A human indoor positioning system and method based on particle filtering

Abstract

The present invention discloses a human body indoor positioning system and method based on particle filtering. The system is fixedly installed on the knee of the target human body, and includes a micro camera, an inertial sensor, a microprocessor and a communication module; the micro camera is used to obtain image data of the target human body during movement, the inertial sensor is used to obtain the position and posture data of the target human body during movement, and the microprocessor is used to fuse the image data and the position and posture data to obtain the positioning result of the target human body. The method integrates the inertial positioning technology with the monocular vision positioning technology through the particle filtering algorithm, and optimizes the method of human body positioning based solely on inertial sensors to a large extent. The positioning result has the disadvantage of continuous accumulation of errors over time, and realizes indoor autonomous and precise positioning and navigation of pedestrians in the absence of satellite navigation signals, and is less affected by the environment, and the reliability of the positioning process is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of human body motion positioning and indoor navigation, and more particularly to a human body indoor positioning system and method based on particle filtering. Background Art

[0002] At present, with the continuous development of emerging fields such as smart elderly care, smart home, sports and fitness, personal indoor positioning technology is particularly important. Since the human body is indoors, the signals of commonly used satellite navigation technologies such as GPS are blocked by buildings, resulting in extremely poor positioning results. Inertial navigation does not rely on the external environment and is a completely autonomous navigation method. Therefore, personal indoor positioning often uses inertial navigation solutions based on inertial sensors.

[0003] However, commonly used MEMS inertial sensors are affected by factors such as device drift, and the positioning error caused by inertial devices will continue to accumulate over time. Therefore, when the positioning time becomes longer, the positioning error of this system that relies solely on MEMS inertial sensors for positioning often increases to an unacceptable level, and the positioning effect is not ideal. The human motion positioning accuracy based on monocular vision is higher than that of inertial navigation solutions, but this positioning method also has high requirements for lighting, indoor environment, etc., and has weak environmental adaptability, making it difficult to adapt to the changing indoor positioning needs.

[0004] Therefore, how to provide a more adaptable, accurate and reliable method for indoor human body positioning is an urgent problem to be solved by those skilled in the art. Summary of the invention

[0005] In view of this, the present invention provides a human indoor positioning system and method based on particle filtering, which realizes accurate positioning of the human body indoors by integrating inertial positioning technology and monocular vision positioning technology, and solves the problems of large errors and weak environmental adaptability of existing indoor positioning methods.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] On the one hand, the present invention provides a human indoor positioning system based on particle filtering, which is fixedly installed on the knee of a target human body, and includes: a micro camera, an inertial sensor, a microprocessor and a communication module, wherein the micro camera and the inertial sensor are both electrically connected to the microprocessor, and the microprocessor is electrically connected to the communication module;

[0008] The micro camera is used to obtain image data of the target human body during movement, the inertial sensor is used to obtain posture data of the target human body during movement, the microprocessor is used to fuse the image data and the posture data to obtain the target human body positioning result, and the communication module is used to send the target human body positioning result to an external terminal device for real-time display.

[0009] Further, the inertial sensor includes an accelerometer and a gyroscope, and the accelerometer and the gyroscope are both electrically connected to the microprocessor;

[0010] The accelerometer is used to collect acceleration data of the target human body during movement in real time, the gyroscope is used to collect posture data of the target human body during movement in real time, and the microprocessor is used to analyze the position and posture state of the target human body during movement according to the acceleration data and the posture data.

[0011] In actual use, the system is bound to the human knee, and the image data obtained by the micro camera is sent to the microprocessor through the video interface. The sensor data obtained by the inertial sensor is also sent to the microprocessor through the sensor interface. After the microprocessor fuses the image signal and the sensor data respectively, the human body positioning result is finally obtained, and the human body positioning result is returned to the smart terminal device through the communication module for real-time display.

[0012] On the other hand, the present invention also provides a method for indoor human body positioning based on particle filtering, which is based on the above-mentioned human body indoor positioning system based on particle filtering, and includes:

[0013] Acquire the position and posture data of the target human body during the movement, and calculate the movement distance of the target human body according to the position and posture data;

[0014] According to the pre-obtained initial position and initial azimuth of the target human body and the calculated movement distance, the position of the target human body after the movement is obtained;

[0015] Acquire image data of the target human body during motion, and estimate the position and posture of the micro-camera during motion based on the image data and the camera internal parameters of the micro-camera used to capture the image;

[0016] According to the estimated position of the micro camera, the position of the micro camera after the movement is obtained, and then the position of the target human body after the movement is obtained;

[0017] The target human position information obtained based on the posture data and the target human position information obtained based on the image data are fused through the particle filtering algorithm to obtain the final target human position after movement.

[0018] Furthermore, the process of obtaining the position and posture data of the target human body during the movement, and calculating the movement distance of the target human body according to the position and posture data specifically includes:

[0019] Acquire the attitude data output by the gyroscope, and construct an attitude conversion matrix from the gyroscope sensor coordinate system to the ground coordinate system according to the attitude data;

[0020] Acquire acceleration data output by the accelerometer, and convert the acceleration data into acceleration in a ground coordinate system through the attitude conversion matrix;

[0021] Calculating the acceleration in the moving direction of the target human body according to the acceleration in the ground coordinate system;

[0022] The acceleration in the moving direction of the target human body is numerically integrated to obtain the moving speed of the target human body, and the moving speed is numerically integrated to obtain the moving distance of the target human body.

[0023] At the beginning of positioning, the initial azimuth of the human body is calculated based on the output data of the accelerometer and gyroscope during the standing stage of the target human body. Knowing the initial position of the subject, when walking, the gyroscope attitude is calculated in real time using the output data of the gyroscope and accelerometer using quaternions, thereby obtaining the attitude conversion matrix from the gyroscope sensor coordinate system to the ground coordinate system. The attitude conversion matrix is:

[0024]

[0025] in, is the yaw angle of the gyroscope, ρ is the pitch angle of the gyroscope, and γ is the roll angle of the gyroscope.

[0026] Furthermore, the process of acquiring image data of the target human body during motion, and estimating the position and posture of the micro camera during motion according to the image data and the camera intrinsic parameters of the micro camera used to capture the image, specifically includes:

[0027] Calibrate the micro camera used to capture images and obtain the camera's intrinsic parameters;

[0028] Based on the camera intrinsic parameters, construct a grayscale residual function with the inter-frame pose as an independent variable;

[0029] Obtain image data of the target human body during motion, and construct a least squares objective function with inter-frame posture as the optimization variable based on all pixel points in the previous frame image;

[0030] The least square objective function is iterated to obtain the optimal inter-frame pose and the pose of the micro camera.

[0031] Furthermore, the grayscale residual function is:

[0032] δI(T k,k-1 ,p i )=I k (p′ i )-I k-1 (p i )

[0033] Where p′ i Represents the map point p i The projection point on the current frame k, T k,k-1 represents the inter-frame pose, and the function I(·) represents the grayscale value of a point in the image;

[0034] Among them, p i and p′ i The relationship between them is:

[0035] p′ i =π(T k,k-1 (π -1 (p i ,z)))

[0036] In the formula, the function π(·) represents the transformation from the camera coordinate system to the pixel coordinate system, π -1 (·) represents the inverse transformation, T k,k-1 (·) is the function representing the posture transformation, T k,k-1 Represents the inter-frame pose.

[0037] Furthermore, the least squares objective function is:

[0038]

[0039] Where, T k,k-1 represents the inter-frame pose, p i Represents a map point, where the map point represents the imaging point whose three-dimensional coordinates in the camera coordinate system are determined by the image pixel corresponding to the image pixel in the camera environment. Function I(·) represents the grayscale value of a point in the image, and δ represents the grayscale difference of the pixel projected by the map point in two adjacent frame images.

[0040] Since the above formula is a nonlinear function, only nonlinear optimization can be used to continuously iterate and update the initial posture to make the objective function optimal. The present invention uses the Gauss-Newton method to continuously iterate the objective function to obtain the inter-frame posture, and the real-time position parameters of the camera can be obtained from the camera posture.

[0041] Furthermore, the target human body position information obtained based on the posture data and the target human body position information obtained based on the image data are fused by a particle filter algorithm to obtain the final target human body position after the movement, which specifically includes:

[0042] In the initialization phase, M particles X are randomly generated according to Gaussian distribution. i , the randomly generated particle state value is X i,0 =(x i,0 ,y i,0 ), i = 1, ..., M, and assign the same weight to each particle;

[0043] According to the principle of pedestrian dead reckoning, the human body positioning state transfer equation is determined;

[0044] According to the human body positioning state transfer equation, obtaining the predicted position of the particle at the next moment;

[0045] The obtained position of the micro camera is used as the observed position of the particle, and the Euclidean distance between the predicted position and the observed position of each particle is calculated;

[0046] Assuming that the observed probability density is Gaussian distribution, calculate the update coefficient of particle weight, update the weight of each particle according to the update coefficient of particle weight, and normalize the particle weight;

[0047] According to the normalized particle weights, the average value of the weighted sum of all particle state values ​​is calculated to obtain the final target human body position after movement.

[0048] Furthermore, the calculation formula for the average value of the weighted sum of all particle state values ​​is:

[0049]

[0050] Where M is the total number of particles, w i,k is the normalized weight of the i-th particle at time k, X i,k is the predicted position of the ith particle at time k.

[0051] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a human indoor positioning system and method based on particle filtering, which integrates inertial positioning technology with monocular vision positioning technology through a particle filtering algorithm, thereby optimizing to a great extent the method of human positioning based solely on inertial sensors, which has the disadvantage of the continuous accumulation of positioning result errors over time. It realizes indoor autonomous and precise positioning and navigation of pedestrians in the absence of satellite navigation signals, and is less affected by the environment, and the positioning process is more reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0053] Figure 1 A schematic diagram of the structural architecture of a human indoor positioning system based on particle filtering provided by the present invention;

[0054] Figure 2 The present invention provides a schematic diagram of the implementation process of a particle filtering-based human body indoor positioning method. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] On the one hand, see Appendix Figure 1 The embodiment of the present invention discloses a human indoor positioning system based on particle filtering. The system is fixedly installed on the knee of the target human body. In this embodiment, the system is bound to the human body knee. The system specifically includes: a micro camera 1, an inertial sensor 2, a microprocessor 3 and a communication module 4. The micro camera 1 and the inertial sensor 2 are both electrically connected to the microprocessor 3, and the microprocessor 3 is electrically connected to the communication module 4;

[0057] The micro camera 1 is used to obtain image data of the target human body during movement, the inertial sensor 2 is used to obtain the position and posture data of the target human body during movement, the microprocessor 3 is used to fuse the image data and the position and posture data to obtain the target human body positioning result, and the communication module 4 is used to send the target human body positioning result to an external terminal device for real-time display.

[0058] Specifically, the inertial sensor 2 includes an accelerometer 21 and a gyroscope 22, and both the accelerometer 21 and the gyroscope 22 are electrically connected to the microprocessor 3;

[0059] The accelerometer 21 is used to collect the acceleration data of the target human body during the movement in real time, the gyroscope 22 is used to collect the posture data of the target human body during the movement in real time, and the microprocessor 3 is used to analyze the posture state of the target human body during the movement according to the acceleration data and the posture data.

[0060] In this embodiment, the communication module 4 adopts a Bluetooth module, and the intelligent terminal device adopts a smart phone, which can realize short-range wireless communication between the system and the smart phone.

[0061] The monocular vision positioning module in the microprocessor is an important part of the whole system, which mainly includes the acquisition of feature points of the previous and next two frames of images, feature point matching, calculation of the camera transformation matrix, and camera position positioning.

[0062] On the other hand, see Appendix Figure 2 The embodiment of the present invention further discloses a method for indoor human body positioning based on particle filtering. The method is based on the above-mentioned human body indoor positioning system based on particle filtering, and includes:

[0063] S1: Obtaining the posture data of the target human body during movement, and calculating the movement distance of the target human body according to the posture data;

[0064] The initial azimuth of the human body is calculated from the output data of the accelerometer and gyroscope during the standing phase of the target human body, and the initial position of the subject is known. At the beginning of walking, in the swing phase, the output of the gyroscope and accelerometer is used to solve the gyroscope attitude in real time using quaternions, thereby obtaining the attitude conversion matrix from the gyroscope sensor coordinate system to the ground coordinate system:

[0065]

[0066] in, is the yaw angle of the gyroscope, ρ is the pitch angle of the gyroscope, and γ is the roll angle of the gyroscope.

[0067] Get the acceleration data output by the accelerometer, and convert the output of the accelerometer into acceleration in the ground coordinate system through the attitude transformation matrix, that is,

[0068]

[0069] The acceleration in the direction of human body movement is obtained as:

[0070]

[0071] Then the motion speed is obtained by numerical integration, that is,

[0072]

[0073] The movement distance is further obtained by numerical integration of the movement speed, and the calculation formula is as follows:

[0074]

[0075] Where, T is the sampling period of the accelerometer; a g (t), vg (t), d g (t) are the acceleration, speed and movement distance at the current moment; a g (tT),v g (tT), d g (tT) are the acceleration, velocity and movement distance at the last sampling period.

[0076] S2: according to the pre-obtained initial position and initial azimuth of the target human body and the calculated movement distance, the position of the target human body after the movement is obtained;

[0077] This embodiment uses the knee motion acceleration amplitude The peak detection algorithm is used to detect the minimum peak value of acceleration a. When the acceleration amplitude reaches the minimum peak value, it is considered to be the standing phase. Since the movement speed is 0 in the standing phase, the movement speed is set to 0 at this time to eliminate the accumulated error of inertial navigation positioning. When the next step comes, the same algorithm is executed to calculate the movement distance and obtain the position of the target human body in the next step. Then the position result of the target human body obtained in the previous step is added to the movement displacement of the current step to obtain the position of the human body after each movement.

[0078] S3: Obtain image data of the target human body during motion, and estimate the position and posture of the micro camera during motion based on the image data and the camera internal parameters of the micro camera used to capture the image;

[0079] Firstly, the micro camera is calibrated using the Matlab calibration toolbox to obtain the camera intrinsic parameter K. The camera pose in the micro camera is estimated using the grayscale invariant assumption.

[0080] Construct the inter-frame pose T k,k-1 Is the gray residual function of the independent variable:

[0081] δI(T k,k-1 ,p i )=I k (p′ i )-I k-1 (p i )

[0082] Among them, p′ i Represents the map point p i At the projection point on the current frame k, the function I(·) represents the gray value of a point in the image, p i and p′ i The relationship between the frame poses T k,k-1 Get, specifically:

[0083] p′ i =π(T k,k-1 (π -1(p i ,z)))

[0084] Among them, the function π(·) represents the transformation from the camera coordinate system to the pixel coordinate system, π -1 (·) represents the inverse transformation, T k,k-1 (·) is a function representing the pose transformation.

[0085] The calculation formula for the inter-frame pose is:

[0086]

[0087] Among them, R k,k-1 represents the rotation matrix of the camera from frame k-1 to frame k, R∈SO(3), t k,k-1 Represents the translation matrix of the camera from frame k-1 to frame k.

[0088] Specifically, firstly, the pixel point p in the previous frame i Combine the depth z of the corresponding map point P to restore it to the camera coordinate system of the previous frame. Then restore the point to the position T k,k-1 Transform it down to the camera coordinate system of the current frame, and finally project the point from the camera coordinate system to the pixel coordinate system to obtain p'.

[0089] Solve the problem of T for all pixels in the previous frame k,k-1 For the least squares problem of optimizing variables:

[0090]

[0091] Since the above formula is a nonlinear function, only nonlinear optimization can be used to continuously iterate and update the initial posture to make the objective function optimal. This embodiment uses the Gauss-Newton method to continuously iterate the objective function to obtain T k,k-1 , the real-time position parameters of the camera can be obtained from the camera pose.

[0092] S4: According to the estimated position of the micro camera, the position of the micro camera after the movement is obtained, and then the position of the target human body after the movement is obtained.

[0093] S5: fusing the target human position information obtained based on the posture data and the target human position information obtained based on the image data through a particle filtering algorithm to obtain the final position of the target human body after the movement;

[0094] This embodiment uses a particle filtering algorithm to fuse the human body positioning result based on the inertial sensor and the human body positioning result based on the monocular vision.

[0095] In the initialization phase, M particles X are randomly generated according to Gaussian distribution. i , the randomly generated particle state value is Xi,0 =(x i,0 ,y i,0 ), i = 1, ..., M. Give each particle the same weight

[0096] According to the principle of pedestrian dead reckoning, the human body positioning state transfer equation is:

[0097]

[0098] Among them, l k and They represent the human body motion step length and yaw angle at time k. They are calculated from the output data of the inertial sensor. Therefore, according to the above human body positioning state transfer equation, the predicted state X of the particle at the next moment is obtained. i,k+1 =(x i,k+1 ,y i,k+1 ).

[0099] Since the inertial sensor has limited accuracy and is affected by drift, after the particle state is updated, the monocular vision positioning result is introduced as measurement information to update the particle weight.

[0100] Calculate the predicted position (x) of each particle i,k+1 ,y i,k+1 ) and observation position (x' k ,y' k )’s Euclidean distance:

[0101]

[0102] In the formula, the observation position (x' k ,y' k ) is the camera position detected by the monocular vision module, where t x,k Represents the translation vector t of the camera from frame k-1 to frame k k,k-1 The x-component of y,k Represents the translation vector t of the camera from frame k-1 to frame k k,k-1 The y component of t k,k-1 Obtained from the camera's pose matrix.

[0103] Assuming that the observed probability density is Gaussian distribution, the update coefficient p of the particle weight is calculated by the following formula: i,k :

[0104]

[0105] Where d is the mean square error of the measurement noise and is a constant set to 5.

[0106] The weight of the i-th particle at time k Update as follows:

[0107]

[0108] Normalizing the particle weights yields:

[0109]

[0110] Where M is the total number of particles. i,k is the normalized weight of the i-th particle at time k.

[0111] Return the weighted average of all particle states as the final state of the system:

[0112]

[0113] Where M is the total number of particles, w i,k is the normalized weight of the i-th particle at time k, X i,k is the predicted position of the ith particle at time k.

[0114] In summary, the above-mentioned system and method disclosed in the embodiments of the present invention can optimize the disadvantage of the continuous accumulation of errors in human body positioning based on inertial sensors over time. It can realize autonomous and precise positioning and navigation of pedestrians indoors without satellite navigation signals. It can realize autonomous indoor positioning of human bodies by using a sensor module fixed on the knee and an embedded processor.

[0115] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0116] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A human indoor positioning method based on particle filtering, It is characterized in that The method is based on a human indoor positioning system based on particle filtering, which is fixedly installed on the knee of the target human body and includes: a micro camera, an inertial sensor, a microprocessor and a communication module, wherein the micro camera and the inertial sensor are both electrically connected to the microprocessor, and the microprocessor is electrically connected to the communication module; The micro camera is used to obtain image data of the target human body during the movement, the inertial sensor is used to obtain the position data of the target human body during the movement, the microprocessor is used to fuse the image data and the position data to obtain the target human body positioning result, and the communication module is used to send the target human body positioning result to an external terminal device for real-time display; The inertial sensor includes an accelerometer and a gyroscope, and the accelerometer and the gyroscope are both electrically connected to the microprocessor; The accelerometer is used to collect acceleration data of the target human body during the movement in real time, the gyroscope is used to collect posture data of the target human body during the movement in real time, and the microprocessor is used to analyze the posture state of the target human body during the movement according to the acceleration data and the posture data; The method includes: S1 obtains the position and posture data of the target human body during the movement, and calculates the movement distance of the target human body according to the position and posture data; S2, according to the pre-obtained initial position and initial azimuth of the target human body and the calculated movement distance, obtaining the position of the target human body after the movement; S3 obtains image data of the target human body during the motion process, and estimates the position and posture of the micro camera during the motion process based on the image data and the camera internal parameters of the micro camera used to capture the image; S4 obtains the position of the micro camera after the movement according to the estimated position and posture of the micro camera, and then obtains the position of the target human body after the movement; S5 uses a particle filter algorithm to fuse the target human position information obtained based on the posture data and the target human position information obtained based on the image data to obtain the final position of the target human body after the movement; In S1, the process of obtaining the posture data of the target human body during the movement and calculating the movement distance of the target human body according to the posture data specifically includes: Acquire the attitude data output by the gyroscope, and construct an attitude conversion matrix from the gyroscope sensor coordinate system to the ground coordinate system according to the attitude data; Acquire acceleration data output by the accelerometer, and convert the acceleration data into acceleration in a ground coordinate system through the attitude conversion matrix; Calculating the acceleration in the moving direction of the target human body according to the acceleration in the ground coordinate system; The acceleration in the moving direction of the target human body is numerically integrated to obtain the moving speed of the target human body, and the moving speed is numerically integrated to obtain the moving distance of the target human body.

2. According to the particle filtering-based human body indoor positioning method of claim 1, It is characterized in that The attitude transformation matrix is: in, is the yaw angle of the gyroscope, ρ is the pitch angle of the gyroscope, and γ is the roll angle of the gyroscope.

3. According to the particle filtering-based human body indoor positioning method of claim 1, It is characterized in that The process of acquiring image data of a target human body during motion, and estimating the position and posture of the micro camera during motion according to the image data and the camera internal parameters of the micro camera used to shoot the image, specifically includes: Calibrate the micro camera used to capture images and obtain the camera's intrinsic parameters; Based on the camera intrinsic parameters, construct a grayscale residual function with the inter-frame pose as an independent variable; Obtain image data of the target human body during motion, and construct a least squares objective function with inter-frame posture as the optimization variable based on all pixel points in the previous frame image; The least square objective function is iterated to obtain the optimal inter-frame pose and the pose of the micro camera.

4. According to the particle filtering-based human body indoor positioning method of claim 3, It is characterized in that The grayscale residual function is: δI(T k,k-1 ,p i )=I k (p’ i )-I k-1 (p i ) Where p' i Represents the map point p i The projection point on the current frame k, T k,k-1 represents the inter-frame pose, and the function I(·) represents the grayscale value of a point in the image; Among them, p i With p' i The relationship between them is: p' i =π(T k,k-1 (p -1 (p i ,z))) In the formula, the function π(·) represents the transformation from the camera coordinate system to the pixel coordinate system, π -1 (·) represents the inverse transformation, T k,k-1 (·) is the function representing the posture transformation, T k,k-1 Represents the inter-frame pose.

5. The method for indoor human body positioning based on particle filtering according to claim 3, It is characterized in that The least squares objective function is: Where, T k,k-1 represents the inter-frame pose, p i represents a map point, function I(·) represents the grayscale value of a point in the image, and δ represents the grayscale difference of the pixel points projected by the map point in two adjacent frame images.

6. The method for indoor human body positioning based on particle filtering according to claim 1, It is characterized in that The process of fusing the target human position information obtained based on the posture data and the target human position information obtained based on the image data through the particle filter algorithm to obtain the final target human position after the movement specifically includes: In the initialization phase, M particles X are randomly generated according to Gaussian distribution. i , the randomly generated particle state value is X i,0 =(x i,0 ,y i,0 ), i = 1, ..., M, and assign the same weight to each particle; According to the principle of pedestrian dead reckoning, the human body positioning state transfer equation is determined; According to the human body positioning state transfer equation, obtaining the predicted position of the particle at the next moment; The obtained position of the micro camera is used as the observed position of the particle, and the Euclidean distance between the predicted position and the observed position of each particle is calculated; Assuming that the observed probability density is Gaussian distribution, calculate the update coefficient of particle weight, update the weight of each particle according to the update coefficient of particle weight, and normalize the particle weight; According to the normalized particle weights, the average value of the weighted sum of all particle state values ​​is calculated to obtain the final target human body position after movement.

7. A method for indoor human body positioning based on particle filtering according to claim 6, It is characterized in that The calculation formula for the average value of the weighted sum of all particle state values ​​is: Where M is the total number of particles, w i,k is the normalized weight of the i-th particle at time k, X i,k is the predicted position of the ith particle at time k.

Images