Intelligent anti-lost tracking system based on bluetooth beacon and behavior recognition

Abstract

The application belongs to the technical field of Internet of Things intelligent terminal and short-distance transmission, and specifically discloses an intelligent anti-lost tracking system based on Bluetooth beacon and behavior recognition. The system comprises an anti-lost label, a mobile terminal cluster, a cloud collaborative analysis platform, a behavior recognition engine and a privacy protection communication unit. The anti-lost label dynamically adjusts the Bluetooth broadcast parameters based on the bionic swarm intelligence algorithm, the mobile terminal passively listens and uploads the anonymous fingerprint data packet verified by zero-knowledge proof, the cloud platform aggregates data to build a trajectory heat map, and the behavior recognition engine fuses multi-modal sensor data to identify the state of the article and optimize the broadcast strategy. Through the above technical scheme, the application realizes continuous tracking, high-precision positioning and adaptive power management after disconnection under the premise of protecting the location privacy of users, and improves the collaboration, security and environmental adaptability of the anti-lost system.

Description

Technical Field

[0001] This invention belongs to the field of IoT smart terminal and short-range transmission technology, specifically relating to a smart anti-loss tracking system based on Bluetooth beacons and behavior recognition. Background Technology

[0002] With the popularization of IoT technology and the evolution of wireless communication protocols, smart anti-loss tracking devices based on Bluetooth Low Energy have become important tools in the fields of personal asset management and security. These systems establish a wireless communication link between mobile terminals and electronic tags to achieve distance monitoring and location alerts for high-value targets such as keys, wallets, and even pets. In the construction of smart cities and digital life, stable and reliable asset location services not only effectively reduce the risk of property loss but also provide fundamental support for refined behavior perception and indoor navigation applications.

[0003] Intelligent tracking systems that integrate behavior recognition and distributed sensing represent a key evolutionary direction for improving search efficiency. These systems utilize Bluetooth beacons to transmit broadcast signals and combine this with sensor data from mobile devices to perform real-time calculations of the target's motion status, aiming to build a dynamically responsive protection network. In practical applications, the system needs to cope with complex indoor environmental interference and long-distance signal attenuation, which places higher demands on the robustness of the positioning algorithm, its multi-source data fusion capabilities, and its trajectory compensation mechanism in offline states.

[0004] Traditional anti-loss systems rely excessively on point-to-point single Bluetooth connections. Once the target exceeds the physical communication threshold, the system enters a perception blind spot, making it difficult to maintain continuous trajectory tracking after connection loss. While existing collaborative search models attempt to introduce crowdsourced positioning, they often require third-party devices to expose precise geographical location information, leading to an irreconcilable conflict between search efficiency and user privacy protection. Traditional solutions lack dynamic adjustment capabilities when dealing with non-linear movement paths and cannot simulate complex collaborative search logic. This results in significant signal conflicts and rapid attenuation of positioning accuracy in multi-device interaction environments, making it difficult to meet the needs of accurate target tracing in complex and dynamic environments. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides an intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition, comprising: The anti-loss tag is configured to dynamically adjust the transmission parameters of the Bluetooth broadcast signal based on a biomimetic swarm intelligence algorithm after entering search mode, and broadcast an encrypted beacon signal containing identification and behavioral status information. A mobile terminal cluster consists of multiple mobile devices distributed within a target area. The mobile devices are configured to listen for broadcast signals from the anti-loss tag in the surrounding environment and activate a local privacy-preserving communication unit to preprocess the signal after a valid beacon is detected. The privacy-preserving communication unit is configured to verify the identity and legitimacy of the mobile device through a zero-knowledge proof protocol, and to encapsulate the received signal strength indication value and the encrypted geofence index information into an anonymous fingerprint data packet. A cloud-based collaborative analysis platform is configured to receive and aggregate the anonymous fingerprint data packets uploaded from the mobile terminal cluster, construct a heat map of the target item's movement trajectory based on a distributed perception model, and inversely deduce the current location of the target item. The behavior recognition engine is configured to combine the motion sensor data built into the anti-loss tag with historical movement patterns to classify and identify the current behavior state of the target item, and feed the recognition results back to the anti-loss tag to optimize the subsequent broadcast strategy.

[0007] Preferably, the anti-loss tag includes: The microcontroller unit, as the core logic control center, is used to interact with various peripheral modules through the internal bus; A low-power Bluetooth radio frequency module, connected to the microcontroller unit, is configured to have a long-distance transmission mode and extended broadcast capability, and is used to alternately transmit the encrypted beacon signal on a preset physical channel; The motion sensing components include a three-axis accelerometer, a three-axis gyroscope, and a geomagnetic sensor, which are used to collect the motion vector data of the target object in three-dimensional space in real time and transmit it to the microcontroller unit; The power management module is used to regulate the voltage and monitor the current of the built-in power supply, and supports switching between deep sleep mode, idle mode and high load search mode. A biomimetic algorithm storage unit is used to store the logic instructions of the biomimetic swarm intelligence algorithm. The microcontroller unit calls the logic instructions based on the data collected by the motion sensing component.

[0008] Preferably, the biomimetic swarm intelligence algorithm built into the anti-loss tag is configured as follows: Real-time acquisition of current motion characteristic parameters, including displacement velocity, turning frequency, and environmental vibration intensity; The preset pheromone concentration function is invoked to define the transmission frequency and transmission power of the broadcast signal as the virtual pheromone concentration in space; When the motion characteristic parameters indicate that the target object is in a moving state, the biomimetic swarm intelligence algorithm increases the pheromone update rate and instructs the low-power Bluetooth radio frequency module to increase the broadcast frequency and increase the transmission power. When the motion characteristic parameters indicate that the target object is stationary, the biomimetic swarm intelligence algorithm reduces the pheromone update rate, instructs the low-power Bluetooth radio frequency module to reduce the broadcast frequency and reduce the transmission power, simulating the evaporation process of pheromones in nature.

[0009] Preferably, each mobile device in the mobile terminal cluster is configured with: A background monitoring process is used to periodically scan for broadcast packets in the surrounding environment, and the ratio of the scanning window to the scanning interval of the background monitoring process is adaptively adjusted according to the remaining battery power of the mobile device. The pass-through module is used to directly send the signal data stream to the privacy-protected communication unit without performing local parsing when the encrypted beacon signal conforming to the preset protocol format is detected. The cache scheduling module is used to temporarily store the generated anonymous fingerprint data packets when the network connection is abnormal, and to perform batch uploads according to the first-in-first-out principle after the network is restored.

[0010] Preferably, the privacy-preserving communication unit includes: The zero-knowledge proof protocol submodule is used to generate a proof message containing random numbers, public parameters, and private credential mapping values ​​to prove to the cloud-based collaborative analysis platform that the mobile device has legitimate search qualifications without disclosing the mobile device's unique identifier. The geofence index conversion module is used to divide the physical space into multiple non-overlapping and equal-area symmetrical polygon logical blocks, obtain the fuzzy identifier of the current network environment, and convert the fuzzy identifier into the corresponding encrypted geofence index through a hash mapping table. The encrypted geofence index is used to characterize the logical block range where the target item is located. The signal strength preprocessing module is used to smooth the signal strength indication values ​​received multiple times in succession using a moving average filter, remove noise interference caused by human body obstruction or wall reflection, and extract the true signal strength fingerprint.

[0011] Preferably, the cloud-based collaborative analysis platform adopts a distributed microservice architecture, including: Access gateway service is used to maintain a long connection with the mobile terminal cluster and to process verification requests for zero-knowledge proof protocol. A data aggregation engine is used to deliver the verified anonymous fingerprint data packets to a distributed message queue and group the data according to the encrypted geofence index and a preset time window. The trajectory heatmap generation unit is used to establish a dynamic weight matrix on a virtual electronic map layer, locate the matrix coordinates according to the encrypted geofence index, and assign corresponding initial weights to the matrix coordinates and their neighborhood according to the magnitude of the signal strength indication value. The position inversion calculation module is used to analyze the displacement vector of the highlighted area in the weight matrix over time, and calculate the weighted geometric centroid of the highlighted area to determine the logical position of the target item.

[0012] Preferably, the trajectory heatmap generation unit introduces a time decay factor when updating the weight matrix, configured as follows: after each preset time step, the existing weights of all elements in the weight matrix are multiplied by a decay coefficient less than 1; so that the weights corresponding to recently generated strong signal data are higher than the weights of historical data, thereby ensuring that the movement trajectory heatmap can reflect the latest location dynamics of the target item in real time.

[0013] Preferably, the behavior recognition engine integrates a multimodal sensor fusion algorithm, configured as follows: The vibration frequency is analyzed by the acceleration data reported by the anti-loss tag, the rotation angle change is captured by the gyroscope data, and the directional drift is corrected by the geomagnetic sensor data. The fused sensor feature vectors are input into a preset deep learning classification model to identify whether the target item is in a stationary state, a walking state, a vehicle transport state, a fallen state, or an abnormal shaking state. When the vehicle is identified as being in transport mode, an instruction is issued requiring the anti-loss tag to be fixed at the preset maximum transmission power and to make the broadcast frequency deviate from the limitation of the movement step frequency.

[0014] Preferably, the anti-loss tag and the mobile terminal cluster support Bluetooth mesh networking communication, configured as follows: When the anti-loss tag is in an area covered by multiple mobile devices, the signal emitted by the anti-loss tag is forwarded through the nearest mobile device as a relay node, forming a multi-hop signal relay link; The system is also equipped with an enhanced search mode. When the cloud-based collaborative analysis platform receives a loss report request, it adds the corresponding identity identifier to the entire network search list. In response to the enhanced search mode, the mobile device located near the target item increases the priority of the background monitoring process and shortens the upload interval of the anonymous fingerprint data packet.

[0015] Preferably, the system further includes: An edge computing gateway, deployed at a pre-defined fixed infrastructure location, integrates an array antenna and a local analysis and processing unit to receive high-priority beacon signals forwarded by the mobile device; The edge computing gateway is configured to utilize multiple received signal strength data streams, combined with the original beam phase information obtained by the array antenna, to perform spatial triangulation locally, and upload the generated local position vector to the cloud-based collaborative analysis platform. The privacy-preserving communication unit processes the geofence index based on homomorphic encryption technology. Before calculating the index, it performs an XOR operation between the original geographic coordinates and a dynamic random mask, and then performs a hash mapping.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The intelligent anti-loss tracking system based on Bluetooth beacon and behavior recognition provided by this invention solves the problem of traditional anti-loss devices relying on point-to-point direct connection. By constructing a distributed sensing network driven by biomimetic swarm intelligence, it realizes the ability to continuously track the trajectory after the owner device loses connection. 2. The present invention introduces a zero-knowledge privacy protocol and a geofence indexing mechanism, which fully utilizes crowdsourced mobile terminal resources while completely avoiding the risk of leakage of assistants' location information, thus solving the problem between search efficiency and privacy protection. 3. By integrating behavior recognition and dynamic broadcast control, this invention enables the anti-loss tag to adaptively adjust its working mode according to the actual scenario, thereby improving the signal coverage robustness and positioning accuracy in complex environments. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention; Figure 2 This is a schematic diagram of the core principle framework of the anti-loss tag in this invention, which is based on a biomimetic swarm intelligence algorithm to achieve adaptive broadcasting. Figure 3 This is a flowchart illustrating the main stages of the privacy-preserving communication unit's construction of anonymous fingerprint data packets in this invention. Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow between the cloud-based collaborative analysis platform and the mobile terminal cluster in this invention; Figure 5 This is a flowchart illustrating the logical flow of the behavior recognition engine in this invention, which combines sensor data to perform state classification and recognition. Detailed Implementation

[0018] Example 1: Reference Figures 1 to 5A smart anti-loss tracking system based on Bluetooth beacons and behavior recognition, including anti-loss tags, mobile terminal clusters, cloud-based collaborative analysis platform, behavior recognition engine, and privacy-protected communication unit; The anti-loss tag is used to dynamically adjust the transmission frequency and transmission power of its Bluetooth broadcast signal according to the bionic swarm intelligence algorithm after entering the search mode, and continuously broadcast an encrypted beacon signal containing identity and behavioral status information. The mobile terminal cluster consists of multiple user mobile devices distributed within the target area. Each mobile device is configured to passively listen to broadcast signals from anti-loss tags in the surrounding environment. After detecting a valid beacon, it activates a local privacy-preserving communication unit to perform preliminary processing on the signal. The privacy-preserving communication unit is used to verify the identity and legitimacy of the mobile terminal through a zero-knowledge proof protocol without acquiring or uploading the precise geographical location, and to encapsulate the received signal strength indication value and the encrypted geofence index information into an anonymous fingerprint data packet. The cloud-based collaborative analysis platform is used to receive anonymous fingerprint data packets uploaded from multiple mobile terminals, aggregate and analyze the data based on a distributed perception model, construct a heat map of the target item's movement trajectory, and inversely deduce its possible current location. The behavior recognition engine is used to classify and identify the current behavior state of the target item by combining the motion sensor data built into the anti-loss tag with historical movement patterns, and feed the recognition results back to the anti-loss tag to optimize its subsequent broadcast strategy.

[0019] The anti-loss tag includes a microcontroller unit, a low-power Bluetooth radio frequency module, a motion sensing component, a power management module, and a bionic algorithm storage unit.

[0020] The microcontroller unit serves as the core logic control center and adopts a 32-bit processor architecture with low power consumption characteristics. It interacts with various peripheral modules at high speed through an internal bus.

[0021] The low-power Bluetooth RF module supports Bluetooth 5.0 and above standards, has a long-distance transmission mode and extended broadcast capability, and can alternately send encrypted beacons on three physical channels (37, 38, and 39) according to the instructions of the microcontroller unit.

[0022] The motion sensing component integrates a three-axis accelerometer, a three-axis gyroscope, and a geomagnetic sensor to collect real-time motion vector data of the target object in three-dimensional space.

[0023] The power management module is responsible for regulating the voltage and monitoring the current of the built-in lithium manganese battery, and supports smooth switching between deep sleep mode, idle mode and high load search mode to extend the standby life of the device.

[0024] The biomimetic swarm intelligence algorithm module embedded in the anti-loss tag is specifically configured to execute a pheromone diffusion and attenuation mechanism that simulates the foraging process of an ant colony. Upon entering search mode, the microcontroller first obtains current motion characteristic parameters from the motion sensing components, including displacement velocity, turning frequency, and environmental vibration intensity. The biomimetic algorithm storage unit has a preset pheromone concentration function, which defines the transmission frequency and power of the broadcast signal as the virtual pheromone concentration in space.

[0025] When motion characteristic parameters indicate that the target object is in a fast-moving state, the biomimetic swarm intelligence algorithm module automatically calculates a high pheromone update rate, instructs the low-power Bluetooth radio frequency module to increase the broadcast frequency to once every 100 milliseconds, and adjusts the transmission power to the maximum gain state, thereby leaving a strong signal trace on the movement trajectory; when the target object is detected to be in a relatively stationary state, the algorithm reduces the pheromone update rate, lowers the broadcast frequency to once every 2000 milliseconds, and switches the transmission power to a low-power mode to simulate the slow evaporation process of pheromones in nature.

[0026] This dynamic adjustment mechanism ensures that the signal forms a dynamic trace in space that can be sensed by multiple nodes, while minimizing energy consumption.

[0027] Each mobile device in the mobile terminal cluster maintains a continuous low-power Bluetooth listening process in the system background through a built-in application programming interface (API). This listening process is configured to periodically scan for broadcast packets in the surrounding environment, with the scanning window and interval adaptively adjusted based on the device's remaining battery power. When a mobile device detects an encrypted beacon signal conforming to a preset protocol format, it does not directly parse the beacon's content but instead directly transmits the signal data stream to its local privacy-preserving communication unit. The mobile device also has a cache scheduling module to temporarily store received anonymous fingerprint data packets when the network connection is unstable and uploads them in batches after the network recovers, using a first-in, first-out (FIFO) principle.

[0028] In its processing flow, the privacy-preserving communication unit first invokes the zero-knowledge proof protocol submodule. This submodule stores digitally signed participant qualification certificates. When communicating with the cloud-based collaborative analysis platform, it does not send the raw byte stream of the participant qualification certificates, but instead generates a proof message containing random numbers, public parameters, and private credential mapping values. This proof message proves to the cloud that "the mobile device is a legitimate search node authorized by the system," but the cloud cannot use this to trace the device's unique identifier or owner information. After successful verification, the privacy-preserving communication unit initiates the geofence index conversion logic.

[0029] The system pre-divides the physical space into several equal-area regular hexagonal logical blocks, each block corresponding to a unique encrypted index value in the database. The privacy-preserving communication unit obtains the fuzzy identifier of the current base station or wireless LAN, and converts it into the corresponding encrypted geofence index through a built-in hash mapping table. The geofence index only identifies which kilometer-level block the target is roughly located in, without including latitude and longitude coordinates.

[0030] The privacy-preserving communication unit also includes a signal strength preprocessing module, which filters the received signal strength indication value in real time. Considering the effects of multipath propagation and shadow fading, the preprocessing module uses a moving average filter to smooth the strength values ​​received five consecutive times, removing sudden noise caused by human occlusion or wall reflections, and extracting the true signal strength fingerprint. The privacy-preserving communication unit encapsulates the processed received signal strength indication value, encrypted geofence index information, and a high-precision timestamp to form the final anonymous fingerprint data packet, which is then sent to the cloud via an encrypted Secure Sockets Layer (SSL) channel.

[0031] The cloud-based collaborative analysis platform adopts a distributed microservice architecture, including an access gateway service, a data aggregation engine, a trajectory heatmap generation unit, and a location reverse calculation module. The access gateway service is responsible for maintaining long-lived connections with tens of thousands of mobile terminals and handling zero-knowledge proof verification requests. Upon successful verification, the anonymous fingerprint data packet is delivered to a distributed message queue. The data aggregation engine extracts data from the message queue and groups it according to geofence indexes and time windows. If multiple mobile terminals report anonymous fingerprints for the same anti-loss tag within the same time period, the data aggregation engine will activate a spatial collaborative filtering algorithm.

[0032] The trajectory heatmap generation unit is configured to establish a dynamic weight matrix on a virtual electronic map layer. Each element in the matrix corresponds to a geospatial unit. When an anonymous fingerprint data packet is received, the unit locates the matrix coordinates based on the encrypted geofence index and assigns corresponding initial weights to the coordinates and their neighborhood according to the magnitude of the reception strength indicator value. The greater the reception strength, the closer the anti-loss tag is to the observation point, and the higher the assigned weight. To simulate the dynamic characteristics of the ant colony algorithm, the generation unit introduces a time decay factor. As time passes, the weights of historical data decay exponentially.

[0033] For example, after each preset time step, the weights of all matrix elements are multiplied by a decay coefficient less than 1. In this way, only areas that have recently generated strong signals will appear as bright "hot spots" on the heatmap, while outdated location information will automatically darken or even disappear.

[0034] The location inversion calculation module utilizes the reconstructed trajectory heatmap, combined with the continuity constraint of target movement, to execute a centroid deduction algorithm based on probability flow. The module calculates the weighted geometric centroid of the highlighted areas in the heatmap as the most likely logical location of the target item. Simultaneously, the location inversion calculation module analyzes the displacement vector of the hotspot areas over time, predicts the target's possible movement trend within the next 10 minutes, and feeds this prediction result back to the owner's terminal as a search command.

[0035] The behavior recognition engine achieves accurate state classification by deeply fusing data from multiple sensors. The engine integrates a pre-trained lightweight deep learning classification model. Accelerometer data collected by the anti-loss tag is used to analyze vibration frequency and amplitude, gyroscope data is used to capture changes in rotation angle, and geomagnetic sensor data is used to correct directional drift. The behavior recognition engine is configured to recognize the following specific behavioral states: stationary state, walking carrying state, vehicle transportation state, falling state, and abnormally violent shaking state. When the recognition engine determines that the target is in a vehicle transportation state, characterized by continuous high-speed linear acceleration and low-frequency large-amplitude vibration, it sends a control command to the anti-loss tag via the cloud, instructing it to enter "vehicle tracking mode." In this mode, the broadcast power is forcibly set to the highest level, and the broadcast frequency is no longer affected by the movement step frequency to counteract the Doppler frequency shift and rapid signal switching caused by high-speed vehicle movement.

[0036] The behavior recognition engine also possesses self-learning capabilities, automatically optimizing recognition parameters based on environmental features in historical movement trajectories. For example, if historical trajectories show the target frequently entering and exiting elevators or basements, the engine will automatically increase the weight of barometric pressure sensor data (if the hardware includes a barometer), using step changes in air pressure to help determine whether the target is in a vertical displacement state. The recognition results are not only used for positioning assistance but also directly affect the power consumption control strategy of the anti-loss tag. If the target is identified as stationary for an extended period, such as when placed in a drawer, the system will automatically extend the broadcast interval to more than 30 seconds, maintaining only a minimum survival heartbeat.

[0037] The anti-loss tag also supports Bluetooth Mesh networking communication with the mobile terminal cluster. When the anti-loss tag is in an area with dense mobile devices, the signal emitted can be received by the nearest mobile terminal and forwarded as a relay node to other mobile terminals at a slightly more distance, forming a multi-hop signal relay network. In the Mesh network protocol stack, each node maintains a routing table to record the link quality of its neighbors. This architecture allows the anti-loss tag's signal to bypass physical obstacles and be transmitted through the mobile terminal nodes of the crowd, even inside large shopping malls or complex office buildings, thus extending the sensing range.

[0038] Furthermore, the system is configured with an enhanced search mode. When the owner clicks the "Report Lost" button in the application, the cloud-based collaborative analysis platform adds the tag's unique hash identifier to the "network-wide search blacklist." All mobile terminals connected to the system, upon scanning the unique hash identifier, prioritize their local monitoring processes and shorten the upload interval of anonymous fingerprint data packets. The cloud pushes a silent command to active mobile terminals near the lost location, requesting them to enable higher-precision reception strength sampling. If the system detects that the target item has been in a preset high-risk area for an extended period, such as a garbage transfer station or a busy secondhand trading area, or if the behavior recognition engine detects abnormal vibrations, the system will automatically trigger an alarm and force the anti-loss tag to drive its built-in miniature piezoelectric buzzer to emit a high-decibel distress sound.

[0039] The encryption process of the geofence index employs a spatial transformation technique based on homomorphic encryption. Before calculating the index, the privacy-preserving communication unit first performs an XOR operation between the geographic coordinates and a dynamically random mask sent from the cloud, and then performs a hash mapping. This means that even if a third party intercepts the encrypted index during transmission, it cannot deduce the actual physical area because it does not know the mask value that changes every minute. This mechanism ensures the "forward security" and "instantaneous validity" of geographic location information; that is, past geographic indexes cannot be used to infer future location changes.

[0040] In practical engineering implementation, the physical dimensions of the anti-loss tag are limited to within 30 mm x 30 mm, and the thickness does not exceed 5 mm, to facilitate attachment or hanging on various assets. The outer shell is made of waterproof and impact-resistant high-strength polycarbonate material, and the internal circuit board uses a four-layer board process to optimize radio frequency interference and signal integrity. The antenna uses a ceramic patch antenna with omnidirectional radiation characteristics, ensuring stable transmission gain under different placement postures.

[0041] When performing zero-knowledge proofs, the mobile devices in the mobile terminal cluster employ an improved non-interactive zero-knowledge proof protocol. The mobile device, acting as the prover, pre-generates a commitment value based on its private key and random system parameters. Upon receiving a beacon, the device uses this commitment value and the beacon fingerprint to perform a mixed operation, generating a 256-bit proof string. The cloud, acting as the verifier, only needs to use a publicly available verification algorithm and the corresponding public key to calculate the proof string. If the result matches, the reported data is confirmed to be authentic, reliable, and originating from an authorized device. The entire process does not require the mobile device to send any data that could expose its hardware ID (such as IMEI or MAC address), fully complying with the principle of minimizing data exposure.

[0042] The cloud-based collaborative analysis platform incorporates environmental terrain compensation logic when constructing trajectory heatmaps. The platform stores indoor and outdoor 3D modeling data of the target area. When the reverse calculation module processes the received intensity indication value, it automatically applies a specific attenuation compensation coefficient based on the known wall distribution and obstacle materials at the location of the anonymous fingerprint. For example, if the known target is located between two reinforced concrete columns, the system automatically increases the expected signal attenuation, thereby automatically correcting spatial geometric offsets when retrieving the centroid position, improving positioning accuracy in complex indoor environments.

[0043] Example 2: Based on the intelligent anti-loss tracking system based on Bluetooth beacon and behavior recognition described in Example 1, this example provides a variant of the system architecture based on edge computing node collaboration, which aims to further reduce the processing pressure on the cloud and improve the real-time response speed.

[0044] In Embodiment 2, the system further includes edge computing gateways distributed throughout urban infrastructure. These gateways are deployed in fixed locations such as lampposts, bus stops, shopping mall entrances, or convenience stores. The edge computing gateway integrates a high-performance Bluetooth array antenna and a local analytics processing unit.

[0045] When the anti-loss tag sends an encrypted beacon, its broadcast payload includes an additional priority flag. When the tag determines, through its built-in behavior recognition logic, that it is moving at high speed and far from a known safe area, it will set the priority flag to high.

[0046] Upon receiving a high-priority beacon, the mobile devices in the mobile terminal cluster, in addition to executing the data upload process described in Example 1, will also prioritize searching for available edge computing gateways in the vicinity. If an edge computing gateway is found, the mobile device will synchronize the encrypted beacon data to the gateway via a low-latency direct communication protocol, such as Wi-Fi sensing or dedicated short-range communication.

[0047] The edge computing gateway incorporates a local collaborative analysis engine. This engine receives beacon data forwarded by all mobile devices within its coverage radius. Unlike cloud platforms, edge computing gateways focus on local spatial triangulation. Because the gateway's location is fixed and known, it can utilize the received RSSI data streams from multiple mobile devices, combined with the raw beam phase information (AoA) received by its own Bluetooth array antenna, to perform high-precision local location calculations.

[0048] After completing the calculations locally, the edge computing gateway only uploads the generated local location vectors and the anonymized processing results to the cloud-based collaborative analysis platform. This architecture changes the original model of "all raw data going to the cloud," offloading a large amount of coordinate conversion and signal filtering work to the network edge. In densely populated and busy areas, this distributed processing mode can reduce positioning latency from seconds to milliseconds.

[0049] In Embodiment 2, the privacy-preserving communication unit is integrated into the hardware security module of the edge computing gateway. When the mobile terminal interacts with the gateway, it also uses a zero-knowledge proof protocol for authentication. When processing local location data, the gateway automatically performs spatial obfuscation. For example, when there are only a few mobile devices (assistants) in the area, to protect the location privacy of these assistants, the gateway will proactively reduce the location resolution until the number of participating devices reaches a preset anonymity threshold, such as at least five devices. This dynamic privacy protection mechanism based on the k-anonymity model ensures a dynamic balance between the accuracy of the search target and the privacy of the participants.

[0050] The behavior recognition engine was also enhanced in Embodiment 2. The edge computing gateway can provide additional environmental context to the behavior recognition engine based on local weather sensor and environmental noise sensor data. For example, if the gateway detects heavy rain, it will broadcast a command to all anti-loss tags within range to increase the signal transmission duty cycle to overcome the strong absorption of 2.4GHz signals by rainwater. This adaptive adjustment based on regional environmental characteristics further improves the reliability of the system in severe weather or special geographical environments.

[0051] Example 3: Based on the intelligent anti-loss tracking system based on Bluetooth beacon and behavior recognition described in Example 1 or 2, this example demonstrates a specific application configuration for industrial-grade valuable asset management and long-term monitoring.

[0052] In Example 3, the anti-loss tag is designed as an industrial-enhanced tag with energy harvesting capabilities. In addition to a built-in battery, a micro-photovoltaic panel and a piezoelectric energy harvesting device are integrated into the tag's outer shell. The piezoelectric energy harvesting device converts mechanical vibrations in industrial environments or bumps during asset transport into electrical energy, which is then stored in an internal supercapacitor.

[0053] The microcontroller unit is configured with a specific "energy-sensing scheduling logic." This logic monitors the voltage level of the supercapacitor in real time. When sufficient energy reserves are detected, the tag actively activates higher-order biomimetic swarm intelligence behavior, such as simulating a flock of birds' "navigator" mode. In this mode, the anti-loss tag not only broadcasts its own identity but also uses its built-in low-power flash memory to record a summary of its movement trajectory over the past 24 hours and embeds these summaries into a cyclic broadcast packet.

[0054] In this embodiment, the mobile terminal cluster includes not only smartphones but also vehicle-mounted intelligent terminals installed on forklifts, freight trucks, or automated guided vehicles. These terminals possess stronger computing power and a more continuous scanning frequency. When the vehicle-mounted terminal receives a broadcast packet containing trajectory summary information, its local behavior recognition engine immediately reconstructs the asset's movement chain over a past period.

[0055] The privacy-preserving communication unit has been customized for industrial scenarios. Since industrial assets may involve trade secrets, an attribute-based encryption access control strategy has been added to the anonymous fingerprint data packets. Only users holding specific role keys can decrypt the high-precision portions of the heatmap after secondary authentication through the cloud-based collaborative analysis platform. For ordinary crowdsourcing contributors, their reported data can only be used in the cloud to construct macroscopic, statistical location distributions.

[0056] The cloud-based collaborative analysis platform, in Example 3, adds a logical consistency verification module. This module analyzes historical asset movement patterns and establishes a predictive model based on a Markov chain. For example, if a valuable piece of equipment typically moves only between Factory 1 and Warehouse 2, the verification module will immediately identify it as "unexpected relocation" and trigger the highest-level alarm if its anonymous fingerprint appears in the office area or off-site area. This logically predictive search mechanism allows the system to proactively detect abnormal asset movements, rather than passively searching for lost items.

[0057] The behavior recognition engine adds the identification of "abnormal stillness" and "resonance patterns" when processing industrial asset data. For example, when a precision instrument is subjected to frequency resonance exceeding its tolerance threshold during transportation, the recognition engine will capture this feature through spectrum analysis and record the event in the asset's digital twin file as a basis for subsequent quality traceability.

[0058] Example 4: Based on the intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition described in the foregoing examples, this example details the robustness enhancement scheme of the system in dealing with extremely complex environments, such as large underground multi-story parking lots or computer rooms with dense metal structures.

[0059] In Example 4, the anti-loss tag employs a dual-band transmission architecture. In addition to the standard 2.4GHz Bluetooth Low Energy beacon, the system is also equipped with a backup long-range broadcast unit based on Sub-1GHz. The microcontroller dynamically determines whether to activate dual-band synchronous broadcasting based on the environmental interference assessment results fed back by the behavior recognition engine.

[0060] Because low-frequency signals (Sub-1GHz) have stronger diffraction capabilities and lower free-space loss, they can penetrate multiple layers of thick reinforced concrete slabs. When the anti-loss tag detects frequent Bluetooth connection interruptions and the motion sensor indicates that it is located deep underground, the system will automatically activate the long-range broadcast unit to send a low-speed pulse signal with a narrower bandwidth and stronger error correction code.

[0061] When a specific device in the mobile terminal cluster, such as an inspection terminal equipped with a corresponding frequency band receiving module, detects the low-speed pulse signal, its privacy-preserving communication unit employs a special long-range geometric fingerprinting algorithm. This long-range geometric fingerprinting algorithm does not rely on a single RSSI value, but instead uses the signal's time-of-flight (ToF) or angle-of-arrival (TDOA) for preliminary ranging.

[0062] The cloud-based collaborative analysis platform employs a hierarchical data fusion framework when processing this cross-frequency band mixed data. During the heatmap generation phase, the system assigns different weight distribution functions to data from different frequency bands. Sub-1GHz data, due to its larger positioning error but wider coverage, is used to generate a large-scale "probability cloud," while Bluetooth data is used for "fine-tuning" within the probability cloud.

[0063] In this scenario, the behavior recognition engine also incorporates geomagnetic fingerprinting assistance logic. Pre-collected geomagnetic distribution maps of the underground space are stored in a cloud database. Because different metal structures in different locations generate specific disturbances to the geomagnetic field, the recognition engine performs feature matching between the encrypted geomagnetic vectors reported by the anti-loss tag and the database, achieving meter-level assisted positioning correction even in environments where GPS signals are completely absent.

[0064] The privacy-preserving communication unit further enhances its resistance to correlation attacks when performing cross-frequency band data encapsulation. To prevent attackers from identifying targets by comparing Bluetooth beacons and Sub-1GHz beacons at the same time, the communication unit introduces random jitter bias on the transmission time axis of these two signals. This temporal decorrelation processing makes it difficult for unauthorized eavesdroppers to attribute signals from different frequency bands to the same anti-loss tag, even at the physical level, providing a deeper level of privacy protection in complex electromagnetic environments.

[0065] The system also features a self-healing perception network mechanism based on swarm intelligence. When the cloud platform detects low activity levels in a certain area, such as a blind spot in a shopping mall, it automatically sends task reward instructions to nearby active users, encouraging them to move into the blind spot to fill the perception gaps with their mobile devices. This deep integration of the incentive mechanism and the technical solution creates a dynamically evolving search ecosystem.

[0066] In all embodiments, the system strictly adheres to encryption standards for data storage and transmission. The firmware inside the anti-loss tag supports secure boot and OTA encrypted upgrades to prevent unauthorized firmware replacement. The cloud database uses fragmented encrypted storage, and the critical identity mapping table is stored in a hardware security module, ensuring that even in the event of a systemic intrusion in the cloud, the correspondence between the owner and the tag, as well as the location privacy of the assistant, remain physically isolated and protected.

[0067] The technical features in the above embodiments can be combined and applied according to actual needs. For example, industrial-grade tags can simultaneously possess edge computing collaboration and dual-band broadcasting functions to cope with the most demanding application environments. Through this modular and layered architecture design, the system achieves full-scenario coverage from personal consumer electronics to professional industrial asset management. While ensuring privacy, it solves the technical bottlenecks of traditional anti-loss systems, such as limited sensing range and loss upon disconnection, through the combination of biomimetic intelligence and crowdsourcing collaboration.

[0068] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition, characterized in that, include: The anti-loss tag is configured to dynamically adjust the transmission parameters of the Bluetooth broadcast signal based on a biomimetic swarm intelligence algorithm after entering search mode, and broadcast an encrypted beacon signal containing identification and behavioral status information. A mobile terminal cluster consists of multiple mobile devices distributed within a target area. The mobile devices are configured to listen for broadcast signals from the anti-loss tag in the surrounding environment and activate a local privacy-preserving communication unit to preprocess the signal after a valid beacon is detected. The privacy-preserving communication unit is configured to verify the identity and legitimacy of the mobile device through a zero-knowledge proof protocol, and to encapsulate the received signal strength indication value and the encrypted geofence index information into an anonymous fingerprint data packet. A cloud-based collaborative analysis platform is configured to receive and aggregate the anonymous fingerprint data packets uploaded from the mobile terminal cluster, construct a heat map of the target item's movement trajectory based on a distributed perception model, and inversely deduce the current location of the target item. A behavior recognition engine is configured to combine the motion sensor data built into the anti-loss tag with historical movement patterns to classify and identify the current behavior state of the target item, and feed the recognition results back to the anti-loss tag to optimize subsequent broadcast strategies. The built-in biomimetic swarm intelligence algorithm of the anti-loss tag is configured as follows: Real-time acquisition of current motion characteristic parameters, including displacement velocity, turning frequency, and environmental vibration intensity; The preset pheromone concentration function is invoked to define the transmission frequency and transmission power of the broadcast signal as the virtual pheromone concentration in space; When the motion characteristic parameters indicate that the target object is in a moving state, the biomimetic swarm intelligence algorithm increases the pheromone update rate and instructs the low-power Bluetooth radio frequency module to increase the broadcast frequency and increase the transmission power. When the motion characteristic parameters indicate that the target object is stationary, the biomimetic swarm intelligence algorithm reduces the pheromone update rate, instructs the low-power Bluetooth radio frequency module to reduce the broadcast frequency and reduce the transmission power, simulating the evaporation process of pheromones in nature.

2. The intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition according to claim 1, characterized in that, The anti-loss tag includes: The microcontroller unit, as the core logic control center, is used to interact with various peripheral modules through the internal bus; A low-power Bluetooth radio frequency module, connected to the microcontroller unit, is configured to have a long-distance transmission mode and extended broadcast capability, and is used to alternately transmit the encrypted beacon signal on a preset physical channel; The motion sensing components include a three-axis accelerometer, a three-axis gyroscope, and a geomagnetic sensor, which are used to collect the motion vector data of the target object in three-dimensional space in real time and transmit it to the microcontroller unit; The power management module is used to regulate the voltage and monitor the current of the built-in power supply, and supports switching between deep sleep mode, idle mode and high load search mode. A biomimetic algorithm storage unit is used to store the logic instructions of the biomimetic swarm intelligence algorithm. The microcontroller unit calls the logic instructions based on the data collected by the motion sensing component.

3. The intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition according to claim 2, characterized in that, Each of the mobile devices in the mobile terminal cluster is configured with: A background monitoring process is used to periodically scan for broadcast packets in the surrounding environment, and the ratio of the scanning window to the scanning interval of the background monitoring process is adaptively adjusted according to the remaining battery power of the mobile device. The pass-through module is used to directly send the signal data stream to the privacy-protected communication unit without performing local parsing when the encrypted beacon signal conforming to the preset protocol format is detected. The cache scheduling module is used to temporarily store the generated anonymous fingerprint data packets when the network connection is abnormal, and to perform batch uploads according to the first-in-first-out principle after the network is restored.

4. The intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition according to claim 3, characterized in that, The privacy-protected communication unit includes: The zero-knowledge proof protocol submodule is used to generate a proof message containing random numbers, public parameters, and private credential mapping values ​​to prove to the cloud-based collaborative analysis platform that the mobile device has legitimate search qualifications without disclosing the mobile device's unique identifier. The geofence index conversion module is used to divide the physical space into multiple non-overlapping and equal-area symmetrical polygon logical blocks, obtain the fuzzy identifier of the current network environment, and convert the fuzzy identifier into the corresponding encrypted geofence index through a hash mapping table. The encrypted geofence index is used to characterize the logical block range where the target item is located. The signal strength preprocessing module is used to smooth the signal strength indication values ​​received multiple times in succession using a moving average filter, remove noise interference caused by human body obstruction or wall reflection, and extract the true signal strength fingerprint.

5. The intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition according to claim 4, characterized in that, The cloud-based collaborative analysis platform adopts a distributed microservice architecture, including: Access gateway service is used to maintain a long connection with the mobile terminal cluster and to process verification requests for zero-knowledge proof protocol. A data aggregation engine is used to deliver the verified anonymous fingerprint data packets to a distributed message queue and group the data according to the encrypted geofence index and a preset time window. The trajectory heatmap generation unit is used to establish a dynamic weight matrix on a virtual electronic map layer, locate the matrix coordinates according to the encrypted geofence index, and assign corresponding initial weights to the matrix coordinates and their neighborhood according to the magnitude of the signal strength indication value. The position inversion calculation module is used to analyze the displacement vector of the highlighted area in the weight matrix over time, and calculate the weighted geometric centroid of the highlighted area to determine the logical position of the target item.

6. The intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition according to claim 5, characterized in that, The trajectory heatmap generation unit introduces a time decay factor when updating the weight matrix. It is configured such that after each preset time step, the existing weights of all elements in the weight matrix are multiplied by a decay coefficient less than 1. This ensures that the weights of recently generated strong signal data are higher than the weights of historical data, thereby ensuring that the movement trajectory heatmap can reflect the latest location dynamics of the target item in real time.

7. The intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition according to claim 6, characterized in that, The behavior recognition engine integrates a multimodal sensor fusion algorithm, configured as follows: The vibration frequency is analyzed by the acceleration data reported by the anti-loss tag, the rotation angle change is captured by the gyroscope data, and the directional drift is corrected by the geomagnetic sensor data. The fused sensor feature vectors are input into a preset deep learning classification model to identify whether the target item is in a stationary state, a walking state, a vehicle transport state, a fallen state, or an abnormal shaking state. When the vehicle is identified as being in transport mode, an instruction is issued requiring the anti-loss tag to be fixed at the preset maximum transmission power and to make the broadcast frequency deviate from the limitation of the movement step frequency.

8. The intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition according to claim 7, characterized in that, The anti-loss tag supports Bluetooth mesh networking communication with the mobile terminal cluster, configured as follows: When the anti-loss tag is in an area covered by multiple mobile devices, the signal emitted by the anti-loss tag is forwarded through the nearest mobile device as a relay node, forming a multi-hop signal relay link; The system is also equipped with an enhanced search mode. When the cloud-based collaborative analysis platform receives a loss report request, it adds the corresponding identity identifier to the entire network search list. In response to the enhanced search mode, the mobile device located near the target item increases the priority of the background monitoring process and shortens the upload interval of the anonymous fingerprint data packet.

9. The intelligent anti-loss tracking system based on Bluetooth beacons and behavior recognition according to claim 8, characterized in that, The system also includes: An edge computing gateway, deployed at a pre-defined fixed infrastructure location, integrates an array antenna and a local analysis and processing unit to receive high-priority beacon signals forwarded by the mobile device; The edge computing gateway is configured to utilize multiple received signal strength data streams, combined with the original beam phase information obtained by the array antenna, to perform spatial triangulation locally, and upload the generated local position vector to the cloud-based collaborative analysis platform. The privacy-preserving communication unit processes the geofence index based on homomorphic encryption technology. Before calculating the index, it performs an XOR operation between the original geographic coordinates and a dynamic random mask, and then performs a hash mapping.

Images