A two-wheeled vehicle theft prevention tracking method and system based on fusion positioning

Abstract

The application discloses a two-wheeled vehicle anti-theft tracking method and system based on fusion positioning, and relates to the field of radio navigation technology. The application distinguishes the coverage interaction conditions by fragmenting and counting historical Bluetooth broadcast interaction event sample sets, constructs an intersection coverage degree expected broadcast data number mapping relationship, quantifies the data receiving capacity under different radio frequency intersection conditions, gets rid of the mode of relying on experience to set the number of message outputs, and solves the problem that the number of observation samples is uncontrollable. Meanwhile, a broadcast database corresponding to the working condition is generated, the message is issued in combination with the message breakpoint connection, the broadcast output scale is dynamically adjusted according to the actual channel condition, and the message transmission integrity in the edge scene is improved. Road normal side deviation analysis is carried out by means of a sliding analysis window, a position drift baseline is constructed, the position data reporting items are adaptively adjusted, the positioning error caused by insufficient samples and the resource loss caused by invalid reporting are avoided, and controllable quality input data is provided for cloud fusion calculation.

Description

Technical Field

[0001] This invention relates to the field of radio navigation technology, specifically to a two-wheeled vehicle anti-theft tracking method and system based on fusion positioning. Background Technology

[0002] Current anti-theft tracking methods for two-wheeled vehicles mostly employ a combination of GNSS satellite positioning and cellular base station positioning to obtain vehicle location. In open environments, GNSS outputs high-precision location results, while in scenarios where GNSS signals are lacking, such as underground garages or buildings with obstructions, cellular base station positioning outputs coarse-grained location. To compensate for the insufficient accuracy of cellular base station positioning, existing technologies borrow from the Bluetooth crowdsourced positioning approach, adding a Bluetooth beacon module to the vehicle. Surrounding third-party auxiliary devices scan and capture the vehicle's Bluetooth broadcast messages and then transmit the sniffing data back to the cloud. The cloud integrates the vehicle-reported location data and the third-party sniffing data, thereby improving the positioning effect of two-wheeled vehicles in obstructed environments and achieving all-weather tracking through the complementary relationship between active positioning capabilities and passive Bluetooth crowdsourced positioning.

[0003] However, two-wheeled vehicles are located at the edge of Bluetooth broadcast coverage, especially when the vehicle and auxiliary equipment are traveling in opposite directions. The radio frequency intersection time between the two is extremely short, resulting in numerous interactions with insufficient coverage. Channel multipath and fading further exacerbate random message loss, leading to poor integrity of the sniffing observation data returned by the auxiliary equipment. This, in turn, results in an unstable number of effective observation samples available for fusion calculation in the cloud, directly reducing the accuracy of crowdsourced fusion positioning and increasing the difficulty of anti-theft tracking. Therefore, there is an urgent need for a two-wheeled vehicle anti-theft tracking method and system based on fusion positioning. Summary of the Invention

[0004] The purpose of this invention is to provide a two-wheeled vehicle anti-theft tracking method and system based on fusion positioning to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a two-wheeled vehicle anti-theft tracking method based on fusion positioning, the two-wheeled vehicle anti-theft tracking method comprising the following steps: Step S1: Obtain a sample set of historical Bluetooth broadcast interaction events, establish a mapping relationship between the expected number of broadcast data items for the degree of intersection coverage, and obtain the expected data volume corresponding to different intersection states; Select any tracked two-wheeled vehicle as the research object, and obtain a historical Bluetooth broadcast interaction event sample set formed by the actual interactions between the tracked two-wheeled vehicle and multiple auxiliary devices during past business operations; the historical Bluetooth broadcast interaction event sample set consists of multiple completed Bluetooth interaction event records, each interaction event record corresponds to a real BLE broadcast interaction process between the tracked two-wheeled vehicle and an auxiliary device, the auxiliary device refers to a third-party smart terminal device equipped with a compatible SDK that can scan and capture the BLE broadcast messages sent by the tracked two-wheeled vehicle, including but not limited to smartphones, smart tablets and portable smart IoT terminals with Bluetooth background scanning capabilities; Each interaction event record contains the interaction information corresponding to the interaction process. The interaction information includes the effective intersection degree of the Bluetooth radio frequency coverage between the auxiliary device and the tracked two-wheeled vehicle, the effective interaction dwell time of the Bluetooth broadcast, the interval of the vehicle Bluetooth broadcast, and the number of observation data entries actually captured during the interaction process. Among them, each observation data includes a set of auxiliary device location data, RSSI ranging data, and message capture timestamp. The interaction conditions are divided into sufficient coverage interaction conditions and insufficient coverage interaction conditions based on the actual degree of intersection between the Bluetooth radio frequency coverage area between the auxiliary device and the tracked two-wheeled vehicle. The aforementioned fully covered interactive condition indicates that the Bluetooth radio frequency coverage areas of the auxiliary device and the tracked two-wheeled vehicle overlap sufficiently, and the signal interaction can maintain a long dwell time, ensuring that the tracked two-wheeled vehicle can complete the broadcast interaction of a preset number of message entries. The insufficient coverage interaction condition indicates that the auxiliary device and the Bluetooth radio frequency coverage area of ​​the tracked two-wheeled vehicle only form a small-scale instantaneous intersection, and the interaction can maintain a short dwell time; multipath and channel fading will affect both types of interaction conditions, resulting in random message loss. The historical Bluetooth broadcast interaction event sample set is segmented according to the preset statistical time slice, and all Bluetooth interaction event records generated by the tracked two-wheeled vehicle and multiple auxiliary devices within the same preset statistical time period are classified into the same overall interaction set. When a set of interactions contains both interaction event records that fully cover interaction conditions and interaction event records that do not fully cover interaction conditions, the statistical results corresponding to the fully covered interaction conditions in the set are used to supplement the data of the interaction event records corresponding to the insufficiently covered interaction conditions. When all interaction event records within a certain overall interaction set belong to the insufficiently covered interaction conditions, the original statistical results corresponding to the insufficiently covered interaction conditions are directly adopted, and no data completion operation is performed. For each overall interaction set obtained after segmentation, the Bluetooth interaction event records under the two types of interaction conditions are counted. Combined with the vehicle BLE broadcast cycle and the background scanning duty cycle of the auxiliary equipment, the expected number of captured observation data points corresponding to the two types of interaction conditions are obtained respectively. For the fully covered interactive working condition, the theoretical number of messages that can be received at the radio frequency link layer is obtained by calculating the ratio between the interaction dwell time and the vehicle Bluetooth beacon broadcast period and rounding it down. The theoretical number of messages that can be received at the radio frequency link layer is compared with the maximum number of messages that can be reported by the auxiliary equipment in a single interaction service, and the smaller value is selected as the expected number of interaction observation data points corresponding to the fully covered interactive working condition. For cases where all interaction event records within the overall interaction set belong to the insufficient coverage interaction condition, based on all historical interaction records belonging to the insufficient coverage interaction condition in the historical Bluetooth broadcast interaction event sample set, the interaction dwell time corresponding to each historical interaction record is extracted and the ratio of the vehicle Bluetooth beacon broadcast period is calculated and rounded down. The result is then multiplied by the channel loss probability coefficient obtained from the historical Bluetooth broadcast interaction event sample set to obtain the theoretical number of packets that can be received at the RF link layer after channel disturbance correction for each historical interaction record. Statistical processing is then performed on the theoretical number of packets that can be received at the RF link layer after channel disturbance correction for all historical interaction records to obtain the expected number of interaction observation data points corresponding to the insufficient coverage interaction condition. A mapping relationship for the expected number of broadcast data entries with overlapping coverage is constructed. This mapping relationship is generated offline based on a historical Bluetooth broadcast interaction event sample set after time slicing and on-demand data completion processing. During real-time service operation, real-time interaction data is synchronously divided according to the preset statistical time slices, and all auxiliary device interactions within the same time period are grouped into a single real-time overall interaction set. During the real-time Bluetooth interaction startup phase, the expected number of interaction observation data entries corresponding to the fully covered interaction condition is used as the criterion by default. After a single Bluetooth interaction is completed, the corresponding interaction condition is recorded and marked according to the actual observation data entries produced in this interaction. A single observation data entry includes auxiliary device location data, RSSI ranging data, and a message capture timestamp. The degree of Bluetooth radio frequency intersection varies greatly in different scenarios. In edge coverage conditions, the interaction dwell time is short and the probability of message loss is high. Directly adopting a fixed message output strategy will result in an uncontrollable number of actual captureable observation samples. By dividing the historical Bluetooth broadcast interaction event sample set into segments according to preset statistical time slices, we can distinguish between interaction conditions with sufficient coverage and those with insufficient coverage. We can perform data supplementation processing on the interaction records as needed, and statistically obtain the expected number of captured observation data under different conditions and construct a mapping relationship between the expected number of broadcast data under the degree of intersection coverage. Step S2: Based on the transmission features extracted from the historical Bluetooth broadcast interaction event sample set, generate a broadcast database corresponding to the working conditions, generate interaction result data packets based on the broadcast database, and upload them to the cloud. Based on the historical Bluetooth broadcast interaction event sample set, extract the data item characteristics of the actual transmission completed by each historical interaction event record belonging to the fully covered interaction working condition, and statistically obtain the maximum amount of data that can be transmitted on average under the fully covered interaction working condition, which is recorded as the full transmission threshold. The tracked two-wheeled vehicle locally persists the overall interaction set record corresponding to the previous preset statistical time slice. The overall interaction set record corresponding to the previous preset statistical time slice at least saves the device identifier, the comprehensive interaction condition mark under that time slice, and the position of the message breakpoint where the transmission was not completed. The tracked two-wheeled vehicle does not locally persist the complete trajectory data, but only temporarily maintains the message slices to be sent in the broadcast database. All location-related observation data are transmitted back to the cloud via the auxiliary device that establishes the Bluetooth interaction. The tracked two-wheeled vehicle locally creates a broadcast database. The broadcast database is a temporary scheduling database and is only used to store the broadcast message slices to be sent out in this interaction. The message entry capacity of the broadcast database is limited by the number of expected observation data entries for the corresponding working condition. Once the auxiliary device establishes a Bluetooth interaction link with the tracked two-wheeled vehicle within the same preset statistical time period, the tracked two-wheeled vehicle reads its own device identifier and the comprehensive interaction condition marker carried by the overall interaction set record corresponding to the previous preset statistical time slice stored locally. Based on the read comprehensive interaction condition marker, it determines the broadcast database to be generated for this interaction. When the previous interaction event record is marked as a fully covered interaction condition, the number of expected interaction observation data entries corresponding to the fully covered interaction condition is called, and the number of expected interaction observation data entries is selected as the message entry capacity standard of the broadcast database to generate the broadcast database corresponding to the fully covered interaction condition. When the previous interaction event record is marked as an insufficient coverage interaction condition, the expected number of interaction observation data entries corresponding to the sufficient coverage interaction condition and the expected number of interaction observation data entries corresponding to the insufficient coverage interaction condition are called. Combined with the uncompleted transmission message breakpoint positions recorded in the previous interaction event record, a broadcast database corresponding to the insufficient coverage interaction condition is generated. The broadcast database selects the full transmission threshold number of messages as the message entry capacity standard for this interaction, and loads the message slice to be transmitted from the message breakpoint position. After the broadcast database is generated, the tracked two-wheeled vehicle transmits the message entries stored in the broadcast database to the auxiliary device via BLE broadcast, prioritizing the transmission of the latest generated message entries first, followed by the transmission of residual message entries from the previous interaction that were not completed. After all Bluetooth interactions corresponding to the current preset statistical time period have ended, the tracked two-wheeled vehicle counts the data entries that were actually successfully broadcast under the overall interaction set of that time period. Based on the data entries that were actually successfully broadcast under the overall interaction set of that time period, the vehicle determines the comprehensive interaction condition corresponding to the preset statistical time slice, writes the corresponding comprehensive interaction condition flag into the overall interaction set record corresponding to the preset statistical time slice, and updates the position of the message breakpoints that were not completed in the overall interaction set record. The tracked two-wheeled vehicle extracts the device identifier of the tracked two-wheeled vehicle corresponding to this interaction, the identification information of the auxiliary device, the number of data entries actually broadcast in this interaction, the interaction status marker of this interaction, the upper limit of the message entry capacity of the broadcast database used in this interaction, and the updated message breakpoint position. All of the above information is encapsulated into an interaction result data packet. The transmission link for the auxiliary device to report the interaction result data packet can rely on the BLE Mesh protocol to realize multi-hop self-organizing network relay transmission. At the same time, the Boneh Boyen Shacham group signature scheme is used to sign the message. The aggregation node can verify that the message comes from a legitimate device cluster, but cannot identify the auxiliary device that sends the message. The interaction result data packet after the signature protection is completed is then transmitted to the cloud. The interaction result data packet is merged into the historical Bluetooth broadcast interaction event sample set maintained on the cloud side. Existing vehicle-mounted terminals lack dynamic scheduling mechanisms for broadcast messages based on actual interaction conditions, making them unable to adapt to data transmission requirements under different radio frequency intersection conditions. Furthermore, the reporting links of third-party auxiliary devices pose a risk of leaking device identity information. Based on the transmission characteristics corresponding to the mapping relationship and combined with locally stored message breakpoint positions, a broadcast database adapted to the current interaction condition is generated. Messages are then sent according to predetermined priorities, and the comprehensive interaction condition flag and message breakpoint positions are updated after the interaction ends. Step S3: The cloud performs road normal lateral deviation analysis on the uploaded interactive result data packet based on the sliding analysis window, constructs the position drift baseline, and adaptively determines the number of position data reporting entries; A sliding analysis window is configured on the cloud side. The sliding analysis window extracts data according to the time dimension of the upload of the interaction result data packet and updates the window content according to the preset sliding step size. The sliding analysis window extracts multiple historical interaction result data packets that have been received and stored on the cloud side. Each interaction result data packet carries multiple sets of location data. Each set of location data is continuously collected according to a fixed collection time interval, and each set of location data is bound to a corresponding collection timestamp. The cloud reads all interactive result data packets captured by the sliding analysis window, parses and extracts all location data and the device identifier of the tracked two-wheeled vehicle carried in the interactive result data packets; for each set of location data belonging to the same tracked two-wheeled vehicle, it reads the road vector data pre-stored in the geographic information database. The road vector data at least saves the coordinate sequence of the road centerline and the coordinate sequence of the two sides of the road. Project the planar coordinates of a single set of location data onto the plane containing the road vector data, and calculate the normal relative offset of the location data towards the corresponding road edge. The normal is the vector direction perpendicular to the local extension direction of the road. The longitudinal displacement component generated by the road extension direction does not participate in the calculation of the normal relative offset. The positive and negative values ​​of the normal relative offset distinguish the different sides of the location data from the road, and the absolute value of the normal relative offset represents the actual distance of the location data from the road edge. The normal relative offsets corresponding to all position data belonging to the same tracked two-wheeled vehicle within the sliding analysis window are traversed. Statistical operations are performed on all normal relative offsets to generate the position drift baseline corresponding to the tracked two-wheeled vehicle. The position drift baseline records the statistical distribution interval corresponding to the normal relative offsets, and this statistical distribution interval is used as the regular offset determination interval. The maximum value of the absolute value of all normal relative offsets within the sliding analysis window is extracted from the cloud and defined as the historical maximum drift baseline. For each set of position data within the sliding analysis window, the absolute value of the relative normal offset is compared with the boundary of the conventional offset judgment interval corresponding to the position drift baseline. When the absolute value of the normal relative offset falls within the boundary of the conventional offset determination interval, determine the number of position data entries for the configuration reference of the interaction result data packet. When the absolute value of the normal relative offset crosses the boundary of the regular offset determination interval, a mapping operation is performed between the current absolute value of the normal relative offset and the historical maximum drift baseline. The mapping operation outputs the value of the target position data entry. The input of the mapping operation is the current absolute value of the normal relative offset, the historical maximum drift baseline, and the number of reference position data entries. The mapping operation outputs the value of the position data entries that the interactive result data packet needs to carry. After the cloud completes the mapping calculation, it generates the target location data entry configuration parameters corresponding to the tracked two-wheeled vehicle; the cloud sends the target location data entry configuration parameters to each base station; after the tracked two-wheeled vehicle establishes a communication connection with the base station, the tracked two-wheeled vehicle reads the target location data entry configuration parameters forwarded by the base station and replaces the original location data entry configuration parameters stored locally on the tracked two-wheeled vehicle. The tracked two-wheeled vehicle extracts the device identifier, auxiliary device identification information set, number of data entries actually broadcast in the overall interaction set corresponding to the current preset statistical time slice, comprehensive interaction condition marker, message entry capacity limit of the broadcast database used in this interaction, and updated message breakpoint position from the overall interaction set corresponding to the current preset statistical time slice. All of the above information is encapsulated into an interaction result data packet. The interaction result data packet is transmitted to the cloud by the auxiliary device that captures the Bluetooth broadcast. Newly added interaction result data packets participate in the subsequent interception and statistical calculation of the sliding analysis window. Two-wheeled vehicles are prone to road normal drift during operation. A fixed number of reported position data entries cannot meet the data demands arising from variations in drift. Insufficient reported samples amplify positioning errors when drift is significant, while excessive reporting wastes transmission resources when drift is minor. This paper addresses this issue by using a sliding analysis window to continuously analyze the road normal lateral offset of uploaded interactive result data packets. Based on road vector data, the paper calculates the normal relative offset of position data, statistically generates position drift baselines and historical maximum drift baselines, and combines the offset with the judgment interval to obtain target position data entry configuration parameters through mapping operations. These parameters are then sent to the base station for the vehicle terminal to read and replace local configurations. The system can adaptively adjust the number of reported position data entries according to the actual road drift degree of the vehicle. It increases reported samples when drift is abnormal and maintains the baseline reporting scale when drift is within the normal range, thus adapting the reported data volume to the actual movement and offset state of the vehicle. Step S4: Based on the determined number of reported entries for fixed-location data, configure parameters to complete the position entry constraints of the interaction result data packet, and perform spatiotemporal alignment preprocessing on the observation data of the multi-auxiliary equipment corresponding to the overall interaction set under each preset statistical time slice; Based on the target location data entry configuration parameters, during the generation of interaction result data packets, the tracked two-wheeled vehicle selects location data according to the number of entries limited by the configuration parameters and completes the location entry constraint processing. The interaction result data packets corresponding to each preset statistical time slice are continuously uploaded to the cloud. The interaction result data packets carry sniffing observation data reported by multiple auxiliary devices within the overall interaction set. Each observation data includes auxiliary device location data, RSSI ranging data, and message capture timestamp. The cloud retrieves the pre-stored Bluetooth broadcast effective detection radius parameters and uses the historical target location estimation results of the tracked two-wheeled vehicle that have been calculated in the previous business process to calculate the geometric distance between the auxiliary device location and the historical target location estimation results in each observation data. Observation data whose geometric distance exceeds the Bluetooth broadcast effective detection radius parameters are no longer involved in coordinate transformation processing. From the auxiliary equipment corresponding to the retained observation data, one auxiliary equipment whose positioning quality meets the preset screening criteria is selected as the reference equipment for spatiotemporal alignment. The latitude, longitude, and altitude position information of the reference equipment in the geodetic coordinate system is obtained and converted to the geocentric-fixed coordinate system. Using the geocentric-fixed coordinate system coordinates of the reference equipment as the reference origin, the latitude, longitude, and altitude position information of the other auxiliary equipment corresponding to the retained observations are uniformly converted to the northeast-sky local coordinate system. The calculation formula for the northeast-sky local coordinate system is as follows: ; In the formula, , and Let them be the east, north, and celestial coordinates of the i-th auxiliary device in the northeast-sky coordinate system with the k-th reference auxiliary device as the origin; The longitude is represented as that of the reference device k; Represented as the latitude of the reference device k; Represented as the geocentric coordinate system coordinates of the i-th auxiliary device; Represented as the geocentric coordinate system coordinates of reference device k; When the spatial distance between the auxiliary device and the reference device involved in the conversion exceeds the preset spatial span threshold, the local reference origin of the corresponding auxiliary device is switched to perform coordinate conversion, thereby completing the coordinate benchmark unification of all auxiliary device position data; The cloud identifies the reporting time differences between the capture timestamps of each observation data message, retrieves the historical motion state estimation results of the tracked two-wheeled vehicle to obtain the motion speed parameters, and combines the historical target position estimation results of the tracked two-wheeled vehicle with the position solution of the auxiliary equipment to obtain the unit direction vector of the auxiliary equipment pointing to the tracked two-wheeled vehicle; selects a pre-set unified reference time as the alignment target time, and based on the motion speed parameters of the tracked two-wheeled vehicle, the unit direction vector, and the original message capture timestamps, performs time dimension interpolation calculation on the RSSI ranging data of each observation data, converts all observation data to the unified reference time, and completes the time alignment of the observation data; after coordinate benchmark unification and time alignment processing, a complete set of aligned observation data is obtained; Step S5: Using the spatiotemporally aligned observation data, perform outlier removal, multi-dimensional confidence contribution factor calculation, and target position fusion calculation of the tracked two-wheeled vehicle; The aligned observation data is used to continue the calculation. Each observation data corresponds to a set of auxiliary equipment positions, distance values ​​converted from RSSI, auxiliary equipment GPS positioning covariance parameters, and distance error variance parameters. Combined with the preliminary position estimation results of the tracked two-wheeled vehicle, the distance residual corresponding to each observation data is calculated. The distance residual is the absolute value of the difference between the observed distance value and the geometric distance between the auxiliary equipment and the tracked two-wheeled vehicle calculated from the preliminary position estimation results. The Mahalanobis distance test value is calculated by combining the distance measurement error variance parameter of a single observation data and the GPS positioning covariance parameter of the auxiliary equipment. The Mahalanobis distance test judgment threshold is retrieved from the cloud for numerical comparison. Observation data with Mahalanobis distance test values ​​greater than the judgment threshold are marked as outliers and removed, while observation data with Mahalanobis distance test values ​​less than or equal to the judgment threshold are retained. For all the remaining observation data after removing outliers, the GPS confidence contribution factor, the distance measurement confidence contribution factor, and the observation geometric contribution factor were calculated respectively. The formula for calculating the GPS confidence contribution factor is as follows: ; In the formula, Let the GPS confidence contribution factor be represented as the one corresponding to the i-th observation; Let represent the variance of the GPS position error of the i-th auxiliary device; This is represented as the preset reference positioning variance parameter; The formula for calculating the ranging confidence contribution factor is as follows: ; In the formula, This represents the ranging confidence contribution factor corresponding to the i-th observation; represents the variance of the ranging error corresponding to the i-th auxiliary device; represents the preset reference ranging variance parameter. The formula for calculating the observation geometric contribution factor is as follows: Let the observation direction be a unit vector. ; Observation geometry matrix: ; ; In the formula, Let represent the observation geometric contribution factor corresponding to the i-th observation data; tr() represents the matrix trace operation, which calculates the sum of the elements on the main diagonal of the matrix; H represents the m×3 dimensional observation geometric matrix composed of all retained observation data, where each row of the matrix corresponds to the transpose of the unit direction vector of an observation. ; It represents the observation geometry matrix formed by the remaining observation data after removing the i-th observation; The comprehensive observation weight corresponding to each retained observation data point is calculated based on three types of confidence contribution factors, and the normalization process of all comprehensive observation weights is completed. The formula for calculating the comprehensive observation weight corresponding to the retained observation data is as follows: ; In the formula, denoted as the normalized comprehensive observation weight of the i-th retained observation; m represents the total number of remaining observations after removing outliers. A Kalman filter solution model is built in the cloud, and a state vector containing the planar position and velocity information of the tracked two-wheeled vehicle is set. The state transition matrix and process noise matrix are determined by combining the motion characteristics of the tracked two-wheeled vehicle. An observation noise covariance matrix is ​​constructed by combining the GPS positioning covariance parameter and the ranging error variance parameter of the auxiliary equipment. The matrix includes the projection component of the positioning error of the auxiliary equipment in the direction from the equipment to the tracked two-wheeled vehicle. Kalman filter prediction processing and Kalman filter update processing are performed sequentially to obtain the target position solution result and the target position covariance result of the tracked two-wheeled vehicle. Then, the positioning accuracy parameter, which characterizes the reliability of this solution, is calculated from the target position covariance result. Step S6: Perform multi-mode anonymized aggregation for privacy protection based on the positioning accuracy parameters, and output the final positioning result; Two encrypted aggregation methods are used in the cloud for parameter calculation. The first is the Google SecAgg secure aggregation protocol, which is based on a hybrid implementation of Shamir secret sharing and Diffie Hellman key exchange. The second is the Paillier homomorphic encryption algorithm, which performs summation of observation parameters from multiple devices in the encrypted domain. Regardless of which method is used, the cloud server can only obtain the statistical results after aggregation is completed. It cannot reverse the aggregation statistical results to reconstruct the original observation data content corresponding to any single auxiliary device. After encrypted and anonymized aggregation, the final location result of the tracked two-wheeled vehicle is obtained. The cloud pushes the final location result to the user's client. The cloud organizes the spatiotemporally aligned observation data, outlier sample labeling information, various confidence contribution factors, comprehensive observation weights, target position calculation results, and positioning accuracy characterization parameters corresponding to this interaction into a new record and stores it in the historical Bluetooth broadcast interaction event sample set maintained on the cloud side. Furthermore, a two-wheeled vehicle anti-theft tracking system based on fusion positioning is provided, which includes a sample mapping construction module, a message scheduling generation module, an offset analysis configuration module, and a positioning solution aggregation module. The sample mapping construction module is used to generate a mapping relationship for the expected number of broadcast data entries with intersecting coverage, and outputs the expected number of captured observation data entries corresponding to the interactive working condition; the message scheduling generation module is used to generate a broadcast database and outputs an interactive result data packet carrying position entry constraints; the offset analysis configuration module is used to perform road normal side offset analysis and outputs target position data entry configuration parameters; the positioning calculation and aggregation module is used to complete observation preprocessing, position fusion calculation and encrypted anonymized aggregation processing, and output the final positioning result of the tracked two-wheeled vehicle. The sample mapping construction module includes a sample set acquisition unit and a working condition mapping calculation unit. The sample set acquisition unit is used to acquire a sample set of historical Bluetooth broadcast interaction events and complete the preset statistical time slice segmentation process to obtain the overall interaction set. The working condition mapping calculation unit is used to distinguish between fully covered interaction working conditions and insufficiently covered interaction working conditions, perform data completion processing, and offline construct the expected number of broadcast data entries mapping relationship of intersecting coverage. The message scheduling generation module includes a database generation unit and a data packet encapsulation unit; the database generation unit is used to read the comprehensive interaction condition marker in the overall interaction set record and generate a broadcast database corresponding to the condition; the data packet encapsulation unit is used to complete the BLE message distribution according to the message transmission priority, encapsulate the interaction result data packet and complete the data packet reporting transmission. The offset analysis configuration module includes a window parsing unit and a parameter distribution unit; the window parsing unit is used to load the sliding analysis window, parse the interactive result data packet to extract the location data and calculate the normal relative offset; the parameter distribution unit is used to generate the location drift baseline and the historical maximum drift baseline, obtain the target location data entry configuration parameters through mapping operation and distribute them to the base station. The positioning and aggregation module includes a spatiotemporal alignment processing unit and a solution aggregation output unit. The spatiotemporal alignment processing unit is used to complete the position entry constraints based on the target position data entry configuration parameters, perform coordinate transformation and time interpolation calculation on the observation data of multiple auxiliary devices, and output the aligned observation data. The solution aggregation output unit is used to perform observation outlier removal, confidence contribution factor calculation, Kalman filter fusion calculation to obtain positioning accuracy parameters, and output the final positioning result after performing multi-mode anonymized aggregation privacy protection processing.

[0006] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention distinguishes between sufficient and insufficient coverage interaction conditions by segmenting and statistically analyzing historical Bluetooth broadcast interaction event sample sets, and constructs a mapping relationship between the expected number of broadcast data messages under the intersection coverage. Based on offline statistical results, it quantifies the data reception capability under different radio frequency intersection conditions, thus getting rid of the method of simply relying on experience to set the number of message outputs. It adapts and optimizes for Bluetooth edge instantaneous interaction conditions, improves the problem of uncontrollable observation sample quantity under edge interaction, and provides reliable data basis for vehicle-mounted broadcast message scheduling.

[0007] 2. This invention generates a broadcast database corresponding to the operating conditions based on the mapping relationship, and combines message interruption continuation to send messages, dynamically adjusting the broadcast output scale according to the actual channel interaction conditions. It adapts message output for edge interaction scenarios with short dwell times, continuing to send message content that was not fully transmitted in the previous round, effectively improving the overall transmission integrity of Bluetooth broadcast messages and mitigating the situation of incomplete data acquisition caused by message loss in instantaneous intersection conditions.

[0008] 3. This invention utilizes a sliding analysis window to perform road normal lateral deviation analysis, constructs a position drift baseline, and combines offset degree mapping calculation to adaptively output position data entries and configure parameters, thereby enabling the reported data volume to be dynamically adjusted according to the actual drift state of the vehicle; when the drift is abnormal, the reported samples are increased, while the baseline reporting scale is maintained in the normal drift range, avoiding positioning errors caused by insufficient samples or invalid reporting that consumes transmission resources, and providing high-quality input data for cloud fusion calculation. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating a two-wheeled vehicle anti-theft tracking method based on fusion positioning according to the present invention. Detailed Implementation

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0011] Example 1: As Figure 1 As shown, this invention provides a technical solution: a two-wheeled vehicle anti-theft tracking method based on fusion positioning, and a unified data base management method including the following steps: Step S1: Obtain a sample set of historical Bluetooth broadcast interaction events, establish a mapping relationship between the expected number of broadcast data items for the degree of intersection coverage, and obtain the expected data volume corresponding to different intersection states; Select any tracked two-wheeled vehicle as the research object, and obtain a historical Bluetooth broadcast interaction event sample set formed by the actual interactions between the tracked two-wheeled vehicle and multiple auxiliary devices during past business operations; the historical Bluetooth broadcast interaction event sample set consists of multiple completed Bluetooth interaction event records, each interaction event record corresponds to a real BLE broadcast interaction process between the tracked two-wheeled vehicle and an auxiliary device, the auxiliary device refers to a third-party smart terminal device equipped with a compatible SDK that can scan and capture the BLE broadcast messages sent by the tracked two-wheeled vehicle, including but not limited to smartphones, smart tablets and portable smart IoT terminals with Bluetooth background scanning capabilities; Each interaction event record contains the interaction information corresponding to the interaction process. The interaction information includes the effective intersection degree of the Bluetooth radio frequency coverage between the auxiliary device and the tracked two-wheeled vehicle, the effective interaction dwell time of the Bluetooth broadcast, the interval of the vehicle Bluetooth broadcast, and the number of observation data entries actually captured during the interaction process. Among them, each observation data includes a set of auxiliary device location data, RSSI ranging data, and message capture timestamp. The interaction conditions are divided into sufficient coverage interaction conditions and insufficient coverage interaction conditions based on the actual degree of intersection between the Bluetooth radio frequency coverage area between the auxiliary device and the tracked two-wheeled vehicle. The aforementioned fully covered interactive condition indicates that the Bluetooth radio frequency coverage areas of the auxiliary device and the tracked two-wheeled vehicle overlap sufficiently, and the signal interaction can maintain a long dwell time, ensuring that the tracked two-wheeled vehicle can complete the broadcast interaction of a preset number of message entries. The insufficient coverage interaction condition indicates that the auxiliary device and the Bluetooth radio frequency coverage area of ​​the tracked two-wheeled vehicle only form a small-scale instantaneous intersection, and the interaction can maintain a short dwell time; multipath and channel fading will affect both types of interaction conditions, resulting in random message loss. The historical Bluetooth broadcast interaction event sample set is segmented according to the preset statistical time slice, and all Bluetooth interaction event records generated by the tracked two-wheeled vehicle and multiple auxiliary devices within the same preset statistical time period are classified into the same overall interaction set. When a set of interactions contains both interaction event records that fully cover interaction conditions and interaction event records that do not fully cover interaction conditions, the statistical results corresponding to the fully covered interaction conditions in the set are used to supplement the data of the interaction event records corresponding to the insufficiently covered interaction conditions. When all interaction event records within a certain overall interaction set belong to the insufficiently covered interaction conditions, the original statistical results corresponding to the insufficiently covered interaction conditions are directly adopted, and no data completion operation is performed. For each overall interaction set obtained after segmentation, the Bluetooth interaction event records under the two types of interaction conditions are counted. Combined with the vehicle BLE broadcast cycle and the background scanning duty cycle of the auxiliary equipment, the expected number of captured observation data points corresponding to the two types of interaction conditions are obtained respectively. For the fully covered interactive working condition, the theoretical number of messages that can be received at the radio frequency link layer is obtained by calculating the ratio between the interaction dwell time and the vehicle Bluetooth beacon broadcast period and rounding it down. The theoretical number of messages that can be received at the radio frequency link layer is compared with the maximum number of messages that can be reported by the auxiliary equipment in a single interaction service, and the smaller value is selected as the expected number of interaction observation data points corresponding to the fully covered interactive working condition. For cases where all interaction event records within the overall interaction set belong to the insufficient coverage interaction condition, based on all historical interaction records belonging to the insufficient coverage interaction condition in the historical Bluetooth broadcast interaction event sample set, the interaction dwell time corresponding to each historical interaction record is extracted and the ratio of the vehicle Bluetooth beacon broadcast period is calculated and rounded down. The result is then multiplied by the channel loss probability coefficient obtained from the historical Bluetooth broadcast interaction event sample set to obtain the theoretical number of packets that can be received at the RF link layer after channel disturbance correction for each historical interaction record. Statistical processing is then performed on the theoretical number of packets that can be received at the RF link layer after channel disturbance correction for all historical interaction records to obtain the expected number of interaction observation data points corresponding to the insufficient coverage interaction condition. A mapping relationship for the expected number of broadcast data entries with overlapping coverage is constructed. This mapping relationship is generated offline based on a historical Bluetooth broadcast interaction event sample set after time slicing and on-demand data completion processing. During real-time service operation, real-time interaction data is synchronously divided according to the preset statistical time slices, and all auxiliary device interactions within the same time period are grouped into a single real-time overall interaction set. During the real-time Bluetooth interaction startup phase, the expected number of interaction observation data entries corresponding to the fully covered interaction condition is used as the criterion by default. After a single Bluetooth interaction is completed, the corresponding interaction condition is recorded and marked according to the actual observation data entries produced in this interaction. A single observation data entry includes auxiliary device location data, RSSI ranging data, and a message capture timestamp. In practice, any tracked two-wheeled vehicle in normal road driving condition is selected. The construction of the historical Bluetooth broadcast interaction event sample set is performed offline in the cloud background, without occupying the computing resources of the two-wheeled vehicle's on-board terminal. During business operation, the actual Bluetooth interaction records generated between the tracked two-wheeled vehicle and various third-party auxiliary devices are continuously collected, and new interaction event records are continuously added to the historical Bluetooth broadcast interaction event sample set. When performing segmentation processing, a fixed preset statistical time slice is used uniformly, and all interaction event records are strictly divided according to time boundaries. Each interaction event record is only included in one overall interaction set, and cross-set duplication is prevented. When distinguishing interaction conditions, the determination is only based on the original records of the radio frequency intersection degree and interaction dwell time actually collected during the interaction process, without manually modifying the condition attribute labels of the interaction event records. Data completion operations only apply to records within the same overall interaction set, and cross-set calls for data completion do not occur between different overall interaction sets. After the expected number of captured observation data records corresponding to the two types of conditions obtained by offline statistics are stored, they are directly available for real-time business processes to read and call. Real-time business processes do not modify the offline statistical results. Step S2: Based on the transmission features extracted from the historical Bluetooth broadcast interaction event sample set, generate a broadcast database corresponding to the working conditions, generate interaction result data packets based on the broadcast database, and upload them to the cloud. Based on the historical Bluetooth broadcast interaction event sample set, extract the data item characteristics of the actual transmission completed by each historical interaction event record belonging to the fully covered interaction working condition, and statistically obtain the maximum amount of data that can be transmitted on average under the fully covered interaction working condition, which is recorded as the full transmission threshold. The tracked two-wheeled vehicle locally persists the overall interaction set record corresponding to the previous preset statistical time slice. The overall interaction set record corresponding to the previous preset statistical time slice at least saves the device identifier, the comprehensive interaction condition mark under that time slice, and the position of the message breakpoint where the transmission was not completed. The tracked two-wheeled vehicle does not locally persist the complete trajectory data, but only temporarily maintains the message slices to be sent in the broadcast database. All location-related observation data are transmitted back to the cloud via the auxiliary device that establishes the Bluetooth interaction. The tracked two-wheeled vehicle locally creates a broadcast database. The broadcast database is a temporary scheduling database and is only used to store the broadcast message slices to be sent out in this interaction. The message entry capacity of the broadcast database is limited by the number of expected observation data entries for the corresponding working condition. Once the auxiliary device establishes a Bluetooth interaction link with the tracked two-wheeled vehicle within the same preset statistical time period, the tracked two-wheeled vehicle reads its own device identifier and the comprehensive interaction condition marker carried by the overall interaction set record corresponding to the previous preset statistical time slice stored locally. Based on the read comprehensive interaction condition marker, it determines the broadcast database to be generated for this interaction. When the previous interaction event record is marked as a fully covered interaction condition, the number of expected interaction observation data entries corresponding to the fully covered interaction condition is called, and the number of expected interaction observation data entries is selected as the message entry capacity standard of the broadcast database to generate the broadcast database corresponding to the fully covered interaction condition. When the previous interaction event record is marked as an insufficient coverage interaction condition, the expected number of interaction observation data entries corresponding to the sufficient coverage interaction condition and the expected number of interaction observation data entries corresponding to the insufficient coverage interaction condition are called. Combined with the uncompleted transmission message breakpoint positions recorded in the previous interaction event record, a broadcast database corresponding to the insufficient coverage interaction condition is generated. The broadcast database selects the full transmission threshold number of messages as the message entry capacity standard for this interaction, and loads the message slice to be transmitted from the message breakpoint position. After the broadcast database is generated, the tracked two-wheeled vehicle transmits the message entries stored in the broadcast database to the auxiliary device via BLE broadcast, prioritizing the transmission of the latest generated message entries first, followed by the transmission of residual message entries from the previous interaction that were not completed. After all Bluetooth interactions corresponding to the current preset statistical time period have ended, the tracked two-wheeled vehicle counts the data entries that were actually successfully broadcast under the overall interaction set of that time period. Based on the data entries that were actually successfully broadcast under the overall interaction set of that time period, the vehicle determines the comprehensive interaction condition corresponding to the preset statistical time slice, writes the corresponding comprehensive interaction condition flag into the overall interaction set record corresponding to the preset statistical time slice, and updates the position of the message breakpoints that were not completed in the overall interaction set record. The tracked two-wheeled vehicle extracts the device identifier of the tracked two-wheeled vehicle corresponding to this interaction, the identification information of the auxiliary device, the number of data entries actually broadcast in this interaction, the interaction status marker of this interaction, the upper limit of the message entry capacity of the broadcast database used in this interaction, and the updated message breakpoint position. All of the above information is encapsulated into an interaction result data packet. The transmission link for the auxiliary device to report the interaction result data packet can rely on the BLE Mesh protocol to realize multi-hop self-organizing network relay transmission. At the same time, the Boneh Boyen Shacham group signature scheme is used to sign the message. The aggregation node can verify that the message comes from a legitimate device cluster, but cannot identify the auxiliary device that sends the message. The interaction result data packet after the signature protection is completed is then transmitted to the cloud. The interaction result data packet is merged into the historical Bluetooth broadcast interaction event sample set maintained on the cloud side. In practice, the full-volume transmission threshold is issued by the cloud through offline statistics based on a sample set of historical Bluetooth broadcast interaction events. The vehicle-mounted terminal of the tracked two-wheeled vehicle is only responsible for receiving and storing the threshold parameter, and does not participate in the statistical calculation of the threshold. The broadcast database is a temporary scheduling object in the vehicle terminal's memory. It is instantiated only within the interaction period of the corresponding preset statistical time slice. After the interaction period ends, all message slices inside the broadcast database are released directly and are not written to the non-volatile storage medium of the vehicle terminal. The message breakpoint position only records the sequence number of the message slice that was not sent in the previous round of interaction, and does not store the message slice itself. The message delivery process strictly follows the preset transmission priority, prioritizing the output of newly generated message slices, and then continuing to process the residual message slices corresponding to the breakpoint position. The interaction end determination is based on the preset statistical time slice time boundary. When the time boundary is reached, the data entries of the actual broadcast output in this round are counted, and the comprehensive interaction condition mark is rewritten and the message breakpoint position is updated. After the interaction result data packet is encapsulated, it is handed over to the auxiliary equipment for reporting. BLE Mesh multi-hop relay transmission and Boneh Boyen Shacham group signature processing are completed on the auxiliary equipment and the aggregation node side. The vehicle terminal does not perform signature calculation. Step S3: The cloud performs road normal lateral deviation analysis on the uploaded interactive result data packet based on the sliding analysis window, constructs the position drift baseline, and adaptively determines the number of position data reporting entries; A sliding analysis window is configured on the cloud side. The sliding analysis window extracts data according to the time dimension of the upload of the interaction result data packet and updates the window content according to the preset sliding step size. The sliding analysis window extracts multiple historical interaction result data packets that have been received and stored on the cloud side. Each interaction result data packet carries multiple sets of location data. Each set of location data is continuously collected according to a fixed collection time interval, and each set of location data is bound to a corresponding collection timestamp. The cloud reads all interactive result data packets captured by the sliding analysis window, parses and extracts all location data and the device identifier of the tracked two-wheeled vehicle carried in the interactive result data packets; for each set of location data belonging to the same tracked two-wheeled vehicle, it reads the road vector data pre-stored in the geographic information database. The road vector data at least saves the coordinate sequence of the road centerline and the coordinate sequence of the two sides of the road. Project the planar coordinates of a single set of location data onto the plane containing the road vector data, and calculate the normal relative offset of the location data towards the corresponding road edge. The normal is the vector direction perpendicular to the local extension direction of the road. The longitudinal displacement component generated by the road extension direction does not participate in the calculation of the normal relative offset. The positive and negative values ​​of the normal relative offset distinguish the different sides of the location data from the road, and the absolute value of the normal relative offset represents the actual distance of the location data from the road edge. The normal relative offsets corresponding to all position data belonging to the same tracked two-wheeled vehicle within the sliding analysis window are traversed. Statistical operations are performed on all normal relative offsets to generate the position drift baseline corresponding to the tracked two-wheeled vehicle. The position drift baseline records the statistical distribution interval corresponding to the normal relative offsets, and this statistical distribution interval is used as the regular offset determination interval. The maximum value of the absolute value of all normal relative offsets within the sliding analysis window is extracted from the cloud and defined as the historical maximum drift baseline. For each set of position data within the sliding analysis window, the absolute value of the relative normal offset is compared with the boundary of the conventional offset judgment interval corresponding to the position drift baseline. When the absolute value of the normal relative offset falls within the boundary of the conventional offset determination interval, determine the number of position data entries for the configuration reference of the interaction result data packet. When the absolute value of the normal relative offset crosses the boundary of the regular offset determination interval, a mapping operation is performed between the current absolute value of the normal relative offset and the historical maximum drift baseline. The mapping operation outputs the value of the target position data entry. The input of the mapping operation is the current absolute value of the normal relative offset, the historical maximum drift baseline, and the number of reference position data entries. The mapping operation outputs the value of the position data entries that the interactive result data packet needs to carry. After the cloud completes the mapping calculation, it generates the target location data entry configuration parameters corresponding to the tracked two-wheeled vehicle; the cloud sends the target location data entry configuration parameters to each base station; after the tracked two-wheeled vehicle establishes a communication connection with the base station, the tracked two-wheeled vehicle reads the target location data entry configuration parameters forwarded by the base station and replaces the original location data entry configuration parameters stored locally on the tracked two-wheeled vehicle. The tracked two-wheeled vehicle extracts the device identifier, auxiliary device identification information set, number of data entries actually broadcast in the overall interaction set corresponding to the current preset statistical time slice, comprehensive interaction condition marker, message entry capacity limit of the broadcast database used in this interaction, and updated message breakpoint position from the overall interaction set corresponding to the current preset statistical time slice. All of the above information is encapsulated into an interaction result data packet. The interaction result data packet is transmitted to the cloud by the auxiliary device that captures the Bluetooth broadcast. Newly added interaction result data packets participate in the subsequent interception and statistical calculation of the sliding analysis window. In practice, the sliding analysis window runs continuously in the cloud background. The sliding step size and the number of data entries that the window can hold are pre-configured. Data is entered into the window according to the upload time of the interaction result data packets. Old data is automatically removed from the window when it reaches the window's elimination criteria. The historical interaction result data packets removed from the window are not directly deleted but are still stored in the cloud database. The geographic information database pre-imports the road vector data of the corresponding area. The coordinate sequences of the road centerline and road edge line are kept in read-only mode. The entire analysis process does not modify the original road vector data. The location data projection and normal relative offset are all calculated in the plane coordinate system. The longitudinal displacement component caused by the road direction is isolated throughout the process and does not participate in the offset statistical calculation. The location drift baseline and the historical maximum drift baseline are updated with the window scrolling. Every time the window slides, the statistical calculation of all offsets is re-executed. The input parameters of the mapping operation are all taken from the results obtained from the current window statistics and do not introduce historical offset data outside the window. The configuration parameters of the generated target location data entries are transmitted to the vehicle terminal via the base station's downlink. After receiving the configuration parameters, the vehicle terminal directly overwrites the local old configuration without making additional parameter corrections. Step S4: Based on the determined number of reported entries for fixed-location data, configure parameters to complete the position entry constraints of the interaction result data packet, and perform spatiotemporal alignment preprocessing on the observation data of the multi-auxiliary equipment corresponding to the overall interaction set under each preset statistical time slice; Based on the target location data entry configuration parameters, during the generation of interaction result data packets, the tracked two-wheeled vehicle selects location data according to the number of entries limited by the configuration parameters and completes the location entry constraint processing. The interaction result data packets corresponding to each preset statistical time slice are continuously uploaded to the cloud. The interaction result data packets carry sniffing observation data reported by multiple auxiliary devices within the overall interaction set. Each observation data includes auxiliary device location data, RSSI ranging data, and message capture timestamp. The cloud retrieves the pre-stored Bluetooth broadcast effective detection radius parameters and uses the historical target location estimation results of the tracked two-wheeled vehicle that have been calculated in the previous business process to calculate the geometric distance between the auxiliary device location and the historical target location estimation results in each observation data. Observation data whose geometric distance exceeds the Bluetooth broadcast effective detection radius parameters are no longer involved in coordinate transformation processing. From the auxiliary equipment corresponding to the retained observation data, one auxiliary equipment whose positioning quality meets the preset screening criteria is selected as the reference equipment for spatiotemporal alignment. The latitude, longitude, and altitude position information of the reference equipment in the geodetic coordinate system is obtained and converted to the geocentric-fixed coordinate system. Using the geocentric-fixed coordinate system coordinates of the reference equipment as the reference origin, the latitude, longitude, and altitude position information of the other auxiliary equipment corresponding to the retained observations are uniformly converted to the northeast-sky local coordinate system. The calculation formula for the northeast-sky local coordinate system is as follows: ; In the formula, , and Let them be the east, north, and celestial coordinates of the i-th auxiliary device in the northeast-sky coordinate system with the k-th reference auxiliary device as the origin; The longitude is represented as that of the reference device k; Represented as the latitude of the reference device k; Represented as the geocentric coordinate system coordinates of the i-th auxiliary device; Represented as the geocentric coordinate system coordinates of reference device k; When the spatial distance between the auxiliary device and the reference device involved in the conversion exceeds the preset spatial span threshold, the local reference origin of the corresponding auxiliary device is switched to perform coordinate conversion, thereby completing the coordinate benchmark unification of all auxiliary device position data; The cloud identifies the reporting time differences between the capture timestamps of each observation data message, retrieves the historical motion state estimation results of the tracked two-wheeled vehicle to obtain the motion speed parameters, and combines the historical target position estimation results of the tracked two-wheeled vehicle with the position solution of the auxiliary equipment to obtain the unit direction vector of the auxiliary equipment pointing to the tracked two-wheeled vehicle; selects a pre-set unified reference time as the alignment target time, and based on the motion speed parameters of the tracked two-wheeled vehicle, the unit direction vector, and the original message capture timestamps, performs time dimension interpolation calculation on the RSSI ranging data of each observation data, converts all observation data to the unified reference time, and completes the time alignment of the observation data; after coordinate benchmark unification and time alignment processing, a complete set of aligned observation data is obtained; In practice, the effective detection radius parameter of Bluetooth broadcast is pre-configured in the cloud. Each round of processing uses this parameter to perform initial screening of observation data. Observation data filtered out by geometric distance conditions is directly excluded from all subsequent processing links and will not participate in coordinate transformation and time interpolation calculations. The reference device is selected from the set of auxiliary devices retained after initial screening. The selection criteria are the GPS positioning covariance index reported by the auxiliary devices. Auxiliary devices with lower positioning uncertainty are selected as reference devices first. The coordinate transformation calculation is performed entirely in the cloud. When the spatial distance between the auxiliary device and the current reference device exceeds the preset spatial span threshold, the reference origin is immediately changed and the transformation is re-executed to avoid the conversion deviation caused by coordinate transformation under large spatial spans. The movement speed of the tracked two-wheeled vehicle used in the time alignment stage is taken from the output result of the previous round of fusion calculation. This speed parameter will be iteratively updated with each round of positioning calculation. Time interpolation calculation is only performed on RSSI ranging values. The original reported values ​​of the auxiliary device position and timestamp fields are retained. After interpolation, only new ranging fields are generated. The original reported content is not rewritten. Step S5: Using the spatiotemporally aligned observation data, perform outlier removal, multi-dimensional confidence contribution factor calculation, and target position fusion calculation of the tracked two-wheeled vehicle; The aligned observation data is used to continue the calculation. Each observation data corresponds to a set of auxiliary equipment positions, distance values ​​converted from RSSI, auxiliary equipment GPS positioning covariance parameters, and distance error variance parameters. Combined with the preliminary position estimation results of the tracked two-wheeled vehicle, the distance residual corresponding to each observation data is calculated. The distance residual is the absolute value of the difference between the observed distance value and the geometric distance between the auxiliary equipment and the tracked two-wheeled vehicle calculated from the preliminary position estimation results. The Mahalanobis distance test value is calculated by combining the distance measurement error variance parameter of a single observation data and the GPS positioning covariance parameter of the auxiliary equipment. The Mahalanobis distance test judgment threshold is retrieved from the cloud for numerical comparison. Observation data with Mahalanobis distance test values ​​greater than the judgment threshold are marked as outliers and removed, while observation data with Mahalanobis distance test values ​​less than or equal to the judgment threshold are retained. For all the remaining observation data after removing outliers, the GPS confidence contribution factor, the distance measurement confidence contribution factor, and the observation geometric contribution factor were calculated respectively. The formula for calculating the GPS confidence contribution factor is as follows: ; In the formula, Let the GPS confidence contribution factor be represented as the one corresponding to the i-th observation; Let represent the variance of the GPS position error of the i-th auxiliary device; This is represented as the preset reference positioning variance parameter; The formula for calculating the ranging confidence contribution factor is as follows: ; In the formula, This represents the ranging confidence contribution factor corresponding to the i-th observation; represents the variance of the ranging error corresponding to the i-th auxiliary device; represents the preset reference ranging variance parameter. The formula for calculating the observation geometric contribution factor is as follows: Let the observation direction be a unit vector. ; Observation geometry matrix: ; ; In the formula, Let represent the observation geometric contribution factor corresponding to the i-th observation data; tr() represents the matrix trace operation, which calculates the sum of the elements on the main diagonal of the matrix; H represents the m×3 dimensional observation geometric matrix composed of all retained observation data, where each row of the matrix corresponds to the transpose of the unit direction vector of an observation. ; It represents the observation geometry matrix formed by the remaining observation data after removing the i-th observation; The comprehensive observation weight corresponding to each retained observation data point is calculated based on three types of confidence contribution factors, and the normalization process of all comprehensive observation weights is completed. The formula for calculating the comprehensive observation weight corresponding to the retained observation data is as follows: ; In the formula, denoted as the normalized comprehensive observation weight of the i-th retained observation; m represents the total number of remaining observations after removing outliers. A Kalman filter solution model is built in the cloud, and a state vector containing the planar position and velocity information of the tracked two-wheeled vehicle is set. The state transition matrix and process noise matrix are determined by combining the motion characteristics of the tracked two-wheeled vehicle. An observation noise covariance matrix is ​​constructed by combining the GPS positioning covariance parameter and the ranging error variance parameter of the auxiliary equipment. The matrix includes the projection component of the positioning error of the auxiliary equipment in the direction from the equipment to the tracked two-wheeled vehicle. Kalman filter prediction processing and Kalman filter update processing are performed sequentially to obtain the target position solution result and the target position covariance result of the tracked two-wheeled vehicle. Then, the positioning accuracy parameter, which characterizes the reliability of this solution, is calculated from the target position covariance result. In practice, the ranging residual is calculated based on the target position estimation result from the previous round and the retained observation data; the residual calculation only involves absolute value determination. The Mahalanobis distance test threshold is pre-configured in the cloud and does not change dynamically with each batch of observation data throughout the entire operation. Outlier samples are marked, and observations marked as outliers no longer participate in confidence factor and Kalman filter related calculations, but the original aligned observation records are completely retained. The GPS confidence contribution factor, ranging confidence contribution factor, observation geometric contribution factor, and comprehensive observation weight are all calculated only for the observation samples retained after outlier screening. The observation geometric matrix is ​​reconstructed in each round of calculation based on the current retained observation set, and the matrix after removing a single observation is generated synchronously to obtain the geometric contribution factor. The state vector, state transition matrix, and process noise matrix parameters used in Kalman filtering are fixed, while the observation noise covariance matrix is ​​dynamically generated by the actual statistical parameters of each round of observations. Step S6: Perform multi-mode anonymized aggregation for privacy protection based on the positioning accuracy parameters, and output the final positioning result; Two encrypted aggregation methods are used in the cloud for parameter calculation. The first is the Google SecAgg secure aggregation protocol, which is based on a hybrid implementation of Shamir secret sharing and Diffie Hellman key exchange. The second is the Paillier homomorphic encryption algorithm, which performs summation of observation parameters from multiple devices in the encrypted domain. Regardless of which method is used, the cloud server can only obtain the statistical results after aggregation is completed. It cannot reverse the aggregation statistical results to reconstruct the original observation data content corresponding to any single auxiliary device. After encrypted and anonymized aggregation, the final location result of the tracked two-wheeled vehicle is obtained. The cloud pushes the final location result to the user's client. The cloud organizes the spatiotemporally aligned observation data, outlier sample labeling information, various confidence contribution factors, comprehensive observation weights, target position calculation results, and positioning accuracy characterization parameters corresponding to this interaction into a new record and stores it in the historical Bluetooth broadcast interaction event sample set maintained on the cloud side.

[0012] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention.

Claims

1. A two-wheeled vehicle anti-theft tracking method based on fusion positioning, characterized in that: The two-wheeled vehicle anti-theft tracking method includes: Obtain a sample set of historical Bluetooth broadcast interaction events and establish a mapping relationship between the expected number of broadcast data entries and the degree of overlapping coverage. Based on the transmission characteristics extracted from the historical Bluetooth broadcast interaction event sample set, a broadcast database corresponding to the working conditions is generated. Based on the broadcast database, an interaction result data packet is generated and uploaded to the cloud. The cloud-based system performs road normal lateral deviation analysis on the uploaded interactive result data packets based on a sliding analysis window, constructs a position drift baseline, and adaptively determines the number of position data reporting entries by combining the absolute value of the relative normal offset. The construction of the location drift baseline includes configuring a sliding analysis window in the cloud and extracting multiple stored historical interaction result data packets according to a preset statistical time slice as the sliding step size; The total normal relative offset within the sliding analysis window is statistically analyzed to generate the position drift baseline of the tracked two-wheeled vehicle. The position drift baseline includes the statistical distribution interval of the normal relative offset as the regular offset determination interval, and the maximum absolute value of the normal relative offset within the sliding analysis window is extracted as the historical maximum drift baseline. Based on the determined number of reported entries for fixed-location data, the configuration parameters are used to complete the position entry constraints of the interaction result data packet, and the spatiotemporal alignment preprocessing is performed on the observation data of multiple auxiliary devices corresponding to the overall interaction set under each preset statistical time slice; Using spatiotemporally aligned observation data, outlier removal, multi-dimensional confidence contribution factor calculation, and target position fusion calculation of the tracked two-wheeled vehicle are performed to obtain positioning accuracy parameters. After privacy protection processing through anonymous aggregation, the final positioning result is output.

2. The two-wheeled vehicle anti-theft tracking method based on fusion positioning according to claim 1, characterized in that: The establishment of the expected number of broadcast data entries for the degree of intersection coverage includes: A sample set of historical Bluetooth broadcast interaction events generated by the tracked two-wheeled vehicle and multiple auxiliary devices is obtained. The sample set of historical Bluetooth broadcast interaction events includes multiple Bluetooth interaction event records. Each Bluetooth interaction event record stores the effective intersection degree of Bluetooth radio frequency coverage, the effective interaction dwell time of Bluetooth broadcast, the vehicle Bluetooth broadcast interval, and the number of observation data entries. Each observation data record stores auxiliary device location data, RSSI ranging data, and message capture timestamp. Based on the degree of overlap of Bluetooth radio frequency coverage, the interaction conditions are divided into those with sufficient coverage and those with insufficient coverage. The historical Bluetooth broadcast interaction event sample set is divided into multiple overall interaction sets according to the preset statistical time slices. Perform data completion or retain original statistical results on the entire interaction set; By combining the vehicle-mounted BLE broadcast cycle and the background scanning duty cycle of the auxiliary equipment, the expected number of captured observation data points for the fully covered interactive working condition and the insufficiently covered interactive working condition are calculated respectively. Based on the processed historical Bluetooth broadcast interaction event sample set, an offline mapping relationship is constructed for the expected number of broadcast data entries with intersecting coverage. In real-time business, real-time interactive data is divided into real-time overall interactive sets according to preset statistical time slices. When an interaction starts, the expected number of captured observation data corresponding to the fully covered interactive conditions is used as the criterion. After a single Bluetooth interaction ends, the corresponding interactive conditions are marked according to the actual output of observation data items.

3. The two-wheeled vehicle anti-theft tracking method based on fusion positioning according to claim 1, characterized in that: The step of generating interactive result data packets based on the broadcast database and uploading them to the cloud includes: The full-volume transmission threshold corresponding to fully covered interactive conditions was obtained based on the statistical analysis of historical Bluetooth broadcast interaction event sample sets. The tracked two-wheeled vehicle locally stores the overall interaction set record corresponding to the previous preset statistical time slice. The overall interaction set record stores at least the device identifier, the comprehensive interaction condition mark, and the position of the message breakpoint that was not completed. The tracked two-wheeled vehicle locally creates a broadcast database to store broadcast message slices to be sent. The message entry capacity of the broadcast database is limited by the number of expected observation data entries for the corresponding working condition. After the auxiliary device establishes a Bluetooth interaction link with the tracked two-wheeled vehicle, it reads the comprehensive interaction condition marker stored locally and generates a broadcast database of the corresponding working conditions based on the comprehensive interaction condition marker. According to the transmission priority of prioritizing the transmission of the latest message entries and then continuing to transmit the remaining message entries, the message entries in the broadcast database are broadcast to the auxiliary equipment via BLE broadcast; After all Bluetooth interactions within the current preset statistical time slice have ended, the overall interaction status corresponding to the preset statistical time slice is determined based on the data entries that were actually successfully broadcast, and the overall interaction status flag and the breakpoint position of the message that was not transmitted are updated in the overall interaction set record. Extract the device identifier of the tracked two-wheeled vehicle, auxiliary device identification information, the number of data entries actually broadcast, interaction condition markers, the upper limit of the broadcast database message entry capacity, and the updated message breakpoint position, encapsulate the interaction result data packet, and upload it to the cloud.

4. The two-wheeled vehicle anti-theft tracking method based on fusion positioning according to claim 1, characterized in that: The method of constructing a position drift baseline and adaptively determining the number of position data reporting entries by combining the absolute value of the normal relative offset includes: Based on the parsing of the intercepted interaction result data packets, the location data belonging to the same tracked two-wheeled vehicle and the device identifier of the tracked two-wheeled vehicle are extracted. The pre-stored road vector data is read, which includes the coordinate sequence of the road centerline and the coordinate sequence of the two sides of the road. The planar coordinates of a single set of location data are projected onto the plane where the road vector data is located, and the normal relative offset of the location data relative to the road edge is calculated. The longitudinal displacement component is not included in the calculation of the normal relative offset.

5. A two-wheeled vehicle anti-theft tracking method based on fusion positioning according to claim 4, characterized in that: The method of adaptively determining the number of location data reporting entries by constructing a location drift baseline and combining it with the absolute value of the normal relative offset also includes: The absolute value of the relative normal offset is compared with the boundary of the regular offset determination interval. When the absolute value of the relative normal offset is within the boundary of the regular offset determination interval, the number of position data entries configured in the interaction result data packet is determined. When the absolute value of the relative normal offset exceeds the boundary of the regular offset determination interval, a mapping operation is performed based on the current absolute value of the relative normal offset, the historical maximum drift baseline, and the number of position data entries to obtain the value of the target position data entry corresponding to the interaction result data packet. Based on the target location data entry values ​​obtained from the mapping operation, target location data entry configuration parameters are generated and sent to each base station. After the tracked two-wheeled vehicle establishes a communication connection with the base station, it reads the target location data entry configuration parameters forwarded by the base station and replaces the original local location data entry configuration parameters.

6. The two-wheeled vehicle anti-theft tracking method based on fusion positioning according to claim 1, characterized in that: The process of performing spatiotemporal alignment preprocessing on the observation data of multiple auxiliary devices corresponding to the overall interaction set under each preset statistical time slice includes: Based on the configuration parameters of the target location data entries, the tracked two-wheeled vehicle selects location data and completes encapsulation according to the number of entries limited by the configuration parameters when generating the interaction result data packet, thereby realizing the location entry constraint; The cloud acquires the interaction result data packets of each preset statistical time slice, which are carried by multiple auxiliary devices sniffing and observing data. It retrieves the effective detection radius parameters of Bluetooth broadcast and the historical target position estimation results of the tracked two-wheeled vehicle to filter the observation data. The selected observation data is spatiotemporally aligned to obtain the geodetic coordinate system latitude, longitude, and altitude information of the reference device and convert it to the geocentric coordinate system. Using the geocentric coordinate system coordinates of the reference device as the reference origin, the latitude, longitude, and altitude information of the auxiliary devices corresponding to the remaining observations are converted to the northeast-sky local coordinate system. When the spatial distance between the auxiliary device and the reference device involved in the conversion exceeds the preset spatial span threshold, the local reference origin of the corresponding auxiliary device is switched to perform coordinate conversion, thus completing the coordinate benchmark unification of all auxiliary device position data. Identify the differences in the reporting timestamps of the capture timestamps of each observation data message, retrieve the historical motion state estimation results of the tracked two-wheeled vehicle to obtain the motion speed parameters, and combine the historical target position estimation results of the tracked two-wheeled vehicle with the position of the auxiliary equipment to solve for the unit direction vector of the auxiliary equipment pointing to the tracked two-wheeled vehicle. A preset unified reference time is selected as the alignment target time. Using the motion speed parameters of the tracked two-wheeled vehicle, the unit direction vector, and the original message capture timestamp, the RSSI ranging data of each observation data is interpolated in the time dimension. All observation data are converted to the unified reference time to obtain observation data in which both the coordinate reference and time are aligned.

7. The two-wheeled vehicle anti-theft tracking method based on fusion positioning according to claim 1, characterized in that: The process of outlier removal, multi-dimensional confidence contribution factor calculation, and target position fusion of the tracked two-wheeled vehicle yields positioning accuracy parameters, including: Acquire spatiotemporally aligned observation data. Each observation data point includes the location of the auxiliary equipment, the ranging value converted from RSSI, the GPS positioning covariance parameter of the auxiliary equipment, and the ranging error variance parameter. Calculate the distance residuals for each observation data based on the preliminary position estimation results of the tracked two-wheeled vehicle; The Mahalanobis distance test value is calculated by combining the distance measurement error variance parameter and the GPS positioning covariance parameter of the auxiliary equipment. The Mahalanobis distance test value is compared with the preset Mahalanobis distance test judgment threshold, and the observation data with the Mahalanobis distance test value greater than the judgment threshold are eliminated. For the retained observation data, the GPS confidence contribution factor, the distance measurement confidence contribution factor, and the observation geometric contribution factor are calculated respectively. The comprehensive observation weight is calculated based on the GPS confidence contribution factor, the ranging confidence contribution factor, and the observation geometric contribution factor, and then normalization is performed on the comprehensive observation weight. A Kalman filter solution model is constructed, a state vector containing the planar position information and motion speed information of the tracked two-wheeled vehicle is set, the state transition matrix and process noise matrix are determined, and an observation noise covariance matrix incorporating the directional projection component of the positioning error of the auxiliary equipment is constructed. Kalman filter prediction processing and Kalman filter update processing are performed sequentially to obtain the target position solution result and target position covariance result of the tracked two-wheeled vehicle. The positioning accuracy parameter is calculated from the target position covariance result. After anonymous aggregation and privacy protection processing, it is pushed to the user client.

8. A two-wheeled vehicle anti-theft tracking system based on fusion positioning, applied to the two-wheeled vehicle anti-theft tracking method based on fusion positioning as described in any one of claims 1-7, characterized in that: The two-wheeled vehicle anti-theft tracking system includes a sample mapping construction module, a message scheduling generation module, an offset analysis configuration module, and a positioning calculation and aggregation module. The sample mapping construction module is used to generate a mapping relationship for the expected number of broadcast data entries with intersecting coverage, and outputs the expected number of captured observation data entries corresponding to the interactive working condition; the message scheduling generation module is used to generate a broadcast database and outputs an interactive result data packet carrying position entry constraints; the offset analysis configuration module is used to perform road normal side offset analysis and outputs target position data entry configuration parameters; the positioning calculation and aggregation module is used to complete observation preprocessing, position fusion calculation and encrypted anonymized aggregation processing, and output the final positioning result of the tracked two-wheeled vehicle.

9. A two-wheeled vehicle anti-theft tracking system based on fusion positioning according to claim 8, characterized in that: The sample mapping construction module includes a sample set acquisition unit and a working condition mapping calculation unit. The sample set acquisition unit is used to acquire a sample set of historical Bluetooth broadcast interaction events and complete the preset statistical time slice segmentation process to obtain the overall interaction set. The working condition mapping calculation unit is used to distinguish between fully covered interaction working conditions and insufficiently covered interaction working conditions, perform data completion processing, and offline construct the expected number of broadcast data entries mapping relationship of intersecting coverage. The message scheduling generation module includes a database generation unit and a data packet encapsulation unit. The database generation unit is used to read the comprehensive interaction condition markers in the overall interaction set record and generate a broadcast database corresponding to the condition. The data packet encapsulation unit is used to complete the BLE message distribution according to the message transmission priority, encapsulate the interaction result data packet, and complete the data packet reporting and transmission.

10. A two-wheeled vehicle anti-theft tracking system based on fusion positioning according to claim 8, characterized in that: The offset analysis configuration module includes a window parsing unit and a parameter distribution unit; the window parsing unit is used to load the sliding analysis window, parse the interactive result data packet to extract position data and calculate the normal relative offset. The parameter distribution unit is used to generate the location drift baseline and the historical maximum drift baseline, obtain the target location data entry configuration parameters through mapping calculation, and distribute them to the base station; The positioning and aggregation module includes a spatiotemporal alignment processing unit and a solution aggregation output unit. The spatiotemporal alignment processing unit is used to complete the position entry constraints based on the target position data entry configuration parameters, perform coordinate transformation and time interpolation calculation on the observation data of multiple auxiliary devices, and output the aligned observation data. The solution aggregation output unit is used to perform observation outlier removal, confidence contribution factor calculation, Kalman filter fusion calculation to obtain positioning accuracy parameters, and output the final positioning result after performing multi-mode anonymized aggregation privacy protection processing.

Images