A method for positioning and tracking a sightseeing vehicle based on a Beidou satellite
By integrating multi-source data and correcting with a semantic topology map library, the problems of easy interference and insufficient accuracy in sightseeing vehicle positioning were solved, achieving high-precision, stable sightseeing vehicle positioning and real-time optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 湖南省特种设备检验检测研究院
- Filing Date
- 2026-03-20
- Publication Date
- 2026-06-23
Smart Images
Figure CN122260370A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of sightseeing vehicle positioning and tracking technology, and in particular to a sightseeing vehicle positioning and tracking method based on the BeiDou satellite system. Background Technology
[0002] With the widespread use of sightseeing vehicles in scenic areas and other locations, higher demands are placed on the accuracy, stability, and anti-interference capabilities of their positioning and tracking. Existing sightseeing vehicle positioning technologies mainly rely on single BeiDou satellite positioning or a fusion of BeiDou and inertial navigation, which has the following shortcomings: 1. Single BeiDou positioning is susceptible to environmental interference (such as obstruction and electromagnetic interference), resulting in large positioning errors (usually ≥1m). It cannot function properly in environments where GNSS is denied. 2. Traditional multi-source fusion uses fixed weights, which cannot adapt to dynamic changes in sensor reliability and has weak anti-interference capabilities; 3. The positioning trajectory is prone to jumps during dynamic driving, affecting the accuracy of trajectory monitoring; 4. Positioning deviation correction is mostly done after the fact, and cannot avoid dynamic deviations in advance, resulting in insufficient accuracy under dynamic driving conditions; 5. Lack of an environment adaptation mechanism for sightseeing vehicle driving scenarios, resulting in poor positioning stability in complex scenic environments. Summary of the Invention
[0003] The purpose of this invention is to provide a method for positioning and tracking sightseeing vehicles based on the BeiDou satellite system, thereby solving the aforementioned technical problems.
[0004] To achieve the above objectives, this invention provides a method for positioning and tracking sightseeing vehicles based on the BeiDou satellite system, comprising the following steps: S1. Combining the Beidou satellite positioning module, vehicle-mounted multi-sensor, environmental vision acquisition module and IMU inertial module, collect the sightseeing vehicle's positioning, operating status, environmental vision and inertial motion data, and construct a multi-source raw dataset through time stamp synchronization calibration; S2. Preprocess the multi-source raw dataset output by S1 to complete the coordinate transformation of Beidou data, temperature drift compensation of vehicle sensor, semantic target extraction of visual data, and output a standardized multi-source dataset. S3. Based on the standardized multi-source dataset output by S2, calculate the dynamic confidence weight of each acquisition module, construct an adaptive weighted Block IEKF block iterative filtering model, complete the fusion calculation of multi-source data, and output the fusion positioning result. S4. Based on the fusion positioning results output by S3, initialize the trajectory tracking state, calculate the association cost by jointly calculating motion consistency and position intersection-union ratio, and use the OCSORT algorithm to complete the trajectory association matching and update optimization, and output a smooth optimized trajectory. S5. Based on the smoothed optimized trajectory output by S4 and the semantic segmentation data output by S2, determine the GNSS denial environment state, construct a semantic topology feature vector and match it with the pre-constructed scenic area semantic topology map library, and perform weighted fusion of reference coordinates and optimized trajectory to complete the positioning correction, and output the corrected positioning result. S6. Based on the corrected positioning results output by S5, construct a time series deviation dataset, train the NARX neural network model, predict the positioning deviation through the model and complete the compensation correction, and output the final positioning and tracking results. S7. Based on the final positioning and tracking results output by S6, perform structuring and format standardization processing on the positioning data, upload the standardized data to the cloud platform through the NB-IoT communication module, determine the abnormal operating status of the sightseeing vehicle in combination with the preset safety threshold and trigger offline voice warning, and output cloud-stored data and local warning control signals. S8, based on the cloud storage data output by S7, generates a sightseeing bus route map through route point fitting and curve optimization. It collects real-time data on the entire sightseeing bus operation process to update the deviation dataset and the scenic area semantic topology map library. It optimizes the parameters of the NARX neural network model and the adaptive weighted Block IEKF filter model online to achieve dynamic iterative optimization of the positioning model and feature library.
[0005] Therefore, the above-mentioned sightseeing vehicle positioning and tracking method based on BeiDou satellite has the following beneficial effects: 1. Improve positioning accuracy: By integrating data from BeiDou, IMU, and vehicle-mounted sensors, and combining Block-IEKF block filtering and NARX bias compensation, multipath error and sensor drift are effectively reduced, and positioning accuracy can reach the centimeter level. 2. Enhanced environmental adaptability: By introducing a semantic topology map library, the positioning is corrected through semantic matching in GNSS-denied environments (such as tunnels, dense forests, and building obstructions), thus solving the problem of traditional positioning failure. 3. Ensure trajectory continuity: OCSORT trajectory optimization + observation center momentum mechanism is adopted to smooth trajectory jumps and avoid trajectory breaks caused by missing data or noise. 4. Improve system robustness: By using adaptive weighted filtering and multi-source redundancy design, the impact of a single sensor failure on the positioning results is reduced, thereby improving system stability; 5. Adapt to specific scenarios: The semantic topology map library can dynamically adjust speed and accuracy constraints according to scenic area scenarios (curves, slopes, speed limit zones) to match the actual working conditions of sightseeing vehicles traveling at low speeds.
[0006] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0007] Figure 1 This is a flowchart of a sightseeing vehicle positioning and tracking method based on the BeiDou satellite according to the present invention. Detailed Implementation
[0008] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.
[0009] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.
[0010] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0011] like Figure 1 As shown, a method for locating and tracking sightseeing vehicles based on the BeiDou satellite system includes the following steps: S1. Combining the Beidou satellite positioning module, vehicle-mounted multi-sensor, environmental vision acquisition module and IMU inertial module, collect the sightseeing vehicle's positioning, operating status, environmental vision and inertial motion data, and construct a multi-source raw dataset through time stamp synchronization calibration; S2. Preprocess the multi-source raw dataset output by S1 to complete the coordinate transformation of Beidou data, temperature drift compensation of vehicle sensor, semantic target extraction of visual data, and output a standardized multi-source dataset. S3. Based on the standardized multi-source dataset output by S2, calculate the dynamic confidence weight of each acquisition module, construct an adaptive weighted Block IEKF block iterative filtering model, complete the fusion calculation of multi-source data, and output the fusion positioning result. S4. Based on the fusion positioning results output by S3, initialize the trajectory tracking state, calculate the association cost by jointly calculating motion consistency and position intersection-union ratio, and use the OCSORT algorithm to complete the trajectory association matching and update optimization, and output a smooth optimized trajectory. S5. Based on the smoothed optimized trajectory output by S4 and the semantic segmentation data output by S2, determine the GNSS denial environment state, construct a semantic topology feature vector and match it with the pre-constructed scenic area semantic topology map library, and perform weighted fusion of reference coordinates and optimized trajectory to complete the positioning correction, and output the corrected positioning result. S6. Based on the corrected positioning results output by S5, construct a time series deviation dataset, train the NARX neural network model, predict the positioning deviation through the model and complete the compensation correction, and output the final positioning and tracking results. S7. Based on the final positioning and tracking results output by S6, perform structuring and format standardization processing on the positioning data, upload the standardized data to the cloud platform through the NB-IoT communication module, determine the abnormal operating status of the sightseeing vehicle in combination with the preset safety threshold and trigger offline voice warning, and output cloud-stored data and local warning control signals. S8, based on the cloud storage data output by S7, generates a sightseeing bus route map through route point fitting and curve optimization. It collects real-time data on the entire sightseeing bus operation process to update the deviation dataset and the scenic area semantic topology map library. It optimizes the parameters of the NARX neural network model and the adaptive weighted Block IEKF filter model online to achieve dynamic iterative optimization of the positioning model and feature library.
[0012] In step S1, the Beidou positioning module installed on the roof of the sightseeing vehicle is used to collect the sightseeing vehicle's positioning data. , They represent the first The longitude, latitude, altitude, positioning accuracy factor, and satellite signal-to-noise ratio of the sub-samples. Indicates the total number of samples; The sightseeing vehicle's operating status is collected using onboard multiple sensors. , They represent the first The driving speed, steering angle, and roll angle were sampled in the next sampling. Images of the driving environment are acquired using an environmental vision acquisition module. , Indicates the first The environmental image was sampled subsampled. Inertial motion data is collected using an IMU inertial module. , Indicates the first Triaxial angular acceleration of the next sample, Indicates the first Triaxial acceleration of the next sample; The original multi-source dataset is formed after synchronization using timestamps. .
[0013] In step S2, the sightseeing vehicle positioning data is... The coordinates were converted to the ECEF coordinate system, and abnormal positioning data were removed based on the positioning accuracy factor and satellite signal-to-noise ratio threshold. Missing data were then filled in using linear interpolation to obtain standardized sightseeing vehicle positioning data. ,in, These represent the standardized longitude, latitude, and altitude, respectively. Represents coordinates in the ECEF coordinate system; The six-sided calibration method is used to eliminate inertial motion data. Zero bias and scale factor errors were eliminated, and the data was synchronized with BeiDou data using timestamps to obtain standardized inertial motion data. ,in, This represents the standardized triaxial angular acceleration. This represents the standardized triaxial acceleration; The operating status of the sightseeing vehicle The driving speed is filtered using a moving average to eliminate noise, and temperature compensation is applied to the steering angle and roll angle to obtain the standardized operating status of the sightseeing vehicle. , These represent the standardized driving speed, steering angle, and sideslip angle, respectively. Using the PAN semantic segmentation network to analyze driving environment images Segmentation is performed, semantic categories are identified, and the center coordinates of each category are extracted. Obtain semantic segmentation data , Indicates the first In the second sampled driving environment image, the first Category labels for each semantic target Indicates the total number of category labels; Output standardized multi-source dataset .
[0014] Step S3 specifically includes the following steps: S31. Calculate weighting factors based on the reliability of data from each sensor: ; in, ; and ; In the formula, Indicates the first Normalized weighting factors for sensor-like devices. These represent the BeiDou positioning module, the IMU inertial module, and the vehicle-mounted multi-sensor, respectively. Indicates the first Reliability of sensor types; , and These respectively represent the reliability of the sightseeing vehicle's positioning data, inertial motion data, and operating status; Indicates the maximum signal-to-noise ratio; This represents the variance of the three-axis angular acceleration; Represents a constant (to prevent extremely small values where the denominator is 0); Indicates the variance of driving speed; S32. Construct an adaptive weighted Block IEKF block iterative filtering model; The state equation expression for the adaptive weighted Block IEKF iterative filtering model is as follows: ; in, ; In the formula, Indicates the first The state vector of a frame; Indicates the first Frame state prediction value; Represents the state transition function. Represents the input vector; Indicates process noise; The observation equation is expressed as follows: ; and, ; In the formula, Represents the observation vector; Represents the observation matrix; Indicates observation noise; Indicates the first The coordinates of the sightseeing vehicle in the ECEF coordinate system; The filter update expression is as follows: ; in, ; In the formula, Indicates the first Frame based on the State prediction values from frame observation results; Represents the state transition function. Indicates the first The state estimate of the frame after observation and update. Indicates the first The input vector of the frame; Indicates the first The covariance matrix of the frame prediction state; Represents the state transition Jacobian matrix; Indicates the first The state covariance matrix of a frame after observation and update; Indicates the transpose operation; Represents the process noise variance matrix; Indicates Kalman gain; Represents the observation noise variance matrix; Indicates the first The final state estimate of the frame after observation and update; Indicates the first The state covariance matrix after frame update; This represents the partial derivative of the state transition function with respect to the state vector; S33. The ECEF coordinate system state vector output by the adaptive weighted Block IEKF block iterative filtering model. Convert to WGS84 latitude and longitude coordinates And calculate the positioning accuracy. : ; ; in, ; In the formula, Indicates the filtered ECEF coordinates; Indicates the radius of curvature of the zonal loop; This represents the first eccentricity of the Earth's ellipsoid; Represents the diagonal elements of the state covariance matrix in the ECEF coordinate system; Represents the semi-major axis of the WGS84 ellipsoid; S34. Output the fused localization results ,and .
[0015] Step S4 specifically includes the following steps: S41. Based on the fused positioning data from the previous 5 frames, initialize the trajectory tracker state and output the initial trajectory state. , These represent the unique identifier of the trajectory, the mean initial position of the trajectory, the mean initial velocity of the trajectory, and the initial covariance matrix of the trajectory, respectively. S42, OCSORT trajectory association and update; S421. Calculate the current frame location point. With historical trajectory Matching cost : ; in, ; In the formula, Indicates the crossover ratio (CRO) weighting coefficient; This represents the intersection-union ratio (IoU) between the current frame's bounding box and the historical trajectory bounding boxes. Indicates the current frame location point With historical trajectory The cost of motion consistency; Indicates the current frame location point speed; Representing historical trajectory The average speed; Indicates the maximum speed of the sightseeing vehicle; S422, Based on all Constructing the cost matrix of current frame localization point-historical trajectory ; S423. Solving the cost matrix using the Hungarian algorithm. Minimum weight matching yields the matching pair. And based on the matching pairs The following determinations are made: like If the match is successful, the current frame location point is determined. Belonging to the historical trajectory Otherwise, if the current frame's location point is determined to be a new trajectory, and there is no match for 3 consecutive frames, then the trajectory is determined to have disappeared. S424. Based on the successfully matched current frame location point and historical trajectory state, an observation center momentum mechanism is introduced, and the trajectory state is updated using Kalman filtering to output a smooth trajectory. ; The Kalman filter update expression is as follows: ; ; ; In the formula, Indicates the first Before the frame, the historical trajectory The predicted state covariance matrix; Indicates the first Frame, historical trajectory The state covariance matrix updated after observation; Represents the observation matrix; Indicates the variance of observation noise; Indicates the first Frame, historical trajectory The updated state estimate after observation; Indicates the first Before the frame, the historical trajectory The predicted state estimate; Indicates the first The frame is the observation value after being corrected by the observation center momentum mechanism, and , Indicates the weight of the current observation. This represents the original observation value of the current frame. , This represents the corrected observation value from the previous frame.
[0016] Step S5 specifically includes the following steps: S51. Semantic topological feature extraction; Based on semantic segmentation data Extracting the distance ratio of semantic targets (Ratio of the distance from the centroid of the semantic target to the image center to the farthest distance), angular difference (The angle difference between the line connecting adjacent semantic targets and the image center) and semantic labels Construct semantic topological feature vectors , Indicates the first The first frame of the image The distance ratio of each semantic target; Indicates the first In the frame image, the first The and the first One semantic goal; Indicates the first The first frame of the image Category labels for each semantic target; S52, Positioning accuracy based on optimized trajectory and the signal-to-noise ratio of BeiDou satellite signals Determine if it is a GNSS denied environment: when and When the environment is determined to be GNSS denied, semantic-assisted correction is triggered, and the corrected positioning result is output. Otherwise, directly output the smooth trajectory. , They represent the first Frames, including the latitude, longitude, altitude, speed, and positioning accuracy of the smooth trajectory; The semantic-assisted correction strategy is as follows: First, a semantic topological map library for scenic areas is pre-built. , This indicates that in the pre-built semantic topological map library of scenic spots, the first... Semantic topological feature vectors of reference points In the semantic topological map database of scenic areas, the first... WGS84 latitude and longitude reference coordinates of each reference point; Then, the semantic topological features of the current frame are calculated. Euclidean distance to features in the map database Find the optimal match ; Finally, based on the reference coordinates of the matched items and the current trajectory position, a weighted fusion method is used to correct the localization result. : ; in, ; ; In the formula, They represent the first The latitude, longitude, altitude, speed, and positioning accuracy of the frame after semantic correction; and These represent the weight coefficients of the smooth trajectory and the weight coefficients of the semantic topology matching reference points, respectively. , , , , These represent the latitude, longitude, altitude, scene speed limit, and positioning accuracy of the semantic matching reference point, respectively.
[0017] Step S6 specifically includes the following steps: S61. Construct the deviation dataset; S611. Calculate the deviation between the corrected positioning result and the BeiDou static positioning reference value: In the formula, , , , and They represent the first The frame corrects the latitude deviation, longitude deviation, altitude deviation, speed deviation, and positioning accuracy deviation between the positioning result and the BeiDou high-precision reference value. , , , and They represent the first The frame contains the latitude, longitude, altitude, speed, and positioning accuracy of the BeiDou static positioning reference values. S612. Select the positioning deviation of the first 20 frames. Smooth trajectory velocity and smooth trajectory steering angle are used as input features. Current frame deviation As output labels Build a training dataset , This represents a pair of samples in the training dataset. This represents the total number of samples in the biased dataset. This indicates the number of historical frames corresponding to the input features; in, ; They represent from Frame to The latitude deviation sequence, longitude deviation sequence, altitude deviation sequence, driving speed deviation sequence, and positioning accuracy deviation sequence of the frame; Indicates from Frame to A smooth trajectory velocity sequence of the frame; Indicates from Frame to The smooth trajectory steering angle sequence of the frame. ; S62. Divide the training dataset into training set, validation set and test set, and train the NARX neural network model using the Levenberg-Marquardt algorithm with MSE as the loss function, iterating until the validation set loss converges. The model prediction expression is as follows: ; In the formula, , , , and They represent the first The frame contains latitude deviation, longitude deviation, altitude deviation, driving speed deviation, and positioning accuracy deviation predicted by the NARX neural network model. , , , , Both represent nonlinear mapping functions; S63. Use the trained NARX neural network model to predict the positioning deviation of the current frame, compensate for the corrected positioning result, and obtain the final positioning and tracking result. ; in, ; In the formula, They represent the first The final positioning latitude, final positioning longitude, final altitude, final driving speed, and final positioning longitude of the frame; This indicates the deviation threshold.
[0018] Step S7 specifically includes the following steps: S71. Digital Format Standardization: Standardize the final location tracking results. Convert to JSON format to obtain standardized transmission data. ; S72. Establish a communication link with the OneNet cloud platform through the BC20 NB-IoT module, and use the UDP protocol to transmit standardized data. It uploads data in real time, with the transmission frequency matching the sampling frequency, and receives control commands from the cloud. During this process, local anomaly warnings are continuously detected: when or or When this occurs, it is determined to be an abnormal state, and the SU-03T1 offline voice module is triggered to play a warning message.
[0019] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for positioning and tracking sightseeing vehicles based on the BeiDou satellite system, characterized in that: Includes the following steps: S1. Combining the Beidou satellite positioning module, vehicle-mounted multi-sensor, environmental vision acquisition module and IMU inertial module, collect the sightseeing vehicle's positioning, operating status, environmental vision and inertial motion data, and construct a multi-source raw dataset through time stamp synchronization calibration; S2. Preprocess the multi-source raw dataset output by S1 to complete the coordinate transformation of Beidou data, temperature drift compensation of vehicle sensor, semantic target extraction of visual data, and output a standardized multi-source dataset. S3. Based on the standardized multi-source dataset output by S2, calculate the dynamic confidence weight of each acquisition module, construct an adaptive weighted Block IEKF block iterative filtering model, complete the fusion calculation of multi-source data, and output the fusion positioning result. S4. Based on the fusion positioning results output by S3, initialize the trajectory tracking state, calculate the association cost by jointly calculating motion consistency and position intersection-union ratio, and use the OCSORT algorithm to complete the trajectory association matching and update optimization, and output a smooth optimized trajectory. S5. Based on the smoothed optimized trajectory output by S4 and the semantic segmentation data output by S2, determine the GNSS denial environment state, construct a semantic topology feature vector and match it with the pre-constructed scenic area semantic topology map library, and perform weighted fusion of reference coordinates and optimized trajectory to complete the positioning correction, and output the corrected positioning result. S6. Based on the corrected positioning results output by S5, construct a time series deviation dataset, train the NARX neural network model, predict the positioning deviation through the model and complete the compensation correction, and output the final positioning and tracking results. S7. Based on the final positioning and tracking results output by S6, perform structuring and format standardization processing on the positioning data, upload the standardized data to the cloud platform through the NB-IoT communication module, determine the abnormal operating status of the sightseeing vehicle in combination with the preset safety threshold and trigger offline voice warning, and output cloud-stored data and local warning control signals. S8, based on the cloud storage data output by S7, generates a sightseeing bus route map through route point fitting and curve optimization. It collects real-time data on the entire sightseeing bus operation process to update the deviation dataset and the scenic area semantic topology map library. It optimizes the parameters of the NARX neural network model and the adaptive weighted Block IEKF filter model online to achieve dynamic iterative optimization of the positioning model and feature library.
2. The sightseeing vehicle positioning and tracking method based on BeiDou satellite according to claim 1, characterized in that: In step S1, the Beidou positioning module installed on the roof of the sightseeing vehicle is used to collect the sightseeing vehicle's positioning data. , They represent the first The longitude, latitude, altitude, positioning accuracy factor, and satellite signal-to-noise ratio of the sub-samples. Indicates the total number of samples; The sightseeing vehicle's operating status is collected using onboard multiple sensors. , They represent the first The driving speed, steering angle, and roll angle were sampled in the next sampling. Images of the driving environment are acquired using an environmental vision acquisition module. , Indicates the first The environmental image was sampled subsampled. Inertial motion data is collected using an IMU inertial module. , Indicates the first Triaxial angular acceleration of the next sample, Indicates the first Triaxial acceleration of the next sample; The original multi-source dataset is formed after synchronization using timestamps. .
3. The sightseeing vehicle positioning and tracking method based on BeiDou satellite according to claim 2, characterized in that: In step S2, the sightseeing vehicle positioning data is... The coordinates were converted to the ECEF coordinate system, and abnormal positioning data were removed based on the positioning accuracy factor and satellite signal-to-noise ratio threshold. Missing data were then filled in using linear interpolation to obtain standardized sightseeing vehicle positioning data. ,in, These represent the standardized longitude, latitude, and altitude, respectively. Represents coordinates in the ECEF coordinate system; The six-sided calibration method is used to eliminate inertial motion data. Zero bias and scale factor errors were eliminated, and the data was synchronized with BeiDou data using timestamps to obtain standardized inertial motion data. ,in, This represents the standardized triaxial angular acceleration. This represents the standardized triaxial acceleration; The operating status of the sightseeing vehicle The driving speed is filtered using a moving average to eliminate noise, and temperature compensation is applied to the steering angle and roll angle to obtain the standardized operating status of the sightseeing vehicle. , These represent the standardized driving speed, steering angle, and sideslip angle, respectively. Using the PAN semantic segmentation network to analyze driving environment images Segmentation is performed, semantic categories are identified, and the center coordinates of each category are extracted. Obtain semantic segmentation data , Indicates the first In the second sampled driving environment image, the first Category labels for each semantic target Indicates the total number of category labels; Output standardized multi-source dataset .
4. The sightseeing vehicle positioning and tracking method based on BeiDou satellite according to claim 3, characterized in that: Step S3 specifically includes the following steps: S31. Calculate weighting factors based on the reliability of data from each sensor: ； in, ； and ; In the formula, Indicates the first Normalized weighting factors for sensor-like devices. These represent the BeiDou positioning module, the IMU inertial module, and the vehicle-mounted multi-sensor, respectively. Indicates the first Reliability of sensor types; , and These respectively represent the reliability of the sightseeing vehicle's positioning data, inertial motion data, and operating status; Indicates the maximum signal-to-noise ratio; This represents the variance of the three-axis angular acceleration; Represents a constant; Indicates the variance of driving speed; S32. Construct an adaptive weighted Block IEKF block iterative filtering model; The state equation expression for the adaptive weighted Block IEKF iterative filtering model is as follows: ； in, ； In the formula, Indicates the first The state vector of a frame; Indicates the first Frame state prediction value; Represents the state transition function. Represents the input vector; Indicates process noise; The observation equation is expressed as follows: ； and, ； In the formula, Represents the observation vector; Represents the observation matrix; Indicates observation noise; Indicates the first The coordinates of the sightseeing vehicle in the ECEF coordinate system; The filter update expression is as follows: ； in, ； In the formula, Indicates the first Frame based on the State prediction values from frame observation results; Represents the state transition function. Indicates the first The state estimate of the frame after observation and update. Indicates the first The input vector of the frame; Indicates the first The covariance matrix of the frame prediction state; Represents the state transition Jacobian matrix; Indicates the first The state covariance matrix of a frame after observation and update; Indicates the transpose operation; Represents the process noise variance matrix; Indicates Kalman gain; Represents the observation noise variance matrix; Indicates the first The final state estimate of the frame after observation and update; Indicates the first The state covariance matrix after frame update; This represents the partial derivative of the state transition function with respect to the state vector; S33. The ECEF coordinate system state vector output by the adaptive weighted Block IEKF block iterative filtering model. Convert to WGS84 latitude and longitude coordinates And calculate the positioning accuracy. : ； ； in, ； In the formula, Indicates the filtered ECEF coordinates; Indicates the radius of curvature of the zonal loop; This represents the first eccentricity of the Earth's ellipsoid; Represents the diagonal elements of the state covariance matrix in the ECEF coordinate system; Represents the semi-major axis of the WGS84 ellipsoid; S34. Output the fused localization results ,and .
5. The sightseeing vehicle positioning and tracking method based on BeiDou satellite according to claim 4, characterized in that: Step S4 Specifically, the following steps are included: S41. Based on the fused positioning data from the previous 5 frames, initialize the trajectory tracker state and output the initial trajectory state. , These represent the unique identifier of the trajectory, the mean initial position of the trajectory, the mean initial velocity of the trajectory, and the initial covariance matrix of the trajectory, respectively. S42, OCSORT trajectory association and update; S421. Calculate the current frame location point. With historical trajectory Matching cost : ； in, ； In the formula, Indicates the crossover ratio (CRO) weighting coefficient; This represents the intersection-union ratio (IoU) between the current frame's bounding box and the historical trajectory bounding boxes. Indicates the current frame location point With historical trajectory The cost of motion consistency; Indicates the current frame location point speed; Representing historical trajectory The average speed; Indicates the maximum speed of the sightseeing vehicle; S422, Based on all Constructing the cost matrix of current frame localization point-historical trajectory ; S423. Solving the cost matrix using the Hungarian algorithm. Minimum weight matching yields the matching pair. And based on the matching pairs The following determinations are made: like If the match is successful, the current frame location point is determined. Belonging to the historical trajectory Otherwise, if the current frame's location point is determined to be a new trajectory, and there is no match for 3 consecutive frames, then the trajectory is determined to have disappeared. S424. Based on the successfully matched current frame location point and historical trajectory state, an observation center momentum mechanism is introduced, and the trajectory state is updated using Kalman filtering to output a smooth trajectory. ; The Kalman filter update expression is as follows: ； ； ； In the formula, Indicates the first Before the frame, the historical trajectory The predicted state covariance matrix; Indicates the first Frame, historical trajectory The state covariance matrix updated after observation; Represents the observation matrix; Indicates the variance of observation noise; Indicates the first Frame, historical trajectory The updated state estimate after observation; Indicates the first Before the frame, the historical trajectory The predicted state estimate; Indicates the first The frame is the observation value after being corrected by the observation center momentum mechanism, and , Indicates the weight of the current observation. This represents the original observation value of the current frame. , This represents the corrected observation value from the previous frame.
6. The sightseeing vehicle positioning and tracking method based on BeiDou satellite according to claim 5, characterized in that: Step S5 specifically includes the following steps: S51. Semantic topological feature extraction; Based on semantic segmentation data Extracting the distance ratio of semantic targets , angle difference and semantic tags Construct semantic topological feature vectors , Indicates the first The first frame of the image The distance ratio of each semantic target; Indicates the first In the frame image, the first The and the first One semantic goal; Indicates the first The first frame of the image Category labels for each semantic target; S52, Positioning accuracy based on optimized trajectory and the signal-to-noise ratio of BeiDou satellite signals Determine if it is a GNSS denied environment: when and When the environment is determined to be GNSS denied, semantic-assisted correction is triggered, and the corrected positioning result is output. Otherwise, directly output the smooth trajectory. , They represent the first Frames, including the latitude, longitude, altitude, speed, and positioning accuracy of the smooth trajectory; The semantic-assisted correction strategy is as follows: First, a semantic topological map library for scenic areas is pre-built. , This indicates that in the pre-built semantic topological map library of scenic spots, the first... Semantic topological feature vectors of reference points In the semantic topological map database of scenic areas, the first... WGS84 latitude and longitude reference coordinates of each reference point; Then, the semantic topological features of the current frame are calculated. Euclidean distance to features in the map database Find the optimal match ; Finally, based on the reference coordinates of the matched items and the current trajectory position, a weighted fusion method is used to correct the localization result. : ； in, ； ； In the formula, They represent the first The latitude, longitude, altitude, speed, and positioning accuracy of the frame after semantic correction; and These represent the weight coefficients of the smooth trajectory and the weight coefficients of the semantic topology matching reference points, respectively. , , , , These represent the latitude, longitude, altitude, scene speed limit, and positioning accuracy of the semantic matching reference point, respectively.
7. A sightseeing vehicle positioning and tracking method based on BeiDou satellite as described in claim 6, characterized in that: Step S6 specifically includes the following steps: S61. Construct the deviation dataset; S611. Calculate the deviation between the corrected positioning result and the BeiDou static positioning reference value: In the formula, , , , and They represent the first The frame corrects the latitude deviation, longitude deviation, altitude deviation, speed deviation, and positioning accuracy deviation between the positioning result and the BeiDou high-precision reference value. , , , and They represent the first The frame contains the latitude, longitude, altitude, speed, and positioning accuracy of the BeiDou static positioning reference values. S612. Select the positioning deviation of the first 20 frames. Smooth trajectory velocity and smooth trajectory steering angle are used as input features. Current frame deviation As output labels Build a training dataset , This represents a pair of samples in the training dataset. This represents the total number of samples in the biased dataset. This indicates the number of historical frames corresponding to the input features; in, ； They represent from Frame to The latitude deviation sequence, longitude deviation sequence, altitude deviation sequence, driving speed deviation sequence, and positioning accuracy deviation sequence of the frame; Indicates from Frame to A smooth trajectory velocity sequence of the frame; Indicates from Frame to The smooth trajectory steering angle sequence of the frame. ; S62. Divide the training dataset into training set, validation set and test set, and train the NARX neural network model using the Levenberg-Marquardt algorithm with MSE as the loss function, iterating until the validation set loss converges. The model prediction expression is as follows: ； In the formula, , , , and They represent the first The frame contains latitude deviation, longitude deviation, altitude deviation, driving speed deviation, and positioning accuracy deviation predicted by the NARX neural network model. , , , , Both represent nonlinear mapping functions; S63. Use the trained NARX neural network model to predict the positioning deviation of the current frame, compensate for the corrected positioning result, and obtain the final positioning and tracking result. ; in, ； In the formula, They represent the first The final positioning latitude, final positioning longitude, final altitude, final driving speed, and final positioning longitude of the frame; This indicates the deviation threshold.
8. A sightseeing vehicle positioning and tracking method based on BeiDou satellite according to claim 7, characterized in that: Step S7 specifically includes the following steps: S71. Digital Format Standardization: Standardize the final location tracking results. Convert to JSON format to obtain standardized transmission data. ; S72. Establish a communication link with the OneNet cloud platform through the BC20 NB-IoT module, and use the UDP protocol to transmit standardized data. It uploads data in real time, with the transmission frequency matching the sampling frequency, and receives control commands from the cloud. During this process, local anomaly warnings are continuously detected: when or or When this occurs, it is determined to be an abnormal state, and the SU-03T1 offline voice module is triggered to play a warning message.