CN106595633B - Indoor positioning method and device - Google Patents

Abstract

The present invention provides a kind of indoor orientation method and device, belongs to indoor positioning technologies field.Method includes: to predict the location information of pedestrian according to the collected data of multiple sensor；Based on indoor sport model, the indoor sport state of pedestrian is obtained；Based on indoor environment cartographic model, is calibrated according to indoor sport state and the location information of indoor default node, the pedestrian position information obtained to prediction, obtain the final position information of pedestrian.The location information that the present invention passes through prediction pedestrian.Based on indoor sport model, the indoor sport state of pedestrian is obtained.Based on indoor environment cartographic model, is calibrated according to indoor sport state and the location information of indoor default node, the pedestrian position information obtained to prediction, obtain the final position information of pedestrian.Due to not having to installation external equipment, to can also reduce hardware cost consumption, cost is relatively low so that expending when indoor positioning while avoiding design complexities higher system.

Description

Indoor positioning method and device

technical field

The present invention relates to the technical field of indoor positioning, and more particularly, to an indoor positioning method and device.

Background technique

In many current application fields, it is very important to obtain the position information of people or objects. With the continuous progress of science and technology and communication technology, the positioning technology based on LBS (Location Based Service, location service) is also developing rapidly. The main problem solved by LBS is to provide existing resources to users and provide specific services according to their surrounding environment.

In an outdoor environment, position information can be obtained through GPS (Global Positioning System, global positioning system). With the rapid development of information technology, GPS positioning technology can meet the daily needs of ordinary people. Accordingly, GPS technology has always been the mainstream of research. Since GPS is positioned by satellites, the best positioning accuracy can be obtained only when four satellites are positioned at the same time. However, in an indoor environment, the strength and quality of satellite signals drop rapidly due to the obstruction of GPS signals by buildings. At the same time, because the indoor environment is very complex, accurate positioning cannot be performed indoors. Therefore, indoor positioning technology is more difficult.

In addition, research shows that people spend 90% of their daily lives indoors, which means they spend most of their time outside of GPS signals. It can be seen that indoor positioning technology is also a key demand of people. With the development of intelligent communication equipment, the research of indoor positioning technology has also developed. Smart mobile terminals such as smart phones and tablet computers used in people's daily life integrate high-precision sensor elements such as MEMS (Micro-Electro-Mechanical System, Micro-Electro-Mechanical System), accelerometers, gyroscopes, and barometers. The extensive use of these sensor elements provides a basis for indoor positioning technology and the accuracy of the sensor elements can also meet the needs of indoor positioning technology. In addition, assisted positioning technology has also been widely and fully studied at the same time. For example, indoor positioning technology based on wireless fingerprint signals, vision-based positioning technology, etc. These approaches are used either in conjunction with sensors or as stand-alone positioning techniques. Accordingly, the accuracy of indoor positioning is also improved. In recent years, companies such as Google have begun to build indoor maps, gradually covering the iconic buildings in some big cities, so that indoor maps can also become an auxiliary means of positioning.

Since 2000, on the basis of the gradual popularization of intelligent terminals, LBS has been gradually applied in the fields of emergency rescue, disaster prevention, logistics management, equipment inspection and medical care. In this context, the need for seamless connection from outdoor to indoor positioning is also raised.

In the field of public security, commercial services and warehousing services, high-precision indoor positioning technology is of great significance. For example, in terms of social and public security, accurate indoor positioning technology can provide indoor navigation services for firefighters, police and other personnel, and locate prisoners in prisons. In commercial services, accurate indoor positioning technology can provide positioning services for families, provide navigation in indoor environments such as shopping malls and museums, and provide nursing positioning services for the elderly, children and patients in hospitals and families. In warehousing services, items can be located. Therefore, indoor positioning technology has a wide range of application scenarios.

On this basis, the research direction of pedestrian indoor positioning and navigation technology is mainly to develop towards low cost, easy portability and improved positioning accuracy. The existing indoor positioning method mainly pre-installs multiple external devices, such as AP (AccessPiont, access point), and locates pedestrians according to the mapping relationship between the signal strength and the distance between the mobile terminal and the external device.

In the process of implementing the present invention, it is found that the prior art has at least the following problems: due to the need to install multiple external devices, the cost of the device is relatively high. In addition, a system with high design complexity is usually required to work together among multiple external devices. Therefore, the cost of indoor positioning is relatively high.

SUMMARY OF THE INVENTION

The present invention provides an indoor positioning method and device that overcomes the above problems or at least partially solves the above problems.

According to an aspect of the present invention, an indoor positioning method is provided, the method comprising:

According to the data collected by multiple sensors, predict the location information of pedestrians;

Obtain the indoor motion state of pedestrians based on the indoor motion model;

Based on the indoor environment map model, the predicted pedestrian position information is calibrated according to the indoor motion state and the position information of the indoor preset nodes, and the final position information of the pedestrian is obtained.

According to another aspect of the present invention, an indoor positioning device is provided, the device comprising:

The prediction module is used to predict the location information of pedestrians according to the data collected by multiple sensors;

The acquisition module is used to acquire the indoor motion state of pedestrians based on the indoor motion model;

The calibration module is used for calibrating the predicted pedestrian position information based on the indoor environment map model, according to the indoor motion state and the position information of the indoor preset nodes, to obtain the final position information of the pedestrian.

The beneficial effects brought by the technical solution proposed by the application are:

The location information of pedestrians is predicted based on the data collected by multiple sensors. Based on the indoor motion model, the indoor motion state of the pedestrian is obtained. Based on the indoor environment map model, the predicted pedestrian position information is calibrated according to the indoor motion state and the position information of the indoor preset nodes, and the final position information of the pedestrian is obtained. Since there is no need to install external equipment, while avoiding designing a system with high complexity, the hardware cost consumption can also be reduced, so that the cost of indoor positioning is lower.

In addition, the accelerometer, barometer and gyroscope data are selected during the classification of motion features, which improves the accuracy of classification of motion features and avoids the occurrence of long-term accumulated errors. Since the positioning process combines the pedestrian dead reckoning algorithm, the indoor pedestrian motion feature and the hidden Markov model matching method, the robustness of indoor positioning can be improved while ensuring a high positioning accuracy.

Description of drawings

1 is a schematic flowchart of an indoor positioning method according to an embodiment of the present invention;

2 is a schematic flowchart of an indoor positioning method according to an embodiment of the present invention;

FIG. 3 is a schematic structural diagram of an indoor positioning device according to an embodiment of the present invention.

Detailed ways

The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. The following examples are intended to illustrate the present invention, but not to limit the scope of the present invention.

The existing indoor positioning method mainly pre-installs multiple external devices, such as AP (Access Piont, access point), and locates pedestrians according to the mapping relationship between the signal strength and the distance between the mobile terminal and the external device. Since multiple external devices need to be installed, the equipment cost is high. In addition, a system with high design complexity is usually required to work together among multiple external devices. Therefore, the cost of indoor positioning is relatively high.

In view of the problems in the prior art, this embodiment provides an indoor positioning method. The method is applied to a mobile terminal, and the mobile terminal includes but is not limited to a mobile phone, a tablet computer, a smart watch, and the like. In addition, since the embodiment of the present invention needs to be applied to the data collected by the sensor, an acceleration sensor, a gyroscope, a barometer, etc. may be installed in the mobile terminal, which is not specifically limited in this embodiment.

Referring to FIG. 1 , the method process provided in this embodiment includes: 101. Predicting the location information of pedestrians according to data collected by multiple sensors; 102. Obtaining the indoor motion state of pedestrians based on an indoor motion model; 103. Based on an indoor environment map model , and calibrate the predicted pedestrian position information according to the indoor motion state and the position information of the indoor preset nodes to obtain the final position information of the pedestrian.

The method provided by the embodiment of the present invention predicts the location information of pedestrians according to the data collected by multiple sensors. Based on the indoor motion model, the indoor motion state of the pedestrian is obtained. Based on the indoor environment map model, the predicted pedestrian position information is calibrated according to the indoor motion state and the position information of the indoor preset nodes, and the final position information of the pedestrian is obtained. Since there is no need to install external equipment, while avoiding designing a system with high complexity, the hardware cost consumption can also be reduced, so that the cost of indoor positioning is lower.

In addition, the accelerometer, barometer and gyroscope data are selected during the classification of motion features, which improves the accuracy of classification of motion features and avoids the occurrence of long-term accumulated errors. Since the positioning process combines the pedestrian dead reckoning algorithm, the indoor pedestrian motion feature and the hidden Markov model matching method, the robustness of indoor positioning can be improved while ensuring a high positioning accuracy.

As an optional embodiment, the location information of pedestrians is predicted according to data collected by multiple sensors, including:

Determining the total number of steps a pedestrian takes between starting to move and stopping;

For each step of the pedestrian's movement, according to the pedestrian's step length, steering angle and the position information before the pedestrian's movement, calculate the position information after the pedestrian moves one step, until the number of calculations reaches the total number of steps, and the final calculation result is used as the pedestrian's position information.

As an optional embodiment, before determining the total number of steps of the pedestrian from starting to moving to stopping, the method further includes:

According to the acceleration of each sampling point, the pedestrian starts to move and stops to move.

As an optional embodiment, according to the acceleration of each sampling point, the detection of the pedestrian starting to move includes:

For any sampling point, when it is detected that the acceleration of any sampling point is not less than the first preset threshold, it is determined that the pedestrian starts to move at any sampling point.

As an optional embodiment, according to the acceleration of each sampling point, the pedestrian stops moving to be detected, including:

For any sampling point, when it is detected that the acceleration of any sampling point is smaller than the first preset threshold, count the number of sampling points that are continuously smaller than the first preset threshold since any sampling point;

When the statistical result reaches the preset number, the last sampling point when the statistical result reaches the preset number is obtained, and it is determined that the pedestrian stops moving at the last sampling point.

As an optional embodiment, determining the total number of steps the pedestrian takes from starting to moving to stopping, including:

For any sampling point during the period from the start of the pedestrian to the stop of movement, when the acceleration of any sampling point is detected to be greater than the second preset threshold, and the acceleration corresponding to the next sampling point of any sampling point is less than the second When the threshold is preset, the previous sampling point of any sampling point, any sampling point and the next sampling point of any sampling point are regarded as one step cycle, and the total number of steps of pedestrians is increased by one.

As an optional embodiment, the previous sampling point of any sampling point, any sampling point and the next sampling point of any sampling point are regarded as a step cycle, and before adding one to the total number of steps of pedestrians, also include:

The second preset threshold is calculated according to the acceleration-related value in the previous step cycle.

As an optional embodiment, before calculating the position information after the pedestrian moves by one step according to the pedestrian's step length, the steering angle and the position information before the pedestrian moves, the method further includes:

Based on the space coordinate system, obtain the angular acceleration of the current pedestrian in three directions;

Based on the projection relationship between the space coordinate system and the ground coordinate system, the steering angle of the pedestrian is calculated according to the current angular acceleration and acceleration of the pedestrian in three directions.

As an optional embodiment, the location information of pedestrians is predicted according to data collected by multiple sensors, including:

Determine the floor where the pedestrian is located according to the air pressure value at the location of the pedestrian.

As an optional embodiment, based on the indoor motion model, the indoor motion state of the pedestrian is obtained, including:

Calculate the eigenvalue corresponding to the first time window according to the acceleration of each sampling point in the first time window;

Based on the motion state classifier, the motion feature of the pedestrian is determined according to the feature value corresponding to the first time window.

As an optional embodiment, based on the indoor motion model, the indoor motion state of the pedestrian is obtained, including:

Calculate the turning angle of the pedestrian according to the turning angle of the pedestrian in the second time window;

According to the turning angle of the pedestrian, the turning characteristic of the pedestrian is determined.

As an optional embodiment, based on the indoor environment map model, according to the indoor motion state and the position information of the indoor preset nodes, the predicted pedestrian position information is calibrated, and the final position information of the pedestrian is obtained, including:

Based on the indoor environment map model, determine the movement probability of pedestrians moving to adjacent preset nodes in the indoor environment map model;

Sort the movement probabilities of each adjacent preset node, and select the two movement probabilities with the largest numerical values in the sorting result, which are respectively the first movement probability and the second movement probability, and the first movement probability is greater than the second movement probability;

When the ratio of the first movement probability to the second movement probability is greater than the third preset threshold, the position information of the adjacent preset nodes corresponding to the first movement probability is used as the final position information of the pedestrian.

As an optional embodiment, based on the indoor environment map model, determining the movement probability of pedestrians moving to adjacent preset nodes in the indoor environment map model includes:

For any adjacent preset node in the indoor environment map model, according to the position information of any adjacent preset node, calculate the emission probability of pedestrian moving to any adjacent preset node;

According to the motion recognition probability matrix, determine the state recognition probability that the motion state of any adjacent preset node represents the indoor motion state;

The product between the emission probability and the state recognition probability is taken as the movement probability of the pedestrian moving to any adjacent preset node.

As an optional embodiment, the movement probabilities of each adjacent preset node are sorted, and the two movement probabilities with the largest numerical values in the sorting result are selected, which are respectively the first movement probability and the second movement probability, and further include:

When the ratio of the first movement probability to the second movement probability is not greater than the third preset threshold, the predicted pedestrian position information is used as the pedestrian's final position information.

All the above-mentioned optional technical solutions can be combined arbitrarily to form optional embodiments of the present invention, which will not be repeated here.

Based on the content provided by the above-mentioned embodiment corresponding to FIG. 1 , an embodiment of the present invention provides an indoor positioning method. Referring to FIG. 2 , the method process provided by this embodiment includes: 201. Determine the total number of steps between the pedestrian starts to move and stops moving; 202. For each step of the pedestrian's movement, according to the pedestrian's step length, the steering angle and the pedestrian's movement 203. Based on the indoor motion model, obtain the indoor motion state of the pedestrian; 204. Based on the indoor motion The environment map model calibrates the predicted pedestrian position information according to the indoor motion state and the position information of the indoor preset nodes, and obtains the final position information of the pedestrian.

Wherein, 201, determine the total number of steps between the pedestrian starts moving and stops moving.

The method provided in this embodiment is mainly a process of first predicting the location information of pedestrians according to data collected by multiple sensors, and then calibrating the predicted location information of pedestrians to realize positioning. Among them, the steps 201 to 202 are mainly the process of predicting the pedestrian position information according to the data collected by the multiple sensors.

Since it is unknown when the pedestrian starts to move indoors and when it stops moving, before step 201 is executed, the pedestrian can also be detected based on the acceleration of each sampling point, which is not performed in this embodiment. Specific restrictions.

This embodiment does not specifically limit the method of detecting the pedestrian's start to move according to the acceleration of each sampling point, including but not limited to: for any sampling point, when the detected acceleration of any sampling point is not less than the first When a preset threshold is set, it is determined that pedestrians start to move at any sampling point.

Among them, the sampling point corresponds to the sampling period of the sensor. The first preset threshold may be set according to the actual situation, which is not specifically limited in this embodiment. For example, take the first preset threshold as 1.5m/s ² as an example. If the sampling period of the acceleration sensor is 20ms, every 20ms is a sampling point. That is, when the acceleration sensor collects the acceleration value at each sampling point, the mobile terminal also determines whether the acceleration value collected at each sampling point is greater than or equal to 1.5m/s ² . If it is detected that the acceleration of a sampling point is greater than 1.5m/s ² , it is determined that the pedestrian starts to move from the sampling point.

This embodiment does not specifically limit the way of detecting pedestrian stop moving according to the acceleration of each sampling point, including but not limited to: for any sampling point, when the detected acceleration of any sampling point is less than the first preset threshold , the number of sampling points that are continuously smaller than the first preset threshold from any sampling point is counted; when the statistical result reaches the preset number, the last sampling point when the statistical result reaches the preset number is obtained, and it is determined that the pedestrian is at the last Stop moving at a sample point. The preset number may also be set according to the actual situation, which is not specifically limited in this embodiment.

For example, the first preset threshold is 1.5 m/s ² and the preset number is 30 as an example. If it is detected that the acceleration of the 10th sampling point is less than 1.5 m/s ² , continue to detect the acceleration of the 11th, 12th, . . . , 40th sampling points. When the acceleration of 30 consecutive sampling points after the 10th sampling point is less than 1.5m/s ² , that is, when the acceleration of the 11th, 12th, ..., 40th sampling point is less than 1.5m/s ² , and obtain the last sampling point when the accumulative number reaches 30 sampling points, that is, the 40th sampling point. Correspondingly, at the 40th sampling point, the pedestrian stops moving.

After it is determined that the pedestrian starts to move and stops to move, the total number of steps the pedestrian walks during the period from the start of the move to the stop of the move can be determined. This embodiment does not specifically limit the manner of determining the total number of steps of a pedestrian from starting to moving, including but not limited to: for any sampling point during the period from the start of the pedestrian to the stop of moving, when the pedestrian is detected When the acceleration of any sampling point is greater than the second preset threshold, and the acceleration corresponding to the next sampling point of any sampling point is less than the second preset threshold, the previous sampling point, any sampling point and The next sampling point of any sampling point is used as a step cycle, and the total number of steps of the pedestrian is increased by one.

When a person is walking, the acceleration of one leg is larger when it is lifted out, and the acceleration is smaller when it is ready to land after being lifted out. Therefore, based on the above principle, two consecutive sampling points can be measured in the above process. detection. Comparing the two sampling points with the second preset threshold, when the former sampling point is greater than the second preset threshold and the latter sampling point is less than the second preset threshold, it can be considered that the pedestrian has taken a step. Correspondingly, for a sampling point whose acceleration is greater than the second preset threshold, when the acceleration corresponding to the next sampling point of the sampling point is smaller than the second preset threshold, the time between the last sampling point and the next sampling point can be calculated. A period of time as a pace cycle. For example, if the acceleration of the third sampling point is greater than the second preset threshold and the acceleration of the fourth sampling point is less than the second preset threshold, the second sampling point, the third sampling point and the fourth sampling point can be point as a pace cycle, and consider the pedestrian to take a step in this pace cycle. Accordingly, the total number of steps may be increased by one. It should be noted that the initial value of the total number of steps is 0 before the total number of steps is counted.

In addition, before the total number of steps is determined, the second preset threshold may also be calculated according to the acceleration related value in the previous step cycle, which is not specifically limited in this embodiment. According to the experimental results of pedestrian dead reckoning, the second preset threshold can be calculated by the dynamic threshold equation, and the dynamic threshold equation (1) is as follows:

Among them, α and β are preset parameters, which may be respectively 0.25 and 0.75, which are not specifically limited in this embodiment. γ is the variance of environmental noise, which can be 0.09, which is not specifically limited in this embodiment. T _old is the second preset threshold value of the previous step cycle, and A ₁ and A ₂ are acceleration-related values. A ₁ and A ₂ may respectively represent the maximum value and the minimum value of the acceleration in the previous step cycle, and A ₁ and A ₂ may also be the mean value or variance of the acceleration, which is not specifically limited in this embodiment.

Among them, 202. For each step of the pedestrian's movement, calculate the position information after the pedestrian moves one step according to the pedestrian's step length, steering angle and the position information before the pedestrian's movement, until the number of calculation times reaches the total number of steps, and use the final calculation result as the pedestrian. location information.

Before executing this step, the step length of the pedestrian, the position information of the pedestrian before moving, and the steering angle of the pedestrian may be obtained. The step length of the pedestrian is estimated according to the height of the pedestrian, which is not specifically limited in this embodiment. Since the method provided in this embodiment is an iterative calculation process, that is, the position information of the pedestrian before the last movement can also be obtained by the method provided in this embodiment, so that when obtaining the position information of the pedestrian before the movement, the position information of the pedestrian before the last movement can be obtained. The calculation results corresponding to the methods provided in the embodiments are sufficient. It should be noted that, when indoor positioning of the pedestrian is performed for the first time, the initial position may be set according to the actual situation, which is not specifically limited in this embodiment.

In addition, since the pedestrian may move in a curved motion, in order to more accurately predict the pedestrian's position, the steering angle of the pedestrian can also be obtained. This embodiment does not specifically limit the way of obtaining the steering angle of the pedestrian, including but not limited to: obtaining the angular acceleration of the current pedestrian in three directions based on the space coordinate system; based on the projection relationship between the space coordinate system and the ground coordinate system , according to the current angular acceleration and acceleration of the pedestrian in three directions, the steering angle of the pedestrian is calculated.

Among them, according to the space coordinate system, it can be divided into angular velocity in three directions of XYZ axis. The angular accelerations in the three directions may be acquired through a gyroscope in the mobile terminal, which is not specifically limited in this embodiment. Before calculating the steering angle of the pedestrian, the angular displacements in the three directions may be calculated according to the angular accelerations in the three directions by means of integration, which is not specifically limited in this embodiment. The integration process can refer to the following formula (2):

in, and are the angular accelerations in the three directions, XYZ, respectively. t _begin is the start time of one move, and t _stop is the end time of one move. θ _x , θ _y and θ _z are the angular displacements in the three directions of XYZ, respectively.

It should be noted that if the angular displacements in the three directions of XYZ in a time window are relatively small, such as less than the angle threshold, it can be considered that the pedestrian is moving in a straight line in this time window. For example, take an angle threshold of 15° as an example. When the angular displacements θ _x , θ _y and θ _z in the three directions of XYZ in a time window are all less than 15°, it can be determined that the pedestrian moves in a straight line in this time window.

Based on the above principles, the pedestrian can calculate the arithmetic mean of accelerations in three directions when walking in a straight line. Correspondingly, the steering angle of the pedestrian can be calculated according to the current arithmetic mean value of the angular displacement and acceleration of the pedestrian in three directions, which is not specifically limited in this embodiment. The above calculation process can be represented by the pedestrian dead reckoning model (3) defined as follows:

where O _z is the steering angle of the pedestrian. θ _x , θ _y and θ _z are angular displacements in three directions, respectively. and are the arithmetic mean of the accelerations in the three directions, respectively.

After the pedestrian's steering angle is calculated, for each step of the pedestrian's movement, the position information after the pedestrian moves one step can be calculated according to the pedestrian's step length, steering angle, and position information before the pedestrian moves. The calculation process can refer to the following formula (4):

Among them, _lk is the step size of the pedestrian. Based on the ground coordinate system, x _k-1 and y _k-1 are the position information of the pedestrian before moving one step, and x _k and y _k are the position information of the pedestrian after moving one step.

Based on the total number of steps determined in the above step 201, the position information of the pedestrian after taking so many steps can be determined according to the above formula (4), that is, the predicted position information of the pedestrian.

It should be noted that the above-mentioned process mainly predicts the location information of pedestrians based on the ground coordinate system. Since the floor of the building where the pedestrian is located may need to be located when locating the pedestrian indoors, the predicted pedestrian location information may also include the floor where the pedestrian is located. This embodiment also provides a method for determining the floor where the pedestrian is located, including but not limited to: determining the floor where the pedestrian is located according to the air pressure value at the location where the pedestrian is located.

Since the air pressure in the earth's atmosphere is inversely proportional to the height, that is, the air pressure decreases when the height increases, so the air pressure measured by the barometer can be used to calculate the height of the pedestrian. The floor where the pedestrian is located can be determined according to the floor height of the building where the pedestrian is located and the height where the pedestrian is located, which is not specifically limited in this embodiment. Based on the ICAO (International Civil Aviation Organization, International Civil Aviation Organization) model, it can be known that the atmospheric pressure decreases by 1 mbar for every increase of about 8.7 meters in altitude. According to the standard atmospheric pressure in 1993, the height of the pedestrian can be calculated by referring to the following formula (5):

Among them, P ₀ represents the standard atmospheric pressure (1013.25mbar). H is the height of the pedestrian, in meters.

Among them, 203, based on the indoor motion model, obtain the indoor motion state of the pedestrian.

Before executing this step, the indoor motion state of the pedestrian may be determined according to the indoor motion habit of the pedestrian, which is not specifically limited in this embodiment. In this embodiment, the motion states of pedestrians are divided into 7 according to actual life scenarios, namely: walking, sitting, standing, going upstairs, going downstairs, turning and U-turn.

Among them, walking, sitting, standing, going upstairs and going downstairs are all motion characteristics of pedestrians in the process of moving. Turns and U-turns are the turning characteristics of pedestrians in the process of moving. Since pedestrians are turning, it may be a small turn to adjust the direction or a big turn to make a U-turn. Therefore, in order to distinguish these two kinds of turns, the above turning features are divided into turns and U-turns. When determining a turn and a U-turn, it can be determined according to the comparison result between the pedestrian's turning angle within the time window and the preset threshold range. For example, when the turning angle is between 50° and 135°, it can be regarded as an ordinary turn. When the turning angle is greater than 135°, it can be considered as a U-turn.

Based on the above content, when acquiring the indoor state of the pedestrian in this step, the motion feature and the turning feature of the pedestrian can be acquired respectively. This embodiment does not specifically limit the way of acquiring the motion features of pedestrians based on the indoor motion model, including but not limited to: calculating the feature value corresponding to the first time window according to the acceleration of each sampling point in the first time window; The state classifier determines the motion characteristics of the pedestrian according to the feature values corresponding to the first time window.

The motion state classifier may be trained by a KNN (k-Nearest-Neighbor, K nearest neighbor) classifier, which is not specifically limited in this embodiment. Specifically, a preset number of experimenters can be selected to practice the above five motion characteristics, namely walking, sitting, standing, upstairs and downstairs, and the characteristic values of the experimenters when they practice the motion characteristics can be obtained. By inputting a motion feature and the feature value of each experimenter each time the motion feature is practiced into the KNN model, the KNN model is continuously trained until the number of training times reaches a preset number of times. Finally, the motion state classifier and motion recognition probability matrix can be obtained according to the training results. Among them, the motion state classifier is used to determine the motion features of pedestrians according to the input feature values.

The motion recognition probability matrix is a probability matrix in which each motion feature of the pedestrian is recognized as well as other motion features. For example, take the pedestrian's current motion feature actually being "sit down" as an example. Based on the motion state classifier, it may be recognized that the pedestrian's motion feature is indeed "sit down" according to the input feature value. At this time, the recognition result is correct, which corresponds to a probability. However, it was originally "sit down" and may also be identified as "walking". At this time, the recognition result is wrong, which also corresponds to a probability. Based on the above theory, through continuous training of the KNN model, the motion recognition probability matrix can also be obtained while the motion state classifier is obtained.

Correspondingly, after the motion state classifier is obtained, only the feature value corresponding to the pedestrian in the first time window is obtained, and then the motion feature of the pedestrian can be determined based on the motion state classifier. The characteristic value may be calculated according to the acceleration of each sampling point in the first time window, which is not specifically limited in this embodiment. The characteristic value may include the mean value and variance of the acceleration in the three directions in the first time window, the air pressure value, and the magnitude of the acceleration in the three directions, etc., which are not specifically limited in this embodiment. The length of the first time window may be 2s to 3s, which is not specifically limited in this embodiment. Considering the continuity of actions when pedestrians walk, the first time window may also be set to overlap coverage of 50%, which is not specifically limited in this embodiment. For example, the first time window may be 0s˜2s, 1s˜3s, 2s˜4s . . .

In addition to determining the pedestrian's motion characteristics, this can determine the pedestrian's turning characteristics. This embodiment does not specifically limit the method of determining the turning feature of the pedestrian, including but not limited to: calculating the turning angle of the pedestrian according to the turning angle of the pedestrian in the second time window; determining the turning feature of the pedestrian according to the turning angle of the pedestrian.

The length of the second time window may also be 2s to 3s, which is not specifically limited in this embodiment. In addition, the lengths of the first time window and the second time window may be the same or different, which are not specifically limited in this embodiment. Specifically, when calculating the turning angle of the pedestrian, the first steering angle at the beginning of the second time window may be calculated first, and then the second steering angle at the end of the second time window may be calculated. The difference between the first steering angle and the second steering angle is taken as the turning angle of the pedestrian. The process of calculating the steering angle may refer to the process of the above formula (3), which will not be repeated here.

For example, the length of the second time window is 2s, and the second time window is 1s˜3s as an example. When calculating the turning angle of a pedestrian, the turning angle of the pedestrian at 1s can be calculated first, and then the turning angle of the pedestrian at 3s can be calculated, so that the difference between the two steering angles can be used as the turning angle of the pedestrian.

Through the above process, the motion characteristics and turning characteristics of pedestrians can be finally obtained. The motion feature is one of walking, sitting, standing, going upstairs and going downstairs, and the turning feature is one of turning and U-turn. Combined with motion features and turning features, the indoor motion state of pedestrians can be obtained.

Wherein, 204, based on the indoor environment map model, according to the indoor motion state and the position information of the indoor preset nodes, calibrate the predicted pedestrian position information to obtain the pedestrian's final position information.

Since the pedestrian position information predicted in the above steps 201 to 202 may have certain errors, in this step, based on the position information of the preset nodes in the indoor environment map model, the pedestrian predicted in the steps 201 to 202 can be used. position information for calibration. This embodiment does not specifically limit the method of calibrating the predicted pedestrian position information based on the indoor environment map model, according to the indoor motion state and the position information of the indoor preset nodes, and obtaining the final position information of the pedestrian, including but not limited to: Based on the indoor environment map model, determine the movement probability of pedestrians moving to adjacent preset nodes in the indoor environment map model; sort the movement probability of each adjacent preset node, and select the two movement probabilities with the largest values in the ranking result, They are the first movement probability and the second movement probability, respectively, and the first movement probability is greater than the second movement probability; when the ratio of the first movement probability and the second movement probability is greater than the third preset threshold, the phase corresponding to the first movement probability is calculated. The position information of the adjacent preset nodes is used as the final position information of the pedestrian.

Among them, the indoor environment map model is constructed from the actual life scene. Specifically, the locations of nodes such as stair corner points and aisle corner points in a building can usually be predetermined. According to the node coordinates of each node, the direction and distance between each node and its spatially accessible adjacent nodes, and the motion state of pedestrians on each node, an indoor environment map model can be established.

Before determining the movement probability of pedestrians moving to adjacent preset nodes in the indoor environment map model based on the indoor environment map model, the location information of the pedestrian before moving can be obtained first. It can be seen from the content in the above steps 201 to 204 that since the method provided in this embodiment is calculated iteratively, that is, the position information of the pedestrian after the next movement is continuously deduced according to the previous calculation result, so this step is obtained before the pedestrian moves. When the location information of the pedestrian is obtained, the location information of the pedestrian calculated by the method provided in this embodiment can be obtained last time, which is not specifically limited in this embodiment.

After determining the position information of the pedestrian before moving, according to the position information of each preset node in the indoor environment map model, it can be determined which preset nodes in the indoor environment map model the pedestrian can directly reach from the position before moving. These preset nodes are the adjacent preset nodes corresponding to the pedestrian's position before moving. Based on the above content, for any adjacent preset node, this embodiment does not specifically limit the way of determining the movement probability of pedestrians moving to adjacent preset nodes in the indoor environment map model, including but not limited to: for indoor environment map models For any adjacent preset node, according to the position information of any adjacent preset node, calculate the emission probability of pedestrian moving to any adjacent preset node; according to the motion recognition probability matrix, determine any adjacent preset node The motion state of , is expressed as the state recognition probability of the indoor motion state; the product between the emission probability and the state recognition probability is taken as the movement probability of pedestrians moving to any adjacent preset node.

Among them, the emission probability is the probability of generating an observation state from a hidden state, the position information of the pedestrian before moving is a hidden state, and the position information of the adjacent preset nodes is the observation state. When calculating the emission probability, the following formula (6) can be referred to:

In the above formula (6), P(z _t |r _i ) is the emission probability. z _t is the position information of the adjacent preset nodes, _ri is the position information before the pedestrian moves, and z _t - _ri is the Euclidean distance between the two. σ is the standard deviation of the calculated value of the pedestrian moving distance, and the value of σ in this embodiment is 0.1.

Based on the relevant content of the motion recognition probability matrix in the above step 203, for any adjacent preset node, when determining the state recognition probability that the motion state of the adjacent preset node is an indoor motion state, it can be determined according to the indoor motion state. The motion feature of , and the motion feature of the adjacent preset node are searched for the corresponding probability in the motion recognition probability matrix. Use the found probability as the state recognition probability.

In addition, when calculating the movement probability of pedestrians moving to any adjacent preset node according to the emission probability and the state recognition probability, the calculation process can refer to the following formula (7):

P(z _t ,m _t |r _i )=P(z _t |r _i )P(m _t |r _i ) (7)

Among them, P(z _t |r _i ) is the emission probability, P(m _t |r _i ) is the state recognition probability, and P(z _t ,m _t |r _i ) is the movement probability.

Since there may be multiple preset nodes adjacent to the pedestrian's position before moving in the indoor environment map model, the movement probability of each adjacent preset node can be calculated according to the above calculation process. After a plurality of movement probabilities are obtained by calculation, all movement probabilities may be sorted, and two movement probabilities with the largest numerical values are selected from the sorting results, and are counted as the first movement probability and the second movement probability. The first movement probability is greater than the second movement probability, that is, the first movement probability is the maximum value of the movement probabilities, and the second movement probability is the second largest value of the movement probabilities.

In order to more accurately determine the position information of the pedestrian, the ratio of the first movement probability to the second movement probability may be calculated. When the ratio is greater than the third preset threshold, it indicates that the possibility of the pedestrian moving to the adjacent preset node corresponding to the first movement probability is much greater than the possibility of moving to the adjacent preset node corresponding to the second movement probability. Therefore, it can be considered that when the pedestrian moves to the adjacent preset node corresponding to the first movement probability, the reliability is high. At this time, the position information of the adjacent preset nodes corresponding to the first movement probability may be used as the final position information of the pedestrian, and the position information predicted in the above steps 201 to 202 is discarded.

In addition, when the ratio of the first movement probability to the second movement probability is not greater than the third preset threshold, the pedestrian position information predicted in the above steps 201 to 202 may be used as the pedestrian's final position information.

In the method provided by the embodiment of the present invention, the total number of steps between the pedestrian starting to move and the stop moving is determined. For each step of the pedestrian's movement, according to the pedestrian's step length, steering angle and the position information before the pedestrian's movement, calculate the position information after the pedestrian moves one step, until the number of calculations reaches the total number of steps, and the final calculation result is used as the pedestrian's position information. Based on the indoor motion model, the indoor motion state of the pedestrian is obtained. Based on the indoor environment map model, the predicted pedestrian position information is calibrated according to the indoor motion state and the position information of the indoor preset nodes, and the final position information of the pedestrian is obtained. Since there is no need to install external equipment, while avoiding designing a system with high complexity, the hardware cost consumption can also be reduced, so that the cost of indoor positioning is lower.

In addition, the accelerometer, barometer and gyroscope data are selected during the classification of motion features, which improves the accuracy of classification of motion features and avoids the occurrence of long-term accumulated errors. Since the positioning process combines the pedestrian dead reckoning algorithm, the indoor pedestrian motion feature and the hidden Markov model matching method, the robustness of indoor positioning can be improved while ensuring a high positioning accuracy.

An embodiment of the present invention provides an indoor positioning apparatus, and the apparatus is configured to execute the indoor positioning method provided in the embodiment corresponding to FIG. 1 or FIG. 2 above. Referring to Figure 3, the device includes:

The prediction module 301 is used for predicting the position information of pedestrians according to the data collected by multiple sensors;

an acquisition module 302, configured to acquire the indoor motion state of the pedestrian based on the indoor motion model;

The calibration module 303 is configured to calibrate the predicted pedestrian position information based on the indoor environment map model, according to the indoor motion state and the position information of the indoor preset nodes, to obtain the final position information of the pedestrian.

As an optional embodiment, the prediction module 301 includes:

A determination unit for determining the total number of steps the pedestrian takes between starting to move and stopping;

The first calculation unit is used to calculate the position information after the pedestrian moves one step according to the step length, the steering angle and the position information before the pedestrian moves for each step of the pedestrian's movement, until the number of calculation times reaches the total number of steps, and the final calculation is performed. The result is used as the location information of the pedestrian.

As an optional embodiment, the prediction module 301 further includes:

The detection unit is used to detect the pedestrian's start and stop movement according to the acceleration of each sampling point.

As an optional embodiment, the detection unit is configured to, for any sampling point, determine that the pedestrian starts to move at any sampling point when it is detected that the acceleration of any sampling point is not less than the first preset threshold.

As an optional embodiment, the detection unit is configured to, for any sampling point, detect that the acceleration of any sampling point is smaller than the first preset threshold value, to detect that the acceleration of any sampling point is less than the first preset threshold value continuously since any sampling point The number of sampling points is counted; when the statistical result reaches the preset number, the last sampling point when the statistical result reaches the preset number is obtained, and it is determined that the pedestrian stops moving at the last sampling point.

As an optional embodiment, the determining unit is configured to, for any sampling point during the period from the start of the pedestrian to the stop of the pedestrian movement, when the acceleration of any sampling point is detected to be greater than the second preset threshold, and any When the acceleration corresponding to the next sampling point of the sampling point is less than the second preset threshold, the previous sampling point of any sampling point, any sampling point and the next sampling point of any sampling point are regarded as a step cycle, and the Add one to the pedestrian's total steps.

As an optional embodiment, the prediction module 301 further includes:

The second calculation unit is configured to calculate the second preset threshold value according to the acceleration related value in the previous step cycle.

As an optional embodiment, the prediction module 301 further includes:

The acquisition unit is used to acquire the angular acceleration of the current pedestrian in three directions based on the space coordinate system;

The third calculation unit is configured to calculate the steering angle of the pedestrian according to the current angular acceleration and acceleration of the pedestrian in three directions based on the projection relationship between the space coordinate system and the ground coordinate system.

As an optional embodiment, the prediction module 301 is configured to determine the floor where the pedestrian is located according to the air pressure value at the location where the pedestrian is located.

As an optional embodiment, the obtaining module 302 is configured to calculate the characteristic value corresponding to the first time window according to the acceleration of each sampling point in the first time window; based on the motion state classifier, according to the acceleration corresponding to the first time window The eigenvalues determine the motion characteristics of pedestrians.

As an optional embodiment, the obtaining module 302 is configured to calculate the turning angle of the pedestrian according to the turning angle of the pedestrian in the second time window; and determine the turning characteristic of the pedestrian according to the turning angle of the pedestrian.

As an optional embodiment, the calibration module 303 includes:

a determining unit, configured to determine the movement probability of pedestrians moving to adjacent preset nodes in the indoor environment map model based on the indoor environment map model;

The selection unit is used to sort the movement probabilities of each adjacent preset node, and selects the two movement probabilities with the largest values in the sorting result, which are respectively the first movement probability and the second movement probability, and the first movement probability is greater than the second movement probability. move probability;

The first comparison unit is configured to use the position information of the adjacent preset nodes corresponding to the first movement probability as the final position information of the pedestrian when the ratio of the first movement probability to the second movement probability is greater than the third preset threshold.

As an optional embodiment, the determining unit is configured to, for any adjacent preset node in the indoor environment map model, calculate the movement of the pedestrian to any adjacent preset node according to the position information of any adjacent preset node According to the motion recognition probability matrix, determine the state recognition probability that the motion state of any adjacent preset node represents the indoor motion state; take the product between the emission probability and the state recognition probability as a pedestrian moving to any adjacent node The movement probability of the preset node.

As an optional embodiment, the calibration module 303 further includes:

The second comparison unit is configured to use the predicted pedestrian position information as the pedestrian's final position information when the ratio of the first movement probability to the second movement probability is not greater than the third preset threshold.

The device provided by the embodiment of the present invention predicts the location information of pedestrians by using data collected by multiple sensors. Based on the indoor motion model, the indoor motion state of the pedestrian is obtained. Based on the indoor environment map model, the predicted pedestrian position information is calibrated according to the indoor motion state and the position information of the indoor preset nodes, and the final position information of the pedestrian is obtained. Since there is no need to install external equipment, while avoiding designing a system with high complexity, the hardware cost consumption can also be reduced, so that the cost of indoor positioning is lower.

In addition, the accelerometer, barometer and gyroscope data are selected during the classification of motion features, which improves the accuracy of classification of motion features and avoids the occurrence of long-term accumulated errors. Since the positioning process combines the pedestrian dead reckoning algorithm, the indoor pedestrian motion feature and the hidden Markov model matching method, the robustness of indoor positioning can be improved while ensuring a high positioning accuracy.

Finally, the method of the present application is only a preferred embodiment, and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (6)

1. an indoor positioning method, is characterized in that, described method comprises: According to the data collected by multiple sensors, predict the location information of pedestrians; obtaining the indoor motion state of the pedestrian based on the indoor motion model; Based on the indoor environment map model, according to the indoor motion state and the position information of the indoor preset nodes, the predicted pedestrian position information is calibrated, and the final position information of the pedestrian is obtained; The indoor environment map model, according to the indoor motion state and the position information of the indoor preset nodes, calibrates the predicted pedestrian position information, and obtains the pedestrian's final position information, including: determining, based on the indoor environment map model, the movement probability of the pedestrian moving to an adjacent preset node in the indoor environment map model; Sort the movement probabilities of each adjacent preset node, and select the two movement probabilities with the largest values in the sorting result, which are the first movement probability and the second movement probability, and the first movement probability is greater than the second movement probability. probability; When the ratio of the first movement probability to the second movement probability is greater than a third preset threshold, the position information of the adjacent preset node corresponding to the first movement probability is used as the final position information of the pedestrian. 2. The method according to claim 1, wherein the obtaining the indoor motion state of the pedestrian based on an indoor motion model comprises: Calculate the characteristic value corresponding to the first time window according to the acceleration of each sampling point in the first time window; Based on the motion state classifier, the motion feature of the pedestrian is determined according to the feature value corresponding to the first time window. 3. The method according to claim 1, wherein the obtaining the indoor motion state of the pedestrian based on an indoor motion model comprises: calculating the turning angle of the pedestrian according to the turning angle of the pedestrian in the second time window; According to the turning angle of the pedestrian, the turning characteristic of the pedestrian is determined. 4 . The method according to claim 1 , wherein the determining the movement probability of the pedestrian moving to an adjacent preset node in the indoor environment map model based on the indoor environment map model comprises: 5 . For any adjacent preset node in the indoor environment map model, according to the position information of the any adjacent preset node, calculate the emission probability of the pedestrian moving to the any adjacent preset node, so The emission probability is the probability of generating an observation state from a hidden state, the position information before the pedestrian moves is a hidden state, and the position information of any adjacent preset node is an observation state; According to the motion recognition probability matrix, the motion state of any adjacent preset node is determined as the state recognition probability of the indoor motion state, and the motion recognition probability matrix is the recognized pairs and recognized pairs of each motion feature of pedestrians is the probability matrix of other motion features; The product between the emission probability and the state identification probability is taken as the moving probability of the pedestrian moving to any of the adjacent preset nodes. 5. The method according to claim 1, characterized in that, the moving probability of each adjacent preset node is sorted, and the two moving probabilities with the largest numerical value in the sorting result are selected, which are respectively the first moving probability and the After the second move probability, it also includes: When the ratio of the first movement probability to the second movement probability is not greater than a third preset threshold, the predicted pedestrian position information is used as the pedestrian's final position information. 6. An indoor positioning device, wherein the device comprises: The prediction module is used to predict the location information of pedestrians according to the data collected by multiple sensors; an acquisition module for acquiring the indoor motion state of the pedestrian based on the indoor motion model; a calibration module, configured to calibrate the predicted pedestrian position information based on the indoor environment map model and according to the indoor motion state and the position information of the indoor preset nodes to obtain the final position information of the pedestrian; The calibration module includes: a determining unit, configured to determine, based on the indoor environment map model, the movement probability of the pedestrian moving to an adjacent preset node in the indoor environment map model; The selection unit is used to sort the movement probability of each adjacent preset node, and selects the two movement probabilities with the largest numerical value in the sorting result, which are respectively the first movement probability and the second movement probability, and the first movement probability is greater than the second movement probability; a first comparison unit, configured to use the position information of the adjacent preset node corresponding to the first movement probability as the pedestrian when the ratio of the first movement probability to the second movement probability is greater than a third preset threshold final location information.