A cross-domain electronic fence cooperative early warning method and system based on intelligent learning
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING FUSION HSBC TECH CO LTD
- Filing Date
- 2026-03-03
- Publication Date
- 2026-06-09
AI Technical Summary
Traditional electronic fences suffer from problems such as repeated alarms, missed alarms, and high false alarm rates when deployed across domains. Furthermore, fixed threshold rules cannot adapt to dynamic environmental changes, leading to delayed warnings and missed intervention windows.
Cross-domain electronic fence collaborative early warning is achieved through intelligent learning methods. Lightweight homomorphic encryption and federated learning are used to achieve cross-domain data collaboration, dynamically adjust thresholds, and adopt encrypted trajectory packet push and cascading message mechanisms to ensure the real-time transmission and accuracy of early warning signals.
It significantly reduced the false alarm rate, achieved real-time and accurate early warning of cross-domain targets, and ensured second-level synchronization of early warning signals and tamper-proof and reliable recording of the entire process.
Smart Images
Figure CN122176841A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of electronic fence technology, and in particular to a cross-domain electronic fence collaborative early warning method and system based on intelligent learning. Background Technology
[0002] With the rapid expansion of scenarios such as hazardous chemical transportation, low-altitude navigation, and waterway law enforcement, the "cross-domain" mode, where targets need to continuously traverse multiple management entities and geographical areas, has become the norm. Traditional electronic fences are deployed independently in each domain, with different coordinate systems, communication protocols, and risk thresholds. This leads to a situation where "repeated alarms and missed alarms" coexist once a target crosses the boundary. At the same time, fixed threshold rules cannot dynamically evolve with weather, traffic flow, or historical boundary crossing probabilities, resulting in a high false alarm rate. Furthermore, different domains cannot share the crucial context of "about to enter," leading to warning delays that generally exceed 3 seconds, missing the window for early intervention.
[0003] In recent years, although artificial intelligence has been able to reduce the false alarm rate to 5% in single-domain video analysis, it is difficult to solve the dual shortcomings of "cross-domain collaboration" and "strategy adaptation" by only using data from this domain. Summary of the Invention
[0004] Based on this, the embodiments of this application provide a cross-domain electronic fence collaborative early warning method and system based on intelligent learning, which can break down data silos between domains through intelligent learning, enable multi-domain computing power collaboration of fences, and greatly reduce the false alarm rate.
[0005] Firstly, a cross-domain electronic fence collaborative early warning method based on intelligent learning is provided, which includes:
[0006] Multi-source sensing data of the target is collected in real time at each domain edge node, and the multi-source sensing data is spatiotemporally aligned and feature extracted to obtain the target trajectory embedding vector in a unified coordinate system.
[0007] The trajectory embedding vector is encrypted into a ciphertext trajectory packet using a lightweight homomorphic encryption algorithm and actively pushed to the downstream domain edge node along the target movement direction to achieve cross-domain identity relay.
[0008] Downstream domain edge nodes perform similarity calculations on the encrypted trajectory packets in encrypted state to generate a boundary crossing risk score, and compare the boundary crossing risk score with a local dynamic threshold to obtain a preliminary warning result;
[0009] Each domain edge node uploads the initial warning results to the federated learning coordination node. The coordination node trains the threshold update model based on global risk samples and sends the encrypted gradient back to each domain to complete dynamic threshold self-learning.
[0010] When the risk score of any domain exceeds the updated dynamic threshold, a collaborative early warning command is triggered, which links the execution mechanism of this domain with the edge nodes of adjacent domains to complete audio-visual interception, path guidance or remote vehicle locking operations.
[0011] Optionally, the multi-source sensing data includes satellite positioning messages, radio frequency identification, video structured features, vehicle bus speed, and visibility from weather sensors;
[0012] The spatiotemporal alignment is achieved through network-wide time synchronization and national standard coordinate transformation. Feature extraction uses a lightweight convolutional network to compress the original high-dimensional data into a fixed-dimensional trajectory embedding vector.
[0013] Optionally, the lightweight homomorphic encryption algorithm adopts a hierarchical homomorphic scheme, which encapsulates the trajectory embedding vector of multiple consecutive frames in a batch processing slot at one time. The encryption latency is lower than the real-time frame interval, and only one set of keys is needed to support concurrent decryption and verification of multiple nodes in the downstream domain.
[0014] Optionally, the ciphertext state similarity calculation uses an inner product approximation kernel function, and completes the trajectory cosine similarity evaluation in the ciphertext domain through polynomial Taylor expansion.
[0015] Optionally, the dynamic threshold self-learning step includes:
[0016] After the coordinating node aggregates the encrypted gradients from each domain, it executes the federated averaging algorithm, adds noise using differential privacy, and trains the threshold update model.
[0017] Each domain edge node downloads the encrypted model parameters and decrypts them in the local trusted execution environment, updating the local dynamic threshold. The update cycle is shorter than the event response cycle.
[0018] Optionally, the collaborative early warning instruction adopts the cascading publish-subscribe mechanism of the message queue telemetry transmission protocol, with the priority queue set to the highest level, the end-to-end transmission delay being lower than the early warning response time limit, and the adjacent domain edge node starts the preset linkage script after receiving the instruction to complete the synchronous execution of the local domain's audible and visual alarm, the barrier gate lowering, and the upstream guidance screen prompt.
[0019] Optionally, the method further includes storing an immutable log of the entire process on the blockchain for evidence preservation, specifically:
[0020] The immutable log on-chain evidence storage adopts a consortium blockchain architecture, which writes the warning timestamp, target anonymity identifier, risk score, threshold version number and execution result into the block. The block writing latency is lower than the event recording time limit, and it supports regulators to conduct verifiable queries under key escrow through smart contracts.
[0021] Secondly, a cross-domain electronic fence collaborative early warning system based on intelligent learning is provided, the system comprising:
[0022] The acquisition module is used to acquire multi-source sensing data of the target in real time at each domain edge node, and to perform spatiotemporal alignment and feature extraction on the multi-source sensing data to obtain the target trajectory embedding vector in a unified coordinate system.
[0023] The encryption module is used to encrypt the trajectory embedding vector into a ciphertext trajectory package using a lightweight homomorphic encryption algorithm, and actively push it to the downstream domain edge node along the target movement direction to realize cross-domain identity relay.
[0024] The comparison module is used by downstream domain edge nodes to calculate the similarity of the encrypted trajectory packet in encrypted state, generate an out-of-bounds risk score, and compare the out-of-bounds risk score with the local dynamic threshold to obtain a preliminary warning result.
[0025] The training module is used by the edge nodes of each domain to upload the initial warning results to the federated learning coordination node. The coordination node trains the threshold update model based on global risk samples and sends the encrypted gradient back to each domain to complete the dynamic threshold self-learning.
[0026] The early warning module is used to trigger a collaborative early warning command when the risk score of any domain exceeds the updated dynamic threshold. This command links the execution mechanism of this domain with the edge nodes of adjacent domains to complete audio-visual interception, path guidance, or remote vehicle locking operations.
[0027] Thirdly, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the methods described in the first aspect above.
[0028] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements any of the methods described in the first aspect above.
[0029] The beneficial effects of the technical solutions provided in this application include at least the following:
[0030] (1) By using the encrypted trajectory packet + cross-domain active push mechanism, the target can deliver the encrypted identity vector to the downstream domain the moment it leaves the upstream domain, realizing identity relay without decryption and without exposing privacy, fundamentally eliminating the risk of plaintext leakage caused by the traditional method of decryption before comparison.
[0031] (2) Federated learning-driven dynamic threshold self-learning enables each edge node to continuously fine-tune the local alarm threshold according to the global risk distribution, which avoids the proliferation of false alarms caused by fixed thresholds and can automatically improve sensitivity in sudden high-risk scenarios, significantly reducing the false alarm rate of the entire system.
[0032] (3) The collaborative early warning command adopts a combination of cascading message mechanism and on-chain evidence storage to ensure that the linkage signal is synchronized between adjacent domains within seconds, and the log of the whole process is tamper-proof, providing a reliable basis for post-event audit. Attached Figure Description
[0033] To more clearly illustrate the embodiments of this application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.
[0034] Figure 1 A flowchart illustrating the steps of a cross-domain electronic fence collaborative early warning method based on intelligent learning, provided in this application embodiment;
[0035] Figure 2 A system block diagram of a cross-domain electronic fence collaborative early warning method based on intelligent learning provided in this application embodiment;
[0036] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation
[0037] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0038] In the description of this application, the terms "comprising," "having," and any variations thereof are intended to cover non-exclusive inclusion, such as a process, method, system, product, or apparatus that includes a series of steps or units, not necessarily limited to those steps or units that are expressly listed, but may also include other steps or units that are not expressly listed but are inherent to these processes, methods, products, or apparatuses, or steps or units added based on further optimizations conceived in this application.
[0039] Please refer to Figure 1 The document illustrates a flowchart of a cross-domain electronic fence collaborative early warning method based on intelligent learning, provided in an embodiment of this application. This method may include the following steps:
[0040] S1 collects multi-source sensing data of the target in real time at the edge nodes of each domain, and performs spatiotemporal alignment and feature extraction on the multi-source sensing data to obtain the target trajectory embedding vector under a unified coordinate system.
[0041] In this embodiment, GNSS receivers, RFID readers, high-definition cameras, and meteorological sensors are deployed at edge nodes such as highway checkpoints, waterway shore bases, or UAV take-off and landing points. Each sensor is uniformly connected to a local industrial control computer via PoE. The industrial control computer has an embedded spatiotemporal alignment module that first completes millisecond-level clock synchronization using NTP messages, and then uniformly converts GNSS latitude and longitude, RFID identification codes, video frame target frames, and meteorological visibility to CGCS2000 plane coordinates. Subsequently, a lightweight 1D-CNN network is called to compress 32 consecutive frames of data into a 256-dimensional floating-point vector, which is stored in the local cache as the target trajectory embedding vector, awaiting the next step of encryption.
[0042] S2 encrypts the trajectory embedding vector into a ciphertext trajectory packet using a lightweight homomorphic encryption algorithm, and actively pushes it to the downstream domain edge node along the target movement direction to achieve cross-domain identity relay.
[0043] In this embodiment, the trajectory embedding vector in the cache is sent to a lightweight homomorphic encryption unit based on the CKKS scheme. This unit pre-generates and resides a set of rotation keys and relinearization keys, and packages 32 frames of vectors into a single ciphertext trajectory packet at once. The size of the ciphertext is approximately 1.8 times that of the original data. After encryption, the edge node parses the downstream domain MQTT topic name according to the target's direction of travel and actively pushes the ciphertext trajectory packet through the TLS tunnel, realizing cross-domain relay of "identity data following the target". The entire process is completed locally without uploading plaintext.
[0044] S3, the downstream domain edge node performs similarity calculation on the ciphertext trajectory packet in ciphertext state, generates a boundary crossing risk score, and compares the boundary crossing risk score with the local dynamic threshold to obtain the preliminary warning result.
[0045] In this embodiment, after receiving the encrypted trajectory packet, the downstream domain edge node directly calls the encrypted operation core embedded in the FPGA to calculate the out-of-bounds risk score between the current packet and the local historical packets in the encrypted domain through Taylor polynomial approximation of cosine similarity, and obtains a floating-point risk value. This value is then compared with the threshold in the local dynamic threshold register. If it is higher than the threshold, a primary warning result is generated and written to the local message queue. If it is lower than the threshold, only a log is recorded. The original trajectory vector does not need to be decrypted throughout the process.
[0046] S4, each domain edge node uploads the initial warning results to the federated learning coordination node. The coordination node trains the threshold update model based on global risk samples and sends the encrypted gradient back to each domain to complete the dynamic threshold self-learning.
[0047] In this embodiment, every five minutes, each domain edge node de-identifies the initial warning results and encapsulates them into gradient messages, which are then uploaded to the federated learning coordination node via bidirectional TLS. After collecting all messages, the coordination node performs FedAvg aggregation, injects differential privacy noise into the aggregated gradient, and then distributes the encrypted model file. The edge node decrypts the model in its local TEE, updates the dynamic threshold register, and completes a self-learning loop of "global risk - local threshold" to ensure that the threshold evolves in real time with the environment and threats.
[0048] S5: When the risk score of any domain exceeds the updated dynamic threshold, a collaborative early warning command is triggered, which links the execution mechanism of this domain with the edge nodes of adjacent domains to complete the sound and light interception, path guidance or remote vehicle locking operation.
[0049] In this embodiment, when the comparison unit detects that the risk score exceeds the latest dynamic threshold, it immediately publishes a "Collaborative Early Warning" topic on the local MQTT. The message carries the event ID, anonymous target code, and recommended handling level. The local audible and visual alarm, barrier gate, or vehicle locking device subscribes to this topic and responds in milliseconds. At the same time, the message is cascaded and forwarded to the edge nodes of adjacent domains, triggering their pre-set linkage scripts to achieve multi-domain synchronous interception, path guidance, and remote vehicle locking. The operation log is also written to the consortium blockchain block to ensure subsequent traceability.
[0050] like Figure 2 This application also provides a cross-domain electronic fence collaborative early warning system based on intelligent learning, which may include:
[0051] The acquisition module is used to acquire multi-source sensing data of the target in real time at each domain edge node, and to perform spatiotemporal alignment and feature extraction on the multi-source sensing data to obtain the target trajectory embedding vector in a unified coordinate system.
[0052] The encryption module is used to encrypt the trajectory embedding vector into a ciphertext trajectory package using a lightweight homomorphic encryption algorithm, and actively push it to the downstream domain edge node along the target movement direction to realize cross-domain identity relay.
[0053] The comparison module is used by downstream domain edge nodes to calculate the similarity of the encrypted trajectory packet in encrypted state, generate an out-of-bounds risk score, and compare the out-of-bounds risk score with the local dynamic threshold to obtain a preliminary warning result.
[0054] The training module is used by the edge nodes of each domain to upload the initial warning results to the federated learning coordination node. The coordination node trains the threshold update model based on global risk samples and sends the encrypted gradient back to each domain to complete the dynamic threshold self-learning.
[0055] The early warning module is used to trigger a collaborative early warning command when the risk score of any domain exceeds the updated dynamic threshold. This command links the execution mechanism of this domain with the edge nodes of adjacent domains to complete audio-visual interception, path guidance, or remote vehicle locking operations.
[0056] Specific limitations regarding the intelligent learning-based cross-domain electronic fence collaborative early warning system can be found in the limitations of the intelligent learning-based cross-domain electronic fence collaborative early warning method described above, and will not be repeated here. Each module in the aforementioned intelligent learning-based cross-domain electronic fence collaborative early warning system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.
[0057] In one embodiment, an electronic device is provided, which may be a computer, and its internal structure diagram may be as follows: Figure 3 As shown, the electronic device includes a processor, memory, and network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for cross-domain electronic fence collaborative early warning data. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a cross-domain electronic fence collaborative early warning method.
[0058] Those skilled in the art will understand that, Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0059] In one embodiment of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described cross-domain electronic fence collaborative early warning method based on intelligent learning.
[0060] In one embodiment of this application, a computer program product is provided, including a computer program / instructions, which, when executed by a processor, implements the steps of the above-described cross-domain electronic fence collaborative early warning method based on intelligent learning.
[0061] The computer-readable storage medium and computer program product provided in this embodiment are similar in implementation principle and technical effect to the above method embodiments, and will not be described again here.
[0062] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.
[0063] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0064] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.
Claims
1. A cross-domain electronic fence collaborative early warning method based on intelligent learning, characterized in that, The method includes: Multi-source sensing data of the target is collected in real time at each domain edge node, and the multi-source sensing data is spatiotemporally aligned and feature extracted to obtain the target trajectory embedding vector in a unified coordinate system. The trajectory embedding vector is encrypted into a ciphertext trajectory packet using a lightweight homomorphic encryption algorithm and actively pushed to the downstream domain edge node along the target movement direction to achieve cross-domain identity relay. Downstream domain edge nodes perform similarity calculations on the encrypted trajectory packets in encrypted state to generate a boundary crossing risk score, and compare the boundary crossing risk score with a local dynamic threshold to obtain a preliminary warning result; Each domain edge node uploads the initial warning results to the federated learning coordination node. The coordination node trains the threshold update model based on global risk samples and sends the encrypted gradient back to each domain to complete dynamic threshold self-learning. When the risk score of any domain exceeds the updated dynamic threshold, a collaborative early warning command is triggered, which links the execution mechanism of this domain with the edge nodes of adjacent domains to complete audio-visual interception, path guidance or remote vehicle locking operations.
2. The method according to claim 1, characterized in that, The multi-source sensing data includes satellite positioning messages, radio frequency identification, video structured features, vehicle bus speed, and visibility from meteorological sensors. The spatiotemporal alignment is achieved through network-wide time synchronization and national standard coordinate transformation. Feature extraction uses a lightweight convolutional network to compress the original high-dimensional data into a fixed-dimensional trajectory embedding vector.
3. The method according to claim 1, characterized in that, The lightweight homomorphic encryption algorithm adopts a hierarchical homomorphic scheme, which encapsulates the trajectory embedding vector of multiple consecutive frames in a batch processing slot at one time. The encryption latency is lower than the real-time frame interval, and only one set of keys is needed to support concurrent decryption and verification of multiple nodes in the downstream domain.
4. The method according to claim 1, characterized in that, The ciphertext state similarity calculation uses an inner product approximation kernel function, and completes the trajectory cosine similarity evaluation in the ciphertext domain through polynomial Taylor expansion.
5. The method according to claim 1, characterized in that, The dynamic threshold self-learning step includes: After the coordinating node aggregates the encrypted gradients of each domain, it executes the federated averaging algorithm, adds noise using differential privacy, and trains the threshold update model. Each domain edge node downloads the encrypted model parameters and decrypts them in the local trusted execution environment, updating the local dynamic threshold. The update cycle is shorter than the event response cycle.
6. The method according to claim 1, characterized in that, The collaborative early warning command adopts the cascading publish-subscribe mechanism of the message queue telemetry transmission protocol, with the priority queue set to the highest level. The end-to-end transmission delay is lower than the early warning response time limit. After receiving the command, the edge node of the adjacent domain starts the preset linkage script to complete the synchronous execution of the audible and visual alarm, the gate lowering, and the upstream guidance screen prompt in the local domain.
7. The method according to claim 1, characterized in that, The method also includes storing immutable logs of the entire process on the blockchain for evidence preservation, specifically: The immutable log on-chain evidence storage adopts a consortium blockchain architecture, which writes the warning timestamp, target anonymity identifier, risk score, threshold version number and execution result into the block. The block writing latency is lower than the event recording time limit, and it supports regulators to conduct verifiable queries under key escrow through smart contracts.
8. A cross-domain electronic fence collaborative early warning system based on intelligent learning, characterized in that, The system includes: The acquisition module is used to acquire multi-source sensing data of the target in real time at each domain edge node, and to perform spatiotemporal alignment and feature extraction on the multi-source sensing data to obtain the target trajectory embedding vector in a unified coordinate system. The encryption module is used to encrypt the trajectory embedding vector into a ciphertext trajectory package using a lightweight homomorphic encryption algorithm, and actively push it to the downstream domain edge node along the target movement direction to realize cross-domain identity relay. The comparison module is used by downstream domain edge nodes to calculate the similarity of the encrypted trajectory packet in encrypted state, generate an out-of-bounds risk score, and compare the out-of-bounds risk score with the local dynamic threshold to obtain a preliminary warning result. The training module is used by the edge nodes of each domain to upload the initial warning results to the federated learning coordination node. The coordination node trains the threshold update model based on global risk samples and sends the encrypted gradient back to each domain to complete the dynamic threshold self-learning. The early warning module is used to trigger a collaborative early warning command when the risk score of any domain exceeds the updated dynamic threshold. This command links the execution mechanism of this domain with the edge nodes of adjacent domains to complete audio-visual interception, path guidance, or remote vehicle locking operations.
9. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, implements the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.