face image, constructing a data packet from the feature values, encrypting the data packet, and transmitting the encrypted data packet to the cloud server; the cloud server receiving the encrypted data packet, decrypting the encrypted data packet, obtaining feature values ​​from the decrypted data packet, clustering and archiving the feature values, performing cloud data aggregation based on the clustering and archiving results, and pushing the clustering and archiving results to the terminal device.

[0024] In the above embodiments, the terminal device refers to a device with camera and network transmission capabilities, including but not limited to mobile phones, desktop computers, laptops, tablets, and in-vehicle computers. This embodiment takes a smartphone as an example, utilizing the smartphone's main camera, front-facing camera, and ultra-wide-angle camera to capture facial images. This multi-device compatibility design reflects the universality and scalability of the solution, covering not only mainstream consumer electronics categories but also reserving access capabilities for new terminals in IoT scenarios, ensuring the broad applicability of the technical solution in practical applications.

[0025] In the above embodiments, preprocessing of the face image includes: calling an image signal processor to perform one or more of the following on the face image: multi-frame noise reduction, HDR synthesis, and face beautification. In these embodiments, to ensure the quality of the input image, the face image captured by the smartphone should be preprocessed before feature value extraction. Specifically, an image signal processor (ISP) can be used for multi-frame noise reduction, HDR synthesis, and face beautification. By calling the image signal processor on the terminal to perform one or more preprocessing operations, the quality of the face image is improved, providing clearer and more standardized input data for subsequent feature extraction. This step fully utilizes the local computing resources of the terminal device, transforming the original image into a high-quality, standardized input, effectively reducing the impact of image quality fluctuations caused by complex ambient lighting, camera shake, or device differences on the accuracy of subsequent feature extraction. This is a crucial foundation for ensuring the overall recognition performance of the system.

[0026] Referring to Figure 2, in the above embodiment, after preprocessing the face image, the method further includes: quality assessment of the face image's blurriness, brightness, and face pose. In this embodiment, the quality assessment step primarily aims to control the quality of the captured image. A built-in rule-based quality assessment module evaluates the blurriness, brightness, and face pose of the face image, filtering out images with low quality or that are not human faces. For face images that meet the quality requirements, the subsequent face image feature extraction step is initiated; for images with low quality or that are not human faces, a retry or abandonment of the current image processing is selected. The image quality assessment stage filters out unqualified face images such as those that are blurry, too dark, or have poor poses, saving unnecessary feature extraction and transmission overhead and improving the overall processing efficiency of the system.

[0027] In the above embodiments, extracting feature values ​​from a face image includes running a deep convolutional neural network to map the face region image in the face image into a fixed-length floating-point vector. In these embodiments, the deep convolutional neural network can be run by calling a dedicated neural network processing unit (NPU) or an AI acceleration engine on the mobile device. The NPU is a dedicated processor designed for artificial intelligence computing needs, accelerating neural network operations through hardware architecture innovation, forming a complementary heterogeneous computing system with CPUs and GPUs. In this embodiment, the deep convolutional neural network run by the NPU is a lightweight yet high-performance deep convolutional neural network suitable for smartphones. This embodiment uses MobileFaceNet and ShuffleNet as examples. MobileFaceNet is a deep convolutional neural network running on mobile devices, based on the MobileNetV2 architecture and featuring a lightweight design suitable for mobile deployment. Its core objective is to significantly reduce model size and computational requirements while maintaining recognition accuracy, adapting to the hardware limitations of mobile devices, achieving a balance between lightweight design and high accuracy in mobile face recognition. ShuffleNet is a lightweight convolutional neural network architecture designed for mobile and embedded devices. It aims to achieve efficient image recognition tasks in computationally limited environments, with the core objective of significantly reducing computational cost and model size while maintaining high classification accuracy. In this embodiment, the deep convolutional neural network maps the detected face region image into a fixed-length, high-dimensional floating-point vector, which represents the face feature value. In this example, 512 dimensions are used. By running the deep convolutional neural network to map the face region into a fixed-length vector, highly discriminative and robust deep features are extracted, making subsequent similarity calculations and clustering more accurate and reliable.

[0028] In the above embodiment, the smartphone generates a structured data packet based on the facial feature values ​​extracted by the deep convolutional neural network. This data packet, constructed from feature values, includes a unique image identifier, facial feature values, image capture time, and device identifier. Specifically, the data packet includes the unique image identifier `image_id`, the facial feature value `face_embedding`, the image capture time `timestamp`, the device identifier `device_id`, and other metadata. This other metadata is optional, such as the geographical location where the image was captured. This allows for the association of feature values ​​with metadata, facilitating data traceability, device differentiation, and more granular aggregation management in the cloud. This structured data packet design encapsulates core features with rich contextual information (metadata), enabling the cloud server to not only process facial features but also perform comprehensive analysis and strategy formulation by combining multi-dimensional information such as time, space, and device. For example, it can implement services such as location-based alerts and multi-device behavior analysis for the same user, expanding the system's application potential.

[0029] In the above embodiments, the smartphone encrypts the generated data packets, using either HTTPS encryption or Chinese cryptographic algorithms. The encrypted data packets are then transmitted to a cloud server, where the image processing results are awaited. This dual-option encryption strategy of HTTPS and Chinese cryptographic algorithms ensures the confidentiality and integrity of sensitive biometric information such as facial features during transmission, effectively defending against network threats such as man-in-the-middle attacks and establishing a reliable secure transmission channel.

[0030] In the above embodiment, after the cloud server receives the encrypted data packet from the smartphone, the cloud server decrypts the encrypted data packet to restore the data packet before encryption. The data packet includes feature values ​​and metadata.

[0031] Referring to Figure 3, in the above embodiment, after obtaining the feature values ​​from the data packet, the feature values ​​are clustered and archived. Specifically, a nearest neighbor search is first performed on the feature values, then a similarity determination is made based on the nearest neighbor search results, and finally, the feature values ​​are clustered and assigned according to the determination results.

[0032] Specifically, the nearest neighbor search for feature values ​​includes using the Hierarchical Navigable Small World (HNSW) algorithm to quickly search for the most similar nearest neighbor features in the global feature graph. In this embodiment, the Hierarchical Navigable Small World (HNSW) is an approximate nearest neighbor search algorithm based on a hierarchical graph structure. This embodiment uses the HNSW algorithm framework as the core of cloud-based face feature retrieval, achieving efficient searching through a multi-layered graph structure. The cloud server maintains a global, dynamically growing HNSW graph index, and each newly uploaded face feature value is inserted as a node into the HNSW graph. In this embodiment, when a new feature value arrives in the cloud, the HNSW algorithm is used to quickly search for its K most similar nearest neighbor features in the existing global feature graph. In this embodiment, by calculating cosine similarity, the 10 features closest to the current feature value are identified and considered as nearest neighbor features. Using the HNSW algorithm for nearest neighbor search is key to the high efficiency of this solution's cloud processing. Compared to traditional linear scanning or tree-based indexing, HNSW can return near-optimal solutions in sublinear time complexity when dealing with massive, high-dimensional feature vectors, perfectly balancing search speed and accuracy. This enables the system to support real-time retrieval and clustering of face feature databases with hundreds of millions or even billions of features, demonstrating excellent scalability.

[0033] Specifically, similarity determination based on nearest neighbor search results includes: calculating the cosine similarity between the feature value and each nearest neighbor feature, and determining whether each cosine similarity exceeds a preset threshold. In this embodiment, the preset threshold is set to 0.7, meaning that among the 10 nearest neighbor features found using the HNSW algorithm, the cosine similarity between each nearest neighbor feature and the new feature value is determined to be greater than 0.7. The introduction of the preset similarity threshold provides a clear and adjustable technical standard for clustering determination. By adjusting this threshold, the "tightness" of system clustering can be flexibly controlled to adapt to the needs of different security levels or application scenarios.

[0034] Referring to Figure 4, specifically, the clustering and assignment of feature values ​​based on the determination result includes: if there are nearest neighbor features with a cosine similarity exceeding a preset threshold, the feature value is added to the cluster to which the nearest neighbor feature with the maximum cosine similarity belongs; if there are no nearest neighbor features with a cosine similarity exceeding the preset threshold, the feature value is added to a new cluster. In this embodiment, after determining whether the cosine similarity between each nearest neighbor feature and the new feature value exceeds 0.7, there will be three possible outcomes. If only one nearest neighbor feature has a cosine similarity exceeding 0.7 with the new feature value, the feature value is added to the cluster to which that nearest neighbor feature belongs; if multiple nearest neighbor features have a cosine similarity exceeding 0.7 with the new feature value, the feature value is added to the cluster to which the nearest neighbor feature with the maximum cosine similarity belongs; if there are no nearest neighbor features with a cosine similarity exceeding 0.7, a new cluster ID is created for this new feature, and the new feature value is added to the new cluster. This dynamic incremental clustering mechanism does not require pre-setting the number of categories and can adaptively discover and create new face categories (i.e., new individuals), making it suitable for practical application scenarios where the face database is dynamically growing.

[0035] In the above embodiments, cloud-based data aggregation is performed based on the clustering archive results, specifically by storing the final aggregation results in a cloud database (such as MySQL or MongoDB). The data structure is roughly as follows: a unique cluster identifier `cluster_id`, representing a person; a list of image IDs `list_of_image_ids`, representing a list of all image IDs belonging to this person; and a representative feature `representative_embedding`, representing the representative feature of this cluster. Cloud-based data aggregation organizes scattered, single-shot facial feature recognition results into structured information centered on "people" and continuously evolving. The `representative_embedding`, obtained by calculating the average or centroid of all features within a cluster, serves as an abstract representation of that cluster, facilitating rapid subsequent comparisons. Simultaneously, `list_of_image_ids` is associated with all original image indexes, achieving decoupling and efficient management of feature data and original image data, providing possibilities for big data analysis and person profiling.

[0036] Finally, the clustering and archiving results are pushed to the terminal device, notifying it whether the currently recognized face image belongs to a specific person identified in the cloud database. This allows the terminal device to execute subsequent business processes or terminate the current one upon receiving the result. In this embodiment, the result push completes the closed loop of end-to-cloud collaboration. After completing perception and feature generation locally, the terminal device leverages the powerful computing capabilities and global data of the cloud to make a "cognitive" decision and feeds the decision back to the terminal to drive specific business applications (such as access control, payment confirmation, and intelligent photo categorization). This collaborative model combines the advantages of real-time performance and privacy of the terminal with the global and intelligent advantages of the cloud, making it an ideal architecture for building an efficient, intelligent, and secure face recognition system.

[0037] Referring to Figure 5, this embodiment also provides an edge-cloud collaborative face image processing system, including a terminal device and a cloud server, wherein: the terminal device includes: a face acquisition module for acquiring face images; a preprocessing module for preprocessing the face images; specifically, the preprocessing module includes an image signal processing unit for performing multi-frame noise reduction, HDR synthesis, and face beautification on the face images; a quality assessment module for assessing the blur, brightness, and face pose of the face images, and filtering images with low quality or non-faces; a feature value extraction module for extracting feature values ​​from the face images; specifically, it runs a deep convolutional neural network to map the face region image in the face image into a fixed-length floating-point vector; a data packet construction module for constructing data packets from the feature values; specifically, the data packet includes a unique image identifier, face feature values, image capture time, and device identifier; and an encryption module for encrypting data... The data packets are encrypted; a sending module transmits the encrypted data packets to the cloud server; the cloud server includes: a receiving module for receiving encrypted data packets; a decryption module for decrypting the encrypted data packets and extracting feature values ​​from the decrypted data packets; a clustering and archiving module for clustering and archiving the feature values, including a nearest neighbor search unit, a similarity determination unit, and a clustering assignment unit; wherein, the nearest neighbor search unit is used to quickly search for several nearest neighbor features most similar to the feature value in the global feature map, the similarity determination unit is used to calculate the cosine similarity between the feature value and each nearest neighbor feature and determine whether each cosine similarity exceeds a preset threshold, and the clustering assignment unit is used to classify the feature value into an existing cluster or a new cluster according to the determination result of the similarity determination unit; a data aggregation module is used to perform cloud data aggregation based on the clustering and archiving results; specifically, the final aggregation results are stored in the cloud database.

[0038] The push module is used to push clustering archive results to terminal devices.

[0039] In summary, this embodiment provides a cloud-edge collaborative face image processing method and system. It extracts and encrypts face image feature values ​​from terminal devices, then uploads them. The cloud server decrypts the data and performs clustering, archiving, and data aggregation. This approach protects the privacy of users' original biometric data while leveraging cloud computing power for efficient and accurate unified management of face data across devices. By adding an image quality assessment step after preprocessing, it filters out unqualified face images such as blurry, overly dark, or poorly posed images, saving unnecessary feature extraction and transmission overhead and improving overall system processing efficiency. This embodiment enables face image preprocessing and feature extraction to be completed on the terminal side, transmitting only encrypted feature data to the cloud. The cloud performs efficient clustering analysis and global archiving, synchronizing the results back to the terminal. This approach significantly reduces network transmission load and cloud processing pressure while fully protecting user privacy and biometric data security, improving the efficiency and consistency of cross-device face data management.

[0040] The embodiments described above are merely one of the preferred embodiments of the present invention. Ordinary variations and substitutions made by those skilled in the art within the scope of the technical solution of the present invention should be included within the protection scope of the present invention.

Claims

1. A face image processing method with edge-cloud collaboration, characterized in that, include: The terminal device acquires a face image, preprocesses the face image, extracts the feature values ​​of the face image, constructs a data packet from the feature values, encrypts the data packet, and transmits the encrypted data packet to the cloud server. The cloud server receives the encrypted data packet, decrypts the encrypted data packet, obtains feature values ​​from the decrypted data packet, clusters and archives the feature values, performs cloud data aggregation based on the clustering and archiving results, and pushes the clustering and archiving results to the terminal device.

2. The face image processing method with edge-cloud collaboration according to claim 1, characterized in that, The preprocessing of the face image includes: calling an image signal processor to perform one or more of the following on the face image: multi-frame noise reduction, HDR synthesis, and face beautification.

3. The face image processing method with edge-cloud collaboration according to claim 2, characterized in that, After preprocessing the face image, the process also includes: evaluating the quality of the face image in terms of blur, brightness, and facial pose.

4. The face image processing method with edge-cloud collaboration according to claim 2, characterized in that, The extraction of feature values ​​from the face image includes: running a deep convolutional neural network to map the face region image in the face image into a fixed-length floating-point vector.

5. The face image processing method with end-to-end cloud collaboration according to claim 4, characterized in that, In the process of constructing the data packet from the feature values, the data packet includes a unique image identifier, facial feature values, image capture time, and device identifier.

6. A face image processing method with edge-cloud collaboration according to any one of claims 1-5, characterized in that, The process of clustering and archiving the feature values ​​includes: performing a nearest neighbor search on the feature values, determining similarity based on the nearest neighbor search results, and assigning the feature values ​​to clusters based on the determination results.

7. The face image processing method with edge-cloud collaboration according to claim 6, characterized in that, The nearest neighbor search for the feature value includes: using the HNSW algorithm to quickly search for several nearest neighbor features that are most similar to the feature value in the global feature map.

8. The face image processing method with edge-cloud collaboration according to claim 7, characterized in that, The similarity determination based on the nearest neighbor search results includes: calculating the cosine similarity between the feature value and each of the nearest neighbor features, and determining whether each cosine similarity exceeds a preset threshold.

9. The face image processing method with edge-cloud collaboration according to claim 8, characterized in that, The clustering and assignment of feature values ​​based on the determination result includes: if there are neighboring features with a cosine similarity exceeding a preset threshold, the feature value is added to the cluster to which the neighboring feature with the maximum cosine similarity belongs; if there are no neighboring features with a cosine similarity exceeding the preset threshold, the feature value is added to a new cluster.

10. A cloud-edge collaborative face image processing system, characterized in that, The system includes a terminal device and a cloud server. The terminal device includes: a face acquisition module for acquiring face images; a preprocessing module for preprocessing the face images; a feature value extraction module for extracting feature values ​​from the face images; a data packet construction module for constructing data packets from the feature values; an encryption module for encrypting the data packets; and a sending module for transmitting the encrypted data packets to the cloud server. The cloud server includes: a receiving module for receiving the encrypted data packets; a decryption module for decrypting the encrypted data packets and extracting feature values ​​from the decrypted data packets; a clustering and archiving module for clustering and archiving the feature values; a data aggregation module for performing cloud data aggregation based on the clustering and archiving results; and a push module for pushing the clustering and archiving results to the terminal device.

Images