Multi-source information fusion method, system, device and medium for pedestrian positioning

Abstract

The present application relates to navigation positioning technical field, especially to a kind of multi-source information fusion method, system, equipment and medium for pedestrian positioning, method includes synchronous acquisition multi-source information data, according to multi-source information data calculates the confidence weight of each sub-filter, and calculates dynamic noise adaptive adjustment factor and updates each sub-filter, through two threshold values and consistency check multi-dimensional fault detection is carried out and through complementary compensation mechanism deletes fault data source, using other data source collaborative compensation error, according to building structured information data pre-judge pedestrian position, to effective sub-filter is modified and calibrated, according to confidence weight carries out dynamic information distribution, through suboptimality correction factor correction information loss, obtains global optimal estimation.The present application greatly improves pedestrian positioning accuracy, robustness and expansion efficiency under complex environment, solves the technical bottleneck that traditional multi-source fusion technology adapts poorly, fault tolerance is weak, utilization is shallow, expansion is difficult.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of navigation positioning, and in particular to a multi-source information fusion method, system, device and medium for pedestrian positioning. BACKGROUND

[0002] The existing technical solutions in the field of pedestrian positioning are mainly divided into single sensor positioning and multi-source sensor fusion. The single sensor positioning scheme relies on a single sensing unit to realize positioning and cannot cope with multi-source interference in complex environments, and the positioning robustness is insufficient. The existing multi-source fusion scheme mainly adopts the combination of an inertial measurement unit (MIMU) and auxiliary sensors such as a magnetic vector sensor and an ultrasonic range finder, but it is difficult to meet the high-precision positioning requirements in complex scenarios in actual application.

[0003] The existing multi-source fusion scheme only focuses on the simple superposition of sensor data, resulting in problems such as penetration of walls and path deviation in the positioning result, especially in complex indoor scenarios where the positioning accuracy is difficult to guarantee. The weights of various sensors are mostly pre-set fixed values, or are simply assigned based on a single dimension of sensor accuracy. When some sensors are disturbed and data anomalies occur, the fixed weight distribution method cannot reduce the influence of abnormal data, and the fusion accuracy is easily significantly reduced. The single threshold judgment method is used for fault detection, which is easy to cause fault misjudgment or omission. After identifying the faulty sensor, the data is directly discarded, resulting in the loss of positioning dimension information corresponding to the faulty data source, and further affecting the continuity and reliability of positioning. In the process of multi-source sensor information distribution and fusion, the existing scheme does not consider the inherent information loss problem when information is interacted between sub-filters, lacks an effective correction mechanism, and the final fusion result has a sub-optimal deviation, making it difficult to achieve global optimal estimation. The existing pedestrian positioning technology has the defects of insufficient positioning accuracy, poor robustness, weak environmental adaptability, limited fault tolerance, etc., whether it is a single sensor scheme or a traditional multi-source fusion scheme, and it is difficult to meet the demand for high-precision, high-reliability real-time positioning of pedestrian positions in complex building environments, emergency rescue and other scenarios. SUMMARY

[0004] The present application aims to at least solve one of the technical problems in the related art. To this end, the present application provides a multi-source information fusion method, system, device and medium for pedestrian positioning, which greatly improves the pedestrian positioning accuracy, robustness and expansion efficiency in complex environments, and solves the technical bottlenecks of poor adaptation, weak fault tolerance, shallow utilization and difficult expansion of traditional multi-source fusion technology.

[0005] The present application provides a multi-source information fusion method for pedestrian positioning, comprising:

[0006] S1: synchronously acquiring micro-inertial measurement unit data, heading data of a magnetic vector sensor, distance measured by an ultrasonic range finder from an obstacle, and building structured information data;

[0007] S2: calculating confidence weights of each sub-filter in real time according to the micro-inertial measurement unit data, the heading data of the magnetic vector sensor, the distance measured by the ultrasonic range finder from the obstacle, and the building structured information data;

[0008] S3: calculating a dynamic noise adaptive adjustment factor according to the confidence weights of each sub-filter, and updating each sub-filter through the dynamic noise adaptive adjustment factor;

[0009] S4: performing multi-dimensional fault detection on the updated each sub-filter through double thresholds and consistency checking, identifying a faulty sub-filter, deleting a faulty data source through a complementary compensation mechanism, and using remaining data sources to compensate for errors caused by the faulty data source, to obtain an effective sub-filter;

[0010] S5: predicting a pedestrian position according to the micro-inertial measurement unit data and the building structured information data;

[0011] S6: correcting and calibrating the effective sub-filter according to the predicted pedestrian position and the building structured information data, and performing dynamic information distribution on the corrected and calibrated effective sub-filter according to the confidence weights;

[0012] S7: correcting information loss in the dynamic information distribution of the effective sub-filter through a suboptimality correction factor, to obtain a globally optimal estimation.

[0013] Further, a calculation expression of the confidence weights of each sub-filter is:

[0014]

[0015] wherein, is a confidence weight of an i-th sub-filter at a t-th moment, is a sensor accuracy of the i-th sub-filter, is a data continuity marker of the i-th sub-filter at the t-th moment, is a historical fusion error of the i-th sub-filter at the t-th moment, is a minimum constant. Further, the S4 step includes: S41: calculating a measurement residual of each sub-filter; S42: performing multi-dimensional fault detection on each sub-filter through double thresholds and consistency checking, identifying a faulty sub-filter, and deleting a faulty data source through a complementary compensation mechanism;

[0016] Further, the S4 step includes:

[0017] S41: calculating a measurement residual of each sub-filter; ​​​​

[0018] S42: setting a dynamic threshold according to the confidence weight of each sub-filter;

[0019] S43: if the absolute value of the measurement residual is greater than the dynamic threshold and the data continuity is 1, performing consistency check;

[0020] The consistency check includes:

[0021] calculating the deviation of the filtering result of each sub-filter from the global estimate;

[0022] If the deviation of the filtering result of the sub-filter from the global estimate is greater than the scene adaptation threshold, the sub-filter is faulty.

[0023] Further, the S5 step includes:

[0024] S51: performing semantic analysis on the building structured information to disassemble it into wall position, passage width, door and window position, and occlusion probability;

[0025] S52: predicting the pedestrian position according to the wall position, passage width, door and window position, occlusion probability, and micro-inertial measurement unit data.

[0026] Further, the effective sub-filter is modified and calibrated according to the predicted pedestrian position and the building structured information data, including:

[0027] S61: filtering abnormal measurement of ultrasonic ranging using the wall position;

[0028] If the deviation of the ultrasonic ranging measurement value from the theoretical distance of the wall position exceeds the deviation threshold, the ultrasonic ranging is abnormal measurement, and the measurement weight of the ultrasonic ranging is reduced;

[0029] S62: compensating for measurement accuracy degradation using the occlusion probability;

[0030] In an occlusion scene, the deviation of the predicted pedestrian position from the measurement value is calculated;

[0031] The measurement equation is updated according to the deviation of the predicted pedestrian position from the measurement value and the occlusion compensation term, and the occlusion compensation term includes the occlusion probability and the measurement mutation correction value corresponding to the door and window position;

[0032] S63: every time a door and window position is passed, the door and window position coordinates are taken as a fixed anchor point, a calibration threshold is set in combination with the occlusion probability, and the deviation of the current positioning at the fixed anchor point is corrected according to the calibration threshold.

[0033] Further, the effective sub-filter after modification and calibration is dynamically allocated information according to the confidence weight to meet the confidence-oriented information conservation, and the calculation expression is:

[0034]

[0035]

[0036] in, for The initial covariance matrix of global fusion at each time step. for The posterior covariance matrix at time t. For the first Sub-filters Confidence weight at each time point.

[0037] Furthermore, by correcting for information loss in the dynamic information allocation of the effective sub-filters through a suboptimal correction factor, the calculation expression for the globally optimal estimate is obtained as follows:

[0038]

[0039] in, for The global optimal estimate at time t. For the first Sub-filters Confidence weight at each moment for Suboptimal correction factor at time, For the first Sub-filters The posterior covariance matrix at time t. for The initial covariance matrix of global fusion at each time step. To find the inverse matrix.

[0040] The present invention also provides a multi-source information fusion system for pedestrian positioning, for performing the above-described multi-source information fusion method for pedestrian positioning, comprising:

[0041] The acquisition unit simultaneously acquires data from the micro inertial measurement unit, heading data from the magnetic vector sensor, distances to obstacles measured by the ultrasonic rangefinder, and structural information data of the building.

[0042] The confidence weight calculation unit calculates the confidence weight of each sub-filter in real time based on the data from the micro inertial measurement unit, the heading data from the magnetic vector sensor, the distance to the obstacle measured by the ultrasonic rangefinder, and the building structure information data.

[0043] A noise adjustment unit calculates a dynamic noise adaptive adjustment factor based on the confidence weight of each sub-filter, and updates each sub-filter using the dynamic noise adaptive adjustment factor.

[0044] A fault detection unit performs multi-dimensional fault detection on the updated sub-filters through double thresholds and consistency checking, identifies faulty sub-filters, deletes faulty data sources through a complementary compensation mechanism, compensates errors caused by the faulty data sources using the remaining data sources, and obtains effective sub-filters;

[0045] A pedestrian position prediction unit predicts a pedestrian position according to micro-inertial measurement unit data and building structured information data;

[0046] A dynamic information distribution unit corrects and calibrates the effective sub-filters according to the predicted pedestrian position and the building structured information data, and dynamically distributes the corrected and calibrated effective sub-filters according to confidence weights;

[0047] A correction unit corrects information loss in the dynamic information distribution of the effective sub-filters through a suboptimality correction factor, and obtains a globally optimal estimate.

[0048] The application further provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the multi-source information fusion method for pedestrian positioning according to any one of the above when executing the program.

[0049] The application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and the computer program implements the steps of the multi-source information fusion method for pedestrian positioning according to any one of the above when executed by a processor.

[0050] The one or more technical solutions described above in the embodiments of the application have at least one of the following technical effects:

[0051] The application integrates confidence evaluation and building feature information analysis into federated Kalman filtering, realizes the upgrade from passive fusion to active adaptation, breaks through the limitations of traditional fault removal through a multi-dimensional fault self-healing mechanism, and improves the robustness of the system in complex environments; through deviation compensation and suboptimality correction, the pedestrian positioning accuracy, robustness and expansion efficiency in complex environments are greatly improved while maintaining low computational complexity, meeting the needs of high-precision, high-reliability and strong expansion of pedestrian positioning, and solving the technical bottlenecks of poor adaptation, weak fault tolerance, shallow utilization and difficult expansion of traditional multi-source fusion technology.

[0052] Additional aspects and advantages of the application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described in the following description are some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without any creative effort.

[0054] Figure 1 is a flow diagram of a multi-source information fusion method for pedestrian positioning provided by the present application.

[0055] Figure 2 is a structural diagram of a multi-source information fusion system for pedestrian positioning provided by the present application.

[0056] Figure 3 is a block diagram of an electronic device provided by the present application.

[0057] Reference signs:

[0058] 101, acquisition unit; 102, confidence weight calculation unit; 103, noise adjustment unit; 104, fault detection unit; 105, pedestrian position pre-judgment unit; 106, dynamic information distribution unit; 107, correction unit; 201, processor; 202, communication bus; 203, communication interface; 204, memory. DETAILED DESCRIPTION

[0059] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort belong to the scope of protection of the present application. The following embodiments are used to illustrate the present application, but cannot be used to limit the scope of the present application.

[0060] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine the different embodiments or examples described in the present application and the features of the different embodiments or examples without contradiction.

[0061] The following will be described in combination withFigures 1 to 3 This invention describes a multi-source information fusion method, system, device, and medium for pedestrian positioning.

[0062] like Figure 1 As shown, a multi-source information fusion method for pedestrian localization includes:

[0063] S1: Simultaneously acquire data from the micro inertial measurement unit (MIMU), magnetic vector sensor, ultrasonic rangefinder, and building structure information;

[0064] It covers four core data sources: MIMU, magnetic vector sensor, ultrasonic rangefinder, and building structured information, making up for the limitations of traditional technologies that rely on a single data source or have insufficient data types. Synchronous acquisition ensures that the data sources are time-aligned, avoiding fusion errors caused by data delays or asynchrony.

[0065] In some specific embodiments of the present invention, the MIMU is used as a common reference system, and four independent sub-filters are set up: MIMU / zero velocity correction, MIMU / magnetometer, MIMU / ultrasonic rangefinder, and MIMU / building feature information. Each sub-filter operates independently.

[0066] This invention uses federated Kalman filtering without reset feedback for filtering; based on the original system's federated filtering architecture, a confidence assessment module and a building feature information analysis module are added to form a closed-loop architecture of perception-assessment-fusion-self-healing-expansion.

[0067] Confidence assessment module: Calculates the confidence weight of each sub-filter in real time, serving as the core basis for information allocation;

[0068] The building feature information parsing module decomposes the building structure information into building feature information, including wall location, passage width, door and window location, and occlusion probability. Combined with the MIMU motion state, it realizes the full-process adaptation of "prediction-constraint-compensation".

[0069] No. The system state vector of each sub-filter ;

[0070] Measurement vector : For MIMU / zero velocity calibration measurement vector, For MIMU / magnetometer measurement vector, For MIMU / ultrasonic rangefinder to measure vectors, For MIMU / building feature information measurement vectors, the first Sub-filters Time-based data continuity markers ,like If the data is continuous, then the data is continuous. , then the data jumps;

[0071] The error covariance matrix is the correlation item of the confidence weight, that is:

[0072]

[0073] wherein, is the confidence weight of the first sub-filter at the time point, is the error covariance matrix of the first sub-filter at the time point, is the error covariance matrix of the first sub-filter at the time point, is the traditional error covariance matrix of the first sub-filter at the time point, is the traditional error covariance matrix of the first sub-filter at the time point, is the confidence weight of the first sub-filter at the time point, is the confidence weight of the first sub-filter at the time point, is the confidence weight of the first sub-filter at the time point, is the confidence weight of the first sub-filter at the time point, is the inverse matrix.

[0074] The application takes the MIMU as a public reference system, each sub-filter (MIMU / zero-speed correction, MIMU / magnetometer, MIMU / ultrasonic range finder, MIMU / architectural feature information) maintains independent work, cooperates with the main filter through the newly added module, takes into account fault tolerance and expansibility, and solves the technical defects of the traditional architecture.

[0075] S2: real-time calculation of the confidence weight of each sub-filter according to the micro inertial measurement unit data, the heading data of the magnetic vector sensor, the distance measured by the ultrasonic range finder and the architectural structured information data;

[0076] The confidence evaluation module combines the sensor accuracy, data continuity and historical fusion error to calculate the confidence weight of each sub-filter, and the calculation expression of the confidence weight of each sub-filter is:

[0077]

[0078] wherein, is the confidence weight of the first sub-filter at the time point, is the confidence weight of the first sub-filter at the time point, is the sensor accuracy of the first sub-filter, is the sensor accuracy of the first sub-filter, is the data continuity mark of the first sub-filter at the time point, is the data continuity mark of the first sub-filter at the time point, is the historical fusion error of the first sub-filter at the time point, is the historical fusion error of the first sub-filter at the time point, is the historical fusion error of the first sub-filter at the time point, is the historical fusion error of the first sub-filter at the time point, is the historical fusion error of the first sub-filter at the time point, is the minimum constant;

[0079] In some specific embodiments of the application, the value of is [0, -1], , To avoid the denominator being 0.

[0080] S3: Calculate the dynamic noise adaptive adjustment factor based on the confidence weight of each sub-filter, and update each sub-filter using the dynamic noise adaptive adjustment factor;

[0081] By using a dynamic noise adaptive adjustment factor, noise accumulation in complex environments is suppressed. The calculation expression is as follows:

[0082]

[0083] in, For the first The system state vector of each sub-filter is in Prior state estimation at time 10:00 For the first Sub-filters in The prior state transition matrix at time t. For the first The system state vector of each sub-filter is in Posterior state estimation at time 10:00. For the first Sub-filters in The prior error covariance matrix at time t. For the first Sub-filters in The posterior error covariance matrix at time t. For the first Sub-filters in The state transition matrix at time t, This is a dynamic noise adaptive adjustment factor. , The lower the value, the stronger the noise suppression. This is the transpose of the matrix.

[0084] S4: Perform multi-dimensional fault detection on each updated sub-filter through dual thresholds and consistency checks, identify faulty sub-filters, delete faulty data sources through complementary compensation mechanisms, and use other data sources to collaboratively compensate for the errors caused by faulty data sources to obtain effective sub-filters.

[0085] Existing fault detection technologies based solely on measurement residuals are prone to misjudgment. Multi-dimensional fault detection, employing dual thresholds and consistency checks, is employed, including:

[0086] S41: Calculate the measurement residuals of each sub-filter. The calculation expression is as follows:

[0087]

[0088] in,

[0089] S42: Set a dynamic threshold according to the confidence weight of the sub-filter, and calculate the expression as follows:

[0090]

[0091]

[0092] S43: If the absolute value of the residual is greater than the dynamic threshold and the data continuity is 1, perform consistency check.

[0093] If and , enter the consistency check.

[0094] The consistency check includes:

[0095] Calculate the deviation of the filtering result of each sub-filter from the global estimation, and calculate the expression as follows:

[0096]

[0097]

[0098] If , the​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​ Several sub-filters are faulty, among which... To adapt the threshold for the scene, Positively correlated with occlusion probability, when there is no occlusion When partially obscured When fully occluded By using a complementary compensation mechanism to remove faulty data sources and using other data sources to collaboratively compensate for the errors caused by the faulty data sources, an effective sub-filter can be obtained.

[0099] In some specific embodiments of the present invention, when the magnetic vector sensor fails due to electromagnetic interference, the attitude angle error is compensated by the building feature information (predicting the direction of the channel) and ultrasonic ranging (real-time ranging of the wall) to ensure that the main filter can still obtain a stable input.

[0100] S5: Predict pedestrian positions based on data from micro inertial measurement units and building structure information data;

[0101] Real-time acquisition of pedestrian motion data to obtain basic motion information for pedestrian position calculation; synchronous acquisition of auxiliary sensor data, including heading data from magnetic vector sensors and real-time distance measurements from ultrasonic rangefinders to obstacles such as walls;

[0102] S51: Perform semantic parsing on the building's structural information, breaking it down into wall locations, passage widths, door and window locations, and occlusion probabilities;

[0103] Obtain the building's structural information, perform semantic parsing on the structural information, and decompose it into wall location, passage width, door and window location, and occlusion probability;

[0104] S52: Predict pedestrian positions based on wall location, passage width, door and window location, obstruction probability, and micro inertial measurement unit data;

[0105] Based on the building's structural information and the velocity information measured by the MIMU, position prediction with physical constraints and occlusion adaptation is achieved. The calculation expression is as follows:

[0106]

[0107] in, for The predicted pedestrian position at any given time. For the first The system state vector of each sub-filter is in Prior state estimation at time 10:00 for Pedestrian speed at any given moment The sampling interval time. for Channel width at any given moment for The probability of occlusion at any given time. Location of the wall. The nearest valid location of the wall boundary;

[0108] Depend on The coordinate range is calculated to avoid predicting wall penetration. Real-time calculations based on building layout and sensor feedback;

[0109] Building layout: such as wall thickness and obstacle distribution; sensor feedback: such as the degree of ultrasonic signal attenuation.

[0110] At that time, there was no obstruction. At times, partial obstruction, At that time, it is completely covered.

[0111] When the pedestrian is near the door or window When the distance is ≤1.5m, predict the path switching status.

[0112] If the scene is unobstructed, increase the weight of the ultrasonic rangefinder in advance;

[0113] If the scene is obscured, the weight of the building's structural information should be increased in advance.

[0114] S6: Correct and calibrate the effective sub-filters based on the predicted pedestrian positions and building structure information data, and dynamically allocate information to the corrected and calibrated effective sub-filters according to the confidence weights;

[0115] Before updating measurements, a closed loop of "prediction-correction-calibration" is added to the wall location. Location of doors and windows Occlusion probability Working together to exert influence;

[0116] S61: Filtering out abnormal measurements in ultrasonic ranging using wall location;

[0117] If the deviation between the ultrasonic ranging measurement value and the theoretical distance to the wall location exceeds the deviation threshold, the ultrasonic ranging is considered an abnormal measurement, and the measurement weight of the ultrasonic ranging is reduced.

[0118] In some specific embodiments of the present invention

[0119] If the ultrasound measurement value and If the theoretical distance deviation exceeds the deviation threshold, it is judged as abnormal, and its measurement weight is reduced. The calculation expression is:

[0120]

[0121] in, For the ultrasonic measurement value after instant correction, is an ultrasonic measurement value adjustment weight;

[0122] In some embodiments of the present application, , the deviation threshold is 30 cm.

[0123] S62: Compensate for measurement accuracy attenuation using the occlusion probability;

[0124] Under the occlusion scenario, calculate the deviation between the predicted pedestrian position and the measurement value;

[0125] Update the measurement equation according to the deviation between the predicted pedestrian position and the measurement value and the occlusion compensation term, which includes the occlusion probability and the measurement mutation correction value corresponding to the door and window position;

[0126] Compensate for measurement accuracy attenuation using the occlusion probability ;

[0127] Under the occlusion scenario, the sensor measurement error increases, and the occlusion compensation term is added in the measurement update, the deviation between the predicted position and the measurement value is calculated, and the calculation expression is:

[0128]

[0129] wherein, is the deviation between the predicted pedestrian position and the measurement value at the instant, is the predicted pedestrian position at the instant, is the measurement value at the instant, is the measurement value at the instant, is the measurement value at the instant, The measurement update equation is corrected as:

[0130]

[0131] wherein, is the posterior state estimate of the system state vector of the i-th sub-filter at the instant,

[0132] is the prior state estimate of the system state vector of the i-th sub-filter at the instant, is the filter gain of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, is the measurement value of the i-th sub-filter at the instant, The shielding probability at the moment, The measured mutation correction value corresponding to the door and window position;

[0133] The greater the building feature information prediction deviation compensation weight is, the higher the compensation weight is, and the greater the shielding causes the measurement error.

[0134] S63: Every time a door and window position is passed, the door and window position coordinates are taken as fixed anchor points, a calibration threshold is set in combination with the shielding probability, and the deviation of the current positioning at the fixed anchor points is corrected according to the calibration threshold.

[0135] Every time a door and window position is passed , the coordinates thereof are taken as fixed anchor points, and calibration logic is optimized in combination with the shielding probability:

[0136] When there is no shielding , the calibration threshold is set to 20 cm;

[0137] When there is partial shielding , the calibration threshold is set to 30 cm, so as to avoid temporary deviation caused by shielding from affecting the calibration effectiveness and ensure that the long-term positioning drift is less than or equal to 5 cm / 100 m.

[0138] Discard the traditional fixed information allocation coefficient, and realize dynamic allocation based on the confidence weight , meet the confidence-oriented information conservation, and the calculation expression is:

[0139]

[0140]

[0141] Among them, is the initial covariance matrix globally fused at the moment, is the posterior covariance matrix at the moment, is the confidence weight of the th sub-filter at the moment.

[0142] The weight is updated in real time, so that high-precision sensors obtain a higher fusion weight;

[0143] For example, in an open space, the confidence of the magnetic vector sensor is high, and the weight is improved;

[0144] For example, in a narrow channel, the confidence of ultrasonic ranging and structured information is high, and the weight accounts for more than 60%.

[0145] S7: Correct the information loss in the dynamic information allocation of the effective sub-filter through the suboptimality correction factor to obtain a globally optimal estimate; ​​​

[0146] The suboptimal correction factor is introduced to compensate for the information loss caused by the sub-filter noise distribution, so that the global estimation is closer to the optimal, and the calculation expression is:

[0147]

[0148] Wherein, is the global optimal estimation at time t, is the confidence weight of the first sub-filter at time t, is the posterior covariance matrix of the first sub-filter at time t, is the suboptimal correction factor at time t, is the posterior covariance matrix of the first sub-filter at time t, is the initial covariance matrix of global fusion at time t, is the inverse matrix.

[0149] In some embodiments of the present application, , is the maximum value function, is adaptively adjusted by the highest confidence weight, ensuring that the correction factor does not interfere with the core fusion result.

[0150] The present application integrates confidence evaluation and building feature information analysis into federated Kalman filter, realizing the upgrade from passive fusion to active adaptation;

[0151] Through multi-dimensional fault detection of the updated sub-filters by double threshold and consistency check, the faulty sub-filters are identified and the faulty data sources are deleted through complementary compensation mechanism, and the errors caused by the faulty data sources are compensated by using other data sources, replacing the traditional single residual fault detection, combining with the building feature information to realize the self-healing process of "fault identification-dynamic compensation-reconstruction fusion", breaking through the limitations of traditional fault removal and improving the robustness of the system in complex environments;

[0152] Through deviation compensation and suboptimal correction, the pedestrian positioning accuracy, robustness and expansion efficiency in complex environments are greatly improved while maintaining low computational complexity, solving the technical bottlenecks of traditional multi-source fusion technology such as poor adaptation, weak fault tolerance, shallow utilization and difficult expansion.

[0153] As shown in Figure 2 , a multi-source information fusion system for pedestrian positioning is used to perform the above-mentioned multi-source information fusion method for pedestrian positioning, comprising:

[0154] ​​​​​​The acquisition unit 101 synchronously acquires the micro-inertial measurement unit data, the heading data of the magnetic vector sensor, the distance measured by the ultrasonic range finder to the obstacle, and the building structured information data;

[0155] The confidence weight calculation unit 102 calculates the confidence weight of each sub-filter in real time according to the micro-inertial measurement unit data, the heading data of the magnetic vector sensor, the distance measured by the ultrasonic range finder to the obstacle, and the building structured information data;

[0156] The noise adjustment unit 103 calculates a dynamic noise adaptive adjustment factor according to the confidence weight of each sub-filter, and updates each sub-filter through the dynamic noise adaptive adjustment factor;

[0157] The fault detection unit 104 performs multi-dimensional fault detection on the updated sub-filters through double thresholds and consistency checking, identifies the faulty sub-filter, deletes the faulty data source through a complementary compensation mechanism, and uses the remaining data sources to compensate for the error caused by the faulty data source, to obtain an effective sub-filter;

[0158] The pedestrian position pre-judgment unit 105 pre-judges the pedestrian position according to the micro-inertial measurement unit data and the building structured information data;

[0159] The dynamic information distribution unit 106 corrects and calibrates the effective sub-filter according to the pre-judged pedestrian position and the building structured information data, and dynamically distributes the corrected and calibrated effective sub-filter according to the confidence weight;

[0160] The correction unit 107 corrects the information loss in the dynamic information distribution of the effective sub-filter through a suboptimality correction factor, to obtain a globally optimal estimate.

[0161] Through the cooperative work of the above-mentioned modules, the confidence evaluation and the building feature information analysis are integrated into the federated Kalman filter, realizing the upgrade from passive fusion to active adaptation; through the multi-dimensional fault self-healing mechanism, the limitations of traditional fault removal are broken through, and the robustness of the system in complex environments is improved; through deviation compensation and suboptimality correction, the positioning accuracy is greatly improved while keeping low computational complexity, fully meeting the high-precision, high-reliability, and strong expansion requirements of pedestrian operations.

[0162] Figure 3 An example of a block diagram of an electronic device is shown as Figure 3As shown, the electronic device can include a processor 201, a communications interface 203, a memory 204 and a communications bus 202, wherein the processor 201, the communications interface 203 and the memory 204 complete the communication with each other through the communications bus 202. The processor 201 can invoke the logic instructions in the memory 204 to execute a multi-source information fusion method for pedestrian positioning.

[0163] In addition, the logic instructions in the memory 204 described above can be realized in the form of a software function unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0164] On the other hand, the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer readable storage medium, and the computer program includes program instructions, when the program instructions are executed by a computer, the computer can execute a multi-source information fusion method for pedestrian positioning provided by the above-mentioned methods.

[0165] In yet another aspect, the present application also provides a non-transitory computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement a multi-source information fusion method for pedestrian positioning provided by the above-mentioned methods.

[0166] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present embodiment scheme according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0167] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0168] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multi-source information fusion method for pedestrian positioning, characterized in that, Comprise: S1: synchronously acquire micro-inertial measurement unit data, magnetic vector sensor heading data, ultrasonic range finder measured distance from obstacles and building structured information data; S2: real-time calculate confidence weight of each sub-filter according to micro-inertial measurement unit data, magnetic vector sensor heading data, ultrasonic range finder measured distance from obstacles and building structured information data; S3: calculate dynamic noise adaptive adjustment factor according to confidence weight of each sub-filter, update each sub-filter through dynamic noise adaptive adjustment factor; S4: multi-dimensional fault detection is carried out on updated each sub-filter through double threshold and consistency check, fault sub-filter is identified and deleted through complementary compensation mechanism, error caused by fault data source is compensated by using remaining data source, and effective sub-filter is obtained; S5: pre-judge pedestrian position according to micro-inertial measurement unit data and building structured information data; Real-time collect pedestrian motion state data, obtain basic motion information for pedestrian position calculation; synchronously acquire auxiliary sensor data, including magnetic vector sensor heading data and ultrasonic range finder real-time measured distance from obstacles such as wall; S51: carry out semantic analysis on building structured information, and disassemble into wall position, passage width, door and window position and shielding probability; Obtain building structured information, carry out semantic analysis on building structured information, and disassemble into wall position, passage width, door and window position and shielding probability; S52: pre-judge pedestrian position according to wall position, passage width, door and window position, shielding probability and micro-inertial measurement unit data; Realize position pre-judgment with physical constraint and shielding adaptation according to building structured information combined with speed information measured by MIMU, and the calculation expression is: in, for The predicted pedestrian position at any given time. For the first The system state vector of each sub-filter is in Prior state estimation at time 10:00 for Pedestrian speed at any given moment The sampling interval time. for Channel width at any given moment for The probability of occlusion at any given time. Location of the wall. The nearest valid location of the wall boundary; by The coordinate range of the wall is calculated to avoid penetration, based on the building layout and sensor feedback real-time calculation; Building layout: such as wall thickness, obstacle distribution, sensor feedback: such as ultrasonic signal attenuation degree; no occlusion, partial occlusion, full occlusion; S6: modify and calibrate effective sub-filter according to pre-judged pedestrian position and building structured information data, and dynamically distribute information to modified and calibrated effective sub-filter according to confidence weight; Modify and calibrate effective sub-filter according to pre-judged pedestrian position and building structured information data, including: S61: filter ultrasonic ranging abnormal measurement by using wall position; If the deviation between ultrasonic ranging measurement value and theoretical distance of wall position exceeds deviation threshold, the ultrasonic ranging is abnormal measurement, and the measurement weight of ultrasonic ranging is reduced; S62: compensate measurement accuracy attenuation by using shielding probability; In a shielding scene, calculate the deviation between pre-judged pedestrian position and measurement value; Update measurement equation according to the deviation between pre-judged pedestrian position and measurement value and shielding compensation term, and the shielding compensation term includes shielding probability and measurement mutation correction value corresponding to door and window position; Utilizing occlusion probability Compensate for measurement accuracy degradation; In a shielding scene, sensor measurement error increases, shielding compensation term is added in measurement update, deviation between pre-judged position and measurement value is calculated, and the calculation expression is: wherein, is the deviation of the pedestrian prediction position at the time instant from the measured value, is the first sub-filter the measured value at the time instant; The modified measurement update equation is: in, For the first The system state vector of each sub-filter is in Posterior state estimation at time 10:

00. For the first Sub-filters Filter gain at time 10:00 This is the correction value for measurement abrupt changes corresponding to the location of doors and windows; The larger the building feature information pre-judgment deviation compensation weight is, the higher the compensation weight is, and the measurement error caused by shielding is offset. S63: every time a door and window position is passed, the door and window position coordinates are taken as fixed anchor points, a calibration threshold is set combined with shielding probability, and the deviation of current positioning to the fixed anchor points is modified according to the calibration threshold; S7: correcting the information loss in the dynamic information distribution of the effective sub-filter by a sub-optimality correction factor to obtain a globally optimal estimation.

2. The multi-source information fusion method for pedestrian positioning according to claim 1, characterized in that, The calculation expression of the confidence weight of each sub-filter is: wherein, is the confidence weight of the time instant, is the number of sub-filters, is the data continuity flag of the time instant, is the sensor accuracy of the time instant, is the number of sub-filters, is the history fusion error of the time instant, is the number of sub-filters, is the data continuity flag of the time instant, is the number of sub-filters, is the history fusion error of the time instant, is the number of sub-filters, is a very small constant.

3. The multi-source information fusion method for pedestrian positioning according to claim 1, wherein, The step S4 comprises: S41: calculating the measurement residual of each sub-filter; S42: setting a dynamic threshold according to the confidence weight of each sub-filter; S43: if the absolute value of the measurement residual is greater than the dynamic threshold and the data continuity is 1, performing consistency check; The consistency check comprises: calculating the deviation of the filtering result of each sub-filter from the global estimation; If the deviation of the filtering result of the sub-filter from the global estimation is greater than the scene adaptation threshold, the sub-filter has a fault.

4. The multi-source information fusion method for pedestrian positioning according to claim 1, characterized in that, The dynamic information distribution of the corrected and calibrated effective sub-filter according to the confidence weight satisfies the confidence-oriented information conservation, and the calculation expression is: wherein, is the initial covariance matrix at time instant is the posterior covariance matrix at time instant is the th sub-filter the confidence weight at time instant 5. The multi-source information fusion method for pedestrian positioning according to claim 1, wherein, The calculation expression of the globally optimal estimation obtained by correcting the information loss in the dynamic information distribution of the effective sub-filter by a sub-optimality correction factor is: wherein is a global optimal estimate at time is the th sub-filter a confidence weight at time is a suboptimality correction factor at time is the th sub-filter a posteriori covariance matrix at time is an initial covariance matrix for global fusion at time is the inverse matrix.

6. A multi-source information fusion system for pedestrian positioning, characterized in that, A computer program product for implementing the multi-source information fusion method for pedestrian positioning according to any one of claims 1 to 5. An acquisition unit configured to synchronously acquire micro inertial measurement unit data, heading data of a magnetic vector sensor, distance measured by an ultrasonic range finder from an obstacle, and building structured information data; A confidence weight calculation unit configured to calculate, in real time, a confidence weight of each sub-filter according to the micro inertial measurement unit data, the heading data of the magnetic vector sensor, the distance measured by the ultrasonic range finder from the obstacle, and the building structured information data; A noise adjustment unit configured to calculate a dynamic noise adaptive adjustment factor according to the confidence weight of each sub-filter, and update each sub-filter by the dynamic noise adaptive adjustment factor; A fault detection unit configured to perform multi-dimensional fault detection on the updated sub-filters by double thresholds and consistency check, identify a faulty sub-filter, delete faulty data sources by a complementary compensation mechanism, and obtain effective sub-filters by using the remaining data sources to compensate for errors caused by the faulty data sources; A pedestrian position pre-judging unit configured to pre-judge a pedestrian position according to the micro inertial measurement unit data and the building structured information data; A dynamic information distribution unit configured to correct and calibrate the effective sub-filters according to the pre-judged pedestrian position and the building structured information data, and perform dynamic information distribution on the corrected and calibrated effective sub-filters according to the confidence weight; A correction unit configured to correct the information loss in the dynamic information distribution of the effective sub-filter by a sub-optimality correction factor to obtain a globally optimal estimation.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the multi-source information fusion method for pedestrian positioning according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the multi-source information fusion method for pedestrian positioning according to any one of claims 1 to 5.

Images