Indoor positioning method based on combination of inertial navigation and indoor structure information

Abstract

The invention provides an indoor positioning method based on combination of inertial navigation and indoor structure information. The method comprises the following steps: taking a measured value of an inertial measurement unit (IMU) and an indoor structure map as input information; the IMU measured value is preliminarily corrected by using a zero-speed correction principle, and the step length and course information of the pedestrian are obtained through back-stepping as prediction of particle filtering, so that the accuracy of obtaining the step length and course by directly using inertial navigation information is improved. The indoor structure information map serves as a weight updating condition of particle filtering and is used for restraining deviation occurring in a pedestrian movement track, and landmarks are set at special positions to improve the positioning precision at the turning position. And inertial navigation and indoor structure information are combined to obtain a final positioning result. Under the indoor complex environment, the problem that positioning errors are large due to the fact that only an inertial sensor is used is solved, and the precision and stability of a positioning system are improved.

Description

Technical Field

[0001] The present invention relates to the field of indoor positioning and navigation of pedestrians, and in particular to an indoor positioning method based on the combination of inertial navigation and indoor structure information. Background Art

[0002] With the modernization of cities, indoor environments have become large and complex, and people spend most of their work and life indoors. The Global Navigation Satellite System (GNSS) cannot provide good positioning services due to signal obstruction in indoor environments. Therefore, indoor positioning technology faces challenges such as signal interference.

[0003] As one of the key technologies in the field of modern navigation, inertial navigation is a fully autonomous navigation system that does not require external input or output to the outside world. Compared with other navigation methods, it has the advantages of low power consumption and low cost. It is widely used in indoor positioning and can provide autonomous, continuous and stable positioning information. Although the accuracy is high in a short period of time, there is still a problem that the error will increase rapidly over time. With the development and popularization of wireless networks, network positioning technologies based on magnetic matching, WiFi, Bluetooth, ultra-wideband, etc. are gradually applied in indoor environments, and fusion positioning methods based on multi-source information are gradually applied to indoor positioning. However, this type of method is not an independent system. Its main disadvantage is that it requires the establishment of infrastructure in advance and is easy to affect indoor network signals. In places with high environmental noise, it is easy to cause the positioning accuracy of wireless networks to be unstable and unreliable.

[0004] Therefore, the use of inertial navigation combined with the existing indoor structure information map has become one of the hot spots of current indoor positioning technology. A common solution for integrating inertial navigation systems with auxiliary information is to use an extended Kalman filter (EKF). However, the EKF has limited ability to provide accurate positioning on nonlinear integrated navigation systems. The particle filter (PF) is a fully nonlinear state estimator that has flexibility in utilizing indoor structure information. However, using PF alone has problems such as computational burden and prone to failure and particle poverty. Summary of the invention

[0005] The present invention aims to solve the shortcomings of the above-mentioned prior art and proposes an indoor positioning method based on the combination of inertial navigation and indoor structural information, aiming to solve the problem of inaccurate positioning of a single inertial sensor, overcome the problem that the positioning accuracy cannot be improved well when using Kalman or particle filtering alone, and solve the impact on positioning accuracy when the direction of pedestrian movement changes greatly at corners. In this way, the demand for long-term stable and high-precision positioning in complex indoor environments can be met, and the stability of the positioning system can be improved.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme:

[0007] The indoor positioning method based on the combination of inertial navigation and indoor structure information of the present invention is characterized in that the platform is set up in the indoor environment only by relying on the IMU sensor and the indoor structure diagram of the building, and does not need to send or receive information to the outside world. The platform has an IMU sensor and a PC terminal. Among them, the IMU sensor includes: a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer; the indoor positioning method is performed according to the following steps:

[0008] Step 1: Fix the IMU sensor on the toe of the shoe, collect the three-axis accelerometer and three-axis gyroscope data of the IMU sensor, and send them to the PC for processing. According to the zero-speed correction principle, Kalman filtering is used to perform a preliminary solution on the measurement data to obtain the first positioning data.

[0009] Step 2: Use the first positioning data to reversely calculate the step length and heading information at that moment, which are used as the state update of the particle filter, and upload the indoor structure diagram of the building to the PC to digitize the indoor structure diagram of the building.

[0010] Step 3: Use the pedestrian's step length and heading information to predict the particle state, and perform wall penetration detection on the predicted particle trajectory to set the particle weight judgment condition and further optimize the pedestrian's motion trajectory.

[0011] Step 4: In order to improve the accuracy caused by large changes in pedestrian movement heading at turns, the turning point of the pedestrian's movement trajectory is detected, and landmarks are set up at special locations of the building structure (such as turning points and ends of corridors). The pedestrian's movement trajectory is corrected through landmark matching to obtain the final positioning data.

[0012] The indoor positioning method based on the combination of inertial navigation and indoor structure information described in the present invention is also characterized in that step 1 comprises:

[0013] Step 1.1, zero speed detection:

[0014] According to the characteristic that pedestrians have a short pause after each step, the speed of pedestrians should be zero at this time. Zero speed detection uses a method combining three conditions: acceleration modulus, acceleration sliding variance, and angular velocity modulus to detect:

[0015] (1)

[0016] In formula (1), Represents the output of the accelerometer in the three axes respectively. is the relative mean, and m is the sliding window size. Respectively represent the output of the gyroscope in three axes. The threshold range is set to 9.41m / s< <10.36m / s, <0.3, <6.8° / s.

[0017] Step 1.2, zero speed correction:

[0018] Ideally, when a pedestrian is stationary, his position remains unchanged and his speed is zero, but due to measurement noise and sensor errors, the actual speed of the pedestrian is not zero. Using zero speed as a pseudo-observation to use Kalman filtering to estimate and correct the results of indoor pedestrian navigation and positioning, suppressing the error accumulation of inertial sensors, this process is called zero-speed correction, and this feature is used to make a preliminary correction to the positioning results of pedestrians.

[0019] The step 2 includes inverting the pedestrian step length and heading information from the first positioning data, and processing the indoor structure map of the building:

[0020] Step 2.1, Pedestrian Dead Reckoning:

[0021] Using the navigation results obtained after zero-speed correction, the pedestrian's step length and heading information are inversely calculated. The step length and heading information obtained in this way are more accurate than the results of step length estimation and heading estimation alone. The propagation equation of pedestrian dead reckoning is:

[0022] (2)

[0023] in, and They are the first The easting and northing coordinates of the step; and Respectively represent pedestrians from Walk to The step length and heading angle during the step.

[0024] The distance between two adjacent steps is taken as the length of one step, and the inertial navigation result obtained in step 1.2 is used to calculate the distance between two adjacent steps. j -1 and j The step position coordinates are used to estimate the step length of the pedestrian:

[0025] (3)

[0026] in,( , ) is the pedestrian’s first j Step position coordinates.

[0027] Similarly, the system state vector of the Kalman filter algorithm is used to obtain the heading change between two adjacent steps:

[0028] (4)

[0029] in, is the first pedestrian in the Kalman filter navigation result The result coordinates are used to calculate the heading value of the pedestrian movement.

[0030] Step 2.2: Image processing:

[0031] The indoor structural diagram of the building uploaded to the PC is processed, the line segment information in the image is filtered using the line segment detection method, and then the digital information of the building wall is extracted to obtain a map in the form of line segments.

[0032] Step 2.3, coordinate registration:

[0033] The polygons obtained in step 2.2 are aligned to obtain the coordinate vector map of the building floors, and the indoor floor map of the building is converted to x and y In the rectangular coordinate system, it is directly applied to an indoor positioning method based on the combination of inertial navigation and indoor structure information.

[0034] The step 3 includes further constraining the movement trajectory of pedestrians by using particle filtering and indoor structure information of buildings on the pedestrian step length and heading information obtained by back-calculating the first positioning data:

[0035] Step 3.1, PDR-based particle filter:

[0036] The state transfer equation of the particle filter based on PDR is:

[0037] (5)

[0038] Among them, the superscript Represents the particle filter Particles. and Respectively represent the first The position of the pedestrian represented by the particle is The easting and northing coordinates of the step, such as: The position coordinates of a particle can be expressed as ( , ). Indicates that pedestrians from Walk to The step length of the step. Indicates the pedestrian's heading information. Walk to The heading changes during the step.

[0039] Step 3.2, wall penetration detection method:

[0040] In particle filtering, in order to update particle weights, solve the problem of valid particle detection, and avoid introducing erroneous particles in the next step, it is necessary to use the particle "wall penetration" detection method to test whether new particles pass through the wall of the building, that is, to transform the particle weight update problem into the relationship problem between two line segments. The judgment method is: use the line segment intersection detection algorithm to detect whether the particle passes through the wall, and the intersection of the line segments means that the particle has passed through the wall.

[0041] The relationship between line segments A, B and their intersection point C is as follows:

[0042] (6)

[0043] and are two coefficients, obtained from formula (7):

[0044] (7)

[0045] when and If it is greater than 0 and 1 at the same time, it is judged that the two line segments intersect.

[0046] Step 3.3: Update particle weights.

[0047] According to the result of the “wall penetration” detection in step 3.2, the weight of the particle is assigned. Formula (8) is the formula for assigning the particle weight:

[0048] (8)

[0049] The weight update of particles based on PDR particle filter is as follows:

[0050] (9)

[0051] in, For the j The particles generated in step The weight of the particle. The particles generated in step i The state of a particle. is the current observation, which represents the indoor digital map information. The indoor map information is used to constrain the propagation direction of the particles and complete the update of the state quantity, which is not a direct observation of the state quantity. It is expressed by formula (10):

[0052] (10)

[0053] in, Indicates The particles generated in step The distance between the position of a particle and the position solved by inertial navigation.

[0054] The step 4 includes further improving the positioning information obtained in the above steps by using landmarks:

[0055] Step 4.1: Inflection point detection and establishment of indoor landmarks:

[0056] When a pedestrian passes a turning point, the body's motion state will change significantly, so the pedestrian's heading data will fluctuate significantly. If the pedestrian's heading at the current moment changes significantly compared to the previous moment, the point where the change occurs is called an inflection point, and between two consecutive inflection points, the pedestrian's heading changes are relatively stable.

[0057] The obtained indoor vector map is used to select landmark points at the turning points and input coordinate data to establish a landmark database.

[0058] Step 4.2, landmark matching:

[0059] In indoor pedestrian navigation and positioning, it is necessary to match the detected turning points of pedestrians with the coordinates in the landmark library in sequence, and correct the turning point positions to the matched landmark points to correct the pedestrian's motion trajectory.

[0060] Use the inflection point detection method in step 4.1 to detect the inflection point of indoor pedestrian movement, and match it with the landmarks in the landmark library using the Euclidean distance. When the Euclidean distance between the inflection point position of the pedestrian movement and the known landmark is less than the set threshold M The match is successful. Then the current turning point position is corrected to the landmark point, and the error between the pedestrian turning point position coordinates and the set landmark point position coordinates is used to correct the path from the last detected turning point to the current turning point, and the corrected turning point is used as the starting point of the subsequent path. The Euclidean distance is calculated according to formula (11):

[0061] (11)

[0062] Compared with the prior art, the beneficial effects of the method of the present invention are embodied in:

[0063] The present invention uses an IMU, and utilizes the preliminary results of pedestrian navigation and positioning that have been corrected for zero speed to infer the step length and heading information of the pedestrian. Combined with the digital map of the indoor structural information of the building, the preliminary positioning results are further constrained using particle filtering, and landmarks are added at special locations of the building to further correct the positioning results of pedestrians moving indoors. The present invention effectively solves the problem of the rapid divergence of the accuracy of a single IMU sensor over time, and overcomes the impact of large changes in heading when pedestrians move on positioning accuracy. Compared to the inertial navigation algorithm, the present invention combines the advantages of Kalman filtering and particle filtering algorithms, and does not need to receive other information from the outside world. Under the premise of complete autonomy, the indoor structure information and all the information of the IMU sensor are utilized to reduce the errors generated by a single sensor, meeting the needs of long-term stable and high-precision positioning in complex indoor environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a schematic diagram of an indoor positioning method system based on the combination of inertial navigation and indoor structure information of the present invention;

[0065] Figure 2 This is the schematic diagram of the Kalman filter;

[0066] Figure 3 It is a schematic diagram of pedestrian dead reckoning;

[0067] Figure 4 It is a graph of line segment intersection;

[0068] Figure 5 It is a flowchart of landmark correction; Implementation

[0069] In this embodiment, Figure 1 and Figure 5 As shown in the figure, an indoor positioning method based on the combination of inertial navigation and indoor structure information is to set up a motion platform in the indoor positioning area, and set up an IMU sensor and a PC on the motion platform. The IMU sensor is fixed on the tip of the shoe to collect data; at the same time, the indoor structure information of the building is uploaded to the PC for map digitization and registration. The data collected by the IMU is first corrected using the Kalman filter according to the zero-speed correction algorithm, and the step length and heading information of the pedestrian's step are inversely calculated; secondly, the particle filter algorithm is used to detect the wall penetration and update the particle weight of the pedestrian's motion trajectory according to the indoor structure information; finally, by establishing landmarks at special positions of the indoor structure, the motion trajectory is corrected by landmark matching to obtain the latest positioning. Specifically, it is carried out in the following steps:

[0070] Step 1: Fix the IMU sensor on the toe of the shoe, collect the three-axis accelerometer and three-axis gyroscope data of the IMU sensor, and send them to the PC for processing. According to the zero-speed correction principle, Kalman filtering is used to perform a preliminary solution on the measurement data to obtain the first positioning data.

[0071] Step 1.1, zero speed detection:

[0072] According to the characteristic that pedestrians have a short pause after each step, the speed of pedestrians should be zero at this time. Zero speed detection uses a method combining three conditions: acceleration modulus, acceleration sliding variance, and angular velocity modulus to detect:

[0073] (1)

[0074] In formula (1), Represents the output of the accelerometer in the three axes respectively. is the relative mean, and m is the sliding window size. Respectively represent the output of the gyroscope in three axes. The threshold range is set to 9.41m / s< <10.36m / s, <0.3, <6.8° / s.

[0075] Step 1.2, zero speed correction:

[0076] Ideally, when a pedestrian is stationary, his position remains unchanged and his speed is zero, but due to measurement noise and sensor errors, the actual speed of the pedestrian is not zero. Using zero speed as a pseudo-observation to use Kalman filtering to estimate and correct the results of indoor pedestrian navigation and positioning, suppressing the error accumulation of inertial sensors, this process is called zero-speed correction, and this feature is used to make a preliminary correction to the positioning results of pedestrians.

[0077] The step 2 includes inverting the pedestrian step length and heading information from the first positioning data, and processing the indoor structure map of the building:

[0078] Step 2.1, Pedestrian Dead Reckoning:

[0079] Using the navigation results obtained after zero-speed correction, the pedestrian's step length and heading information are inversely calculated. The step length and heading information obtained in this way are more accurate than the results of step length estimation and heading estimation alone. The propagation equation of pedestrian dead reckoning is:

[0080] (2)

[0081] in, and They are the first The easting and northing coordinates of the step; and Respectively represent pedestrians from Walk to The step length and heading angle during the step.

[0082] The distance between two adjacent steps is taken as the length of one step, and the inertial navigation result obtained in step 1.2 is used to calculate the distance between two adjacent steps. j -1 and j The step position coordinates are used to estimate the step length of the pedestrian:

[0083] (3)

[0084] in,( , ) is the pedestrian’s first j Step position coordinates.

[0085] Similarly, the system state vector of the Kalman filter algorithm is used to obtain the heading change between two adjacent steps:

[0086] (4)

[0087] in, is the first pedestrian in the Kalman filter navigation result The result coordinates are used to calculate the heading value of the pedestrian movement.

[0088] Step 2.2: Image processing:

[0089] The indoor structural diagram of the building uploaded to the PC is processed, the line segment information in the image is filtered using the line segment detection method, and then the digital information of the building wall is extracted to obtain a map in the form of line segments.

[0090] Step 2.3, coordinate registration:

[0091] The polygons obtained in step 2.2 are aligned to obtain the coordinate vector map of the building floors, and the indoor floor map of the building is converted to x and y In the rectangular coordinate system, it is directly applied to an indoor positioning method based on the combination of inertial navigation and indoor structure information.

[0092] The step 3 includes further constraining the movement trajectory of pedestrians by using particle filtering and indoor structure information of buildings on the pedestrian step length and heading information obtained by back-calculating the first positioning data:

[0093] Step 3.1, PDR-based particle filter:

[0094] The state transfer equation of the particle filter based on PDR is:

[0095] (5)

[0096] Among them, the superscript Represents the particle filter Particles. and Respectively represent the first The position of the pedestrian represented by the particle is The easting and northing coordinates of the step, such as: The position coordinates of a particle can be expressed as ( , ). Indicates that pedestrians from Walk to The step length of the step. Indicates the pedestrian's heading information. Walk to The heading changes during the step.

[0097] Step 3.2, wall penetration detection method:

[0098] In particle filtering, in order to update particle weights, solve the problem of valid particle detection, and avoid introducing erroneous particles in the next step, it is necessary to use the particle "wall penetration" detection method to test whether new particles pass through the wall of the building, that is, to transform the particle weight update problem into the relationship problem between two line segments. The judgment method is: use the line segment intersection detection algorithm to detect whether the particle passes through the wall, and the intersection of the line segments means that the particle has passed through the wall.

[0099] The relationship between line segments A, B and their intersection point C is as follows:

[0100] (6)

[0101] and are two coefficients, obtained from formula (7):

[0102] (7)

[0103] when and If it is greater than 0 and 1 at the same time, it is judged that the two line segments intersect.

[0104] Step 3.3: Particle weight update

[0105] According to the result of the “wall penetration” detection in step 3.2, the weight of the particle is assigned. Formula (8) is the formula for assigning the particle weight:

[0106] (8)

[0107] The weight update of particles based on PDR particle filter is as follows:

[0108] (9)

[0109] in, For the j The particles generated in step The weight of the particle. The particles generated in step i The state of a particle. is the current observation, which represents the indoor digital map information. The indoor map information is used to constrain the propagation direction of the particles and complete the update of the state quantity, which is not a direct observation of the state quantity. It is expressed by formula (10):

[0110] (10)

[0111] in, Indicates The particles generated in step The distance between the position of a particle and the position solved by inertial navigation.

[0112] The step 4 includes further improving the positioning information obtained in the above steps by using landmarks:

[0113] Step 4.1: Inflection point detection and establishment of indoor landmarks:

[0114] When a pedestrian passes a turning point, the body's motion state will change significantly, so the pedestrian's heading data will fluctuate significantly. If the pedestrian's heading at the current moment changes significantly compared to the previous moment, the point where the change occurs is called an inflection point, and between two consecutive inflection points, the pedestrian's heading changes are relatively stable.

[0115] The obtained indoor vector map is used to select landmark points at the turning points and input coordinate data to establish a landmark database.

[0116] Step 4.2, landmark matching:

[0117] In indoor pedestrian navigation and positioning, it is necessary to match the detected turning points of pedestrians with the coordinates in the landmark library in sequence, and correct the turning point positions to the matched landmark points to correct the pedestrian's motion trajectory.

[0118] Use the inflection point detection method in step 4.1 to detect the inflection point of indoor pedestrian movement, and match it with the landmarks in the landmark library using the Euclidean distance. When the Euclidean distance between the inflection point position of the pedestrian movement and the known landmark is less than the set thresholdM The match is successful. Then the current turning point position is corrected to the landmark point, and the error between the pedestrian turning point position coordinates and the set landmark point position coordinates is used to correct the path from the last detected turning point to the current turning point, and the corrected turning point is used as the starting point of the subsequent path. The Euclidean distance is calculated according to formula (11):

[0119] (11).

Claims

1. An indoor positioning method based on the combination of inertial navigation and indoor structure information, characterized in that: The platform is set up in an indoor environment only by relying on the IMU sensor and the indoor structure diagram of the building, and does not need to send or receive information to the outside world. The platform has an IMU sensor and a PC. Among them, the IMU sensor includes: a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer; the indoor positioning method is carried out in the following steps: Step 1: Fix the IMU sensor on the toe of the shoe, collect the three-axis accelerometer and three-axis gyroscope data of the IMU sensor, send it to the PC for processing, and use Kalman filtering to perform preliminary calculations on the measured data based on the zero-speed correction principle to obtain the first positioning data. Step 2: Use the first positioning data to reversely calculate the step length and heading information at that moment, which is used as the state update of the particle filter. Upload the indoor structure diagram of the building to the PC and digitize it. Step 3: Use the pedestrian's step length and heading information to predict the particle state, and perform wall penetration detection on the predicted particle trajectory to set the particle weight judgment condition and further optimize the pedestrian's motion trajectory. Step 4: In order to improve the accuracy caused by large changes in pedestrian movement heading at turns, the turning point of the pedestrian's movement trajectory is detected, and landmarks are set up at special locations of the building structure (such as turning points and ends of corridors). The pedestrian's movement trajectory is corrected through landmark matching to obtain the final positioning data.

2. The indoor positioning method based on the combination of inertial navigation and indoor structure information according to claim 1 is characterized in that: The step 1 comprises: Step 1.1, zero speed detection: According to the characteristic that pedestrians have a short pause after each step, the speed of pedestrians should be zero at this time. Zero speed detection uses a method combining three conditions: acceleration modulus, acceleration sliding variance, and angular velocity modulus to detect: (1) In formula (1), Represents the output of the accelerometer in the three axes respectively. is the relative mean, and m is the sliding window size. Respectively represent the output of the gyroscope in three axes. The threshold range is set to 9.41m / s< <10.36m / s, <0.3, <6.8° / s Step 1.2, zero speed correction: Ideally, when a pedestrian is stationary, his position remains unchanged and his speed is zero, but due to measurement noise and sensor errors, the actual speed of the pedestrian is not zero. Using zero speed as a pseudo-observation to use Kalman filtering to estimate and correct the results of indoor pedestrian navigation and positioning, suppressing the error accumulation of inertial sensors, this process is called zero-speed correction, and this feature is used to make a preliminary correction to the positioning results of pedestrians.

3. The indoor positioning method based on the combination of inertial and indoor structure information according to claim 1, characterized in that: The step 2 includes inverting the pedestrian step length and heading information from the first positioning data, and processing the indoor structure map of the building: Step 2.1, Pedestrian Dead Reckoning: Using the navigation results obtained after zero-speed correction, the pedestrian's step length and heading information are inversely calculated. The step length and heading information obtained in this way are more accurate than the results of step length estimation and heading estimation alone. The propagation equation of pedestrian dead reckoning is: (2) in, and They are the first The easting and northing coordinates of the step; and Respectively represent pedestrians from Walk to Step length and heading angle The distance between two adjacent steps is taken as the length of one step, and the inertial navigation result obtained in step 1.2 is used to calculate the distance between two adjacent steps. j -1 and j The step position coordinates are used to estimate the step length of the pedestrian: (3) in,( , ) is the pedestrian’s first j Step position coordinates Similarly, the system state vector of the Kalman filter algorithm is used to obtain the heading change between two adjacent steps: (4) in, is the first pedestrian in the Kalman filter navigation result Step 1 The result coordinates are used to calculate the heading value of the pedestrian movement Step 2.2: Image processing: The indoor structure diagram of the building uploaded to the PC is processed, the line segment information in the image is filtered using the line segment detection method, and then the digital information of the building wall is extracted to obtain a map in the form of line segments. Step 2.3, coordinate registration: The polygons obtained in step 2.2 are aligned to obtain the coordinate vector map of the building floors, and the indoor floor map of the building is converted to and In the rectangular coordinate system, it is directly applied to an indoor positioning method based on the combination of inertial navigation and indoor structure information.

4. The indoor positioning method based on the combination of inertial navigation and indoor structure information according to claim 1, characterized in that: The step 3 includes further constraining the movement trajectory of the pedestrian using particle filtering and indoor structure information of the building based on the pedestrian step length and heading information obtained by back-calculating the first positioning data: Step 3.1, PDR-based particle filter: The state transfer equation of the particle filter based on PDR is: (5) Among them, the superscript Indicates the particle filter Particles. and Respectively represent the first The position of the pedestrian represented by the particle is The easting and northing coordinates of the step, such as: The position coordinates of a particle can be expressed as ( , ). Indicates that pedestrians from Walk to The step length of the step. Indicates the pedestrian's heading information. Walk to Heading change during walking Step 3.2, wall penetration detection method: In particle filtering, in order to update particle weights, solve the problem of valid particle detection, and avoid introducing erroneous particles in the next step, it is necessary to use the particle "wall penetration" detection method to test whether new particles pass through the wall of the building, that is, to transform the particle weight update problem into the relationship problem between two line segments. The judgment method is: use the line segment intersection detection algorithm to detect whether the particle passes through the wall. The intersection of the line segments means that the particle has passed through the wall. The relationship between line segments A, B and their intersection point C is as follows: (6) and are two coefficients, obtained from formula (7): (7) when and If it is greater than 0 and 1 at the same time, then the two line segments intersect. Step 3.3: Particle weight update According to the result of the "wall penetration" detection in step 3.2, the weight of the particle is assigned. Formula (8) is the formula for assigning the particle weight: (8) The weight update of particles based on PDR particle filter is as follows: (9) in, For the j The particles generated in step The weight of the particle. The particles generated in step i The state of a particle. is the current observation, which represents the indoor digital map information. The indoor map information is used to constrain the propagation direction of the particles and complete the update of the state quantity, which is not a direct observation of the state quantity. It is expressed by formula (10): (10) in, Indicates The particles generated in step The distance between the position of a particle and the position solved by inertial navigation.

5. The indoor positioning method based on the combination of inertial navigation and indoor structure information according to claim 1, characterized in that: The step 4 includes further improving the positioning information obtained in the above steps by using landmarks: Step 4.1: Inflection point detection and establishment of indoor landmarks: When a pedestrian passes a turning point, the body's motion state will change significantly, so the pedestrian's heading data will fluctuate significantly. If the pedestrian's heading at the current moment changes significantly compared to the previous moment, the point where the change occurs is called an inflection point. Between two consecutive inflection points, the pedestrian's heading changes are relatively stable. Use the obtained indoor vector map to select landmark points at the turning points and input coordinate data to establish a landmark database Step 4.2, landmark matching: In indoor pedestrian navigation and positioning, it is necessary to match the detected pedestrian turning points with the coordinates in the landmark library in sequence, and correct the turning point position to the matched landmark point to correct the pedestrian's motion trajectory. Use the inflection point detection method in step 4.1 to detect the inflection point of indoor pedestrian movement, and match it with the landmarks in the landmark library using the Euclidean distance. When the Euclidean distance between the inflection point position of the pedestrian movement and the known landmark is less than the set threshold M The match is successful. Then the current turning point position is corrected to the landmark point, and the error between the pedestrian turning point position coordinates and the set landmark point position coordinates is used to correct the path from the last detected turning point to the current turning point, and the corrected turning point is used as the starting point of the subsequent path. The Euclidean distance is calculated according to formula (11): (11)。

Images