A multi-scenario adaptive B5G three-dimensional positioning method for smart medical care

Abstract

This invention discloses a multi-scenario adaptive B5G three-dimensional positioning method for smart healthcare, comprising: using the LBP-SVM algorithm to train a classification decision model based on signal data and category labels of reference points within a hospital during the offline phase; using an improved particle filter algorithm to perform position calculation based on the classification decision model to obtain positioning information of the terminal to be located; and feeding this positioning information back to the terminal to be located. The invention classifies the terminals to be located based on the hospital's zoning and uses particle filtering to convert motion states into static states, thereby improving positioning accuracy and, to a certain extent, being more adaptable to complex environments.

Description

Technical Field

[0001] The present invention belongs to the field of indoor positioning technology, and specifically relates to a multi-scenario adaptive B5G three-dimensional positioning method for smart medical care. Background Art

[0002] In large hospitals, real-time personnel location and navigation are the most common applications of positioning systems. Hospitals are mobile environments, and medical staff's locations are constantly changing due to work demands. Lack of clear personnel management in medical institutions often leads to the following problems: wasted manpower, resulting in significant daily time spent searching for nurses and clinicians; delayed patient care, with required doctors and nurses unable to be found promptly, delaying emergency care; and low service quality, with patients unable to obtain timely medical information, leading to inefficient services and medical accidents. By implementing a real-time personnel location guidance system, doctors can view patients' real-time location information on a positioning server interface. This saves valuable time for diagnosis and treatment, and also enables management of medical staff's work status, such as automatically generating and querying information about a doctor's ward rounds and their movements. Furthermore, it can provide guidance for patient examination routes and family visits. The goal is to improve patient care, reduce the frequency of accidents, and enhance hospital management.

[0003] Hospitals are complex environments. To provide better navigation services, whether for medical staff or patients, it is crucial to be able to determine the location of mobile individuals. Wireless positioning technology is currently well-established in the outdoor research field and has been widely applied in various applications. Satellite-based GPS is currently the most popular outdoor positioning technology. However, due to satellite signal obstruction, GPS cannot guarantee proper operation in complex indoor areas, making indoor positioning unavailable. With the increasing demand for indoor positioning in various fields, there is an urgent need for the development of indoor positioning technology. Although fingerprint-based indoor positioning methods offer satisfactory accuracy, they often face unreliability challenges due to the temporal and spatial variations of RSS measurements and the influence of human movement and obstacles. Therefore, there is an urgent need to develop indoor positioning and tracking technologies with sufficient accuracy and limited latency. Particle filtering (PF) technology has been used in Wi-Fi location fingerprinting systems to assist in target tracking, reducing the mobile tracking problem to a sequence of position estimates for stationary individuals.

[0004] The invention with publication number CN114390462A discloses an indoor positioning method based on RSSI ranging, including: step 1 constructing a map, establishing a geographic information layer according to the road area type, and storing it as a geographic information layer file in a fixed format; step 2 knowing the distance from the signal transmitter to the reference node, measuring the RSSI value at the reference node location, and using a wireless positioning algorithm based on ranging to obtain the position coordinates of the target node to be measured based on a logarithmic path loss model, thereby realizing indoor positioning based on RSSI ranging; step 3 according to the type of road area where the target is located, bringing the geographic information of the road as a constraint condition into the corresponding Kalman filter equation, and constructing a target motion model based on heading constraints; step 4 according to the type of road area where the target is located, bringing the geographic information of the road as a constraint condition into the corresponding Kalman filter equation, and performing particle filtering based on map constraints. The invention, with publication number CN112153569A, discloses an optimization method based on indoor fingerprint positioning, including: S1, discretely gridding the positioning area, deploying N APs, and determining appropriate grid point spacing; S2, selecting M reference points based on the spacing, collecting RSSIs between each reference point and the AP to form a dataset X, which is then preprocessed; S3, dividing the reference point region IDs and constructing a fingerprint database; S4, selecting a real path, collecting RSSIs between T to-be-located points and the APs along the real path; S5, using the SVM algorithm to determine the ID of the to-be-located region based on the dataset X and the reference point region ID; S6, calculating the RSSI similarity between the to-be-located point and the reference points within the region after region classification using Euclidean distance, Manhattan distance, and Chebyshev distance to obtain a position estimate; S7, using the position estimate as an observation value and combining it with the PDR algorithm for particle filtering to obtain accurate positioning coordinates. This invention utilizes a region classification algorithm to reduce the search space size and improve efficiency, while also optimizing the matching algorithm to enhance positioning accuracy.

[0005] However, most indoor wireless positioning algorithms based on wireless maps use static wireless maps, where RSS is assumed to be stable over time, ignoring the fingerprint's sensitivity to environmental changes. This can lead to large positioning errors. Traditional tracking methods only use current observations to estimate a single location, without effectively utilizing continuous motion information. Traditional particle filter sampling methods consider the information provided by individual motion trajectories. In smart healthcare ward scenarios, medical institutions suffer from wasted manpower, delayed patient services, low service quality, and low positioning accuracy. Summary of the Invention

[0006] Technical problems solved: In response to the above technical problems, the present invention provides a semi-spherical area model for smart medical ward scenarios that can improve the positioning accuracy in the online stage; the multi-scenario adaptive B5G three-dimensional positioning method for smart medical care of the present invention can not only provide better accuracy and communication speed, but also cover a wider range, and achieve higher cost performance and system stability.

[0007] Technical solution:

[0008] A multi-scenario adaptive B5G three-dimensional positioning method for smart medical care, the positioning method comprising the following steps:

[0009] S1, obtaining the signal data of the positioning point currently collected by the terminal to be positioned in the hospital;

[0010] S2: Obtain the signal data and positions of reference points in the hospital during the offline phase, and label the reference points according to regions. Perform LBP feature extraction on the signal data of the reference points in the region to obtain the LBP data features of the reference points in the region to which they belong, and calculate the membership of the reference points with respect to each region. Input the RSSI signal data of the reference points in the region and the membership of the reference points with respect to each region into the SVM, and train K classification decision models, where K is the number of regions in the hospital.

[0011] S3: Based on the normal distribution of specific parameters, particle RSSI signal data is generated based on the RSSI signal data of the historical positioning points of the terminal to be located. Based on the classification decision model, it is determined whether the category of the area to which the terminal to be located belongs is consistent with the category of the area to which the generated particles belong, and a higher weight is assigned to the generated particles that belong to the same area. Based on the movement trend changes described by the historical positioning points of the terminal to be located, according to the movement patterns of the human body, and according to the area to which the terminal to be located belongs, an optimal sampling area with a large posterior probability density is constructed, and a higher weight is assigned to the generated particles within the optimal sampling area. Based on the resampling concept, the particle positions are weighted according to the particle weight to obtain the positioning information of the terminal to be located.

[0012] S4: Feedback the acquired positioning information to the terminal to be positioned.

[0013] Furthermore, in step S2, the process of obtaining the LBP data features of the reference point in the region to which it belongs and calculating the membership of the reference point with respect to each region includes the following sub-steps:

[0014] S201, select one of the reference point RSSI values and use it as the center RSSI value;

[0015] S202, obtaining several adjacent RSSI values centered on the selected RSSI value based on a circular LBP operator, and sequentially comparing the obtained several adjacent RSSI values with the central RSSI value;

[0016] S203, marking adjacent RSSI values whose RSSI values are greater than the central RSSI value as 1, and marking adjacent RSSI values whose RSSI values are less than the central RSSI value as 0; extracting the marks of the adjacent RSSI values in a predetermined order to form a binary digital string, and using the binary digital string as the LBP feature of the central RSSI value;

[0017] S204, repeating steps S201 to S203, traversing each RSSI value in the area, and obtaining the LBP feature of the RSSI value of each reference point in the area;

[0018] S205, using the log-likelihood ratio similarity to calculate the membership degree of each reference point in the region according to the following formula:

[0019]

[0020]

[0021]

[0022] s l =2*(matrixEntropy-rowEntropy-columEntropy)

[0023] in, is the nth RSSI eigenvalue of the zth reference point in the region, Z is the number of reference points in the region, N is the number of RSSI eigenvalues of the reference points; rowEntropy is the row entropy; columEntropy is the column entropy; matrixEntropy is the matrix entropy; s l is the log-likelihood ratio similarity.

[0024] Furthermore, in step S2, the RSSI signal data of the reference points in the region and the membership of the reference points with respect to each region are input into the SVM, and the process of training and obtaining K classification decision models includes the following sub-steps:

[0025] S211, using the reference point RSSI signal data as a training set for a support vector machine to obtain a regional classification decision model; the input of the classification decision model is the reference point RSSI and the membership degree s of the reference point to each region. l ; Assume that the reference points belonging to this area are positive samples, and the rest are negative samples; the optimization problem of the support vector to plane distance formula is as follows:

[0026]

[0027]

[0028] Among them, the parameter ω k , b k are the normal vector and intercept of the hyperplane of the kth class respectively; Indicates mapping the input of the linearly inseparable low-dimensional space to a linearly separable high-dimensional space, y l The label class of the RSSI value of the lth reference point, RSSI l is the RSSI value of the lth reference point, θ l is the relaxation factor, C is the penalty factor; l=1,2,…,L, L is the total number of reference points;

[0029] S212, introduce the kernel function K(X l ,X), generalize the linear classification problem to the nonlinear classification problem, and obtain the optimal solution of the original optimization problem by solving the dual problem equivalent to the optimization problem:

[0030]

[0031] st0≤α l ≤s l C

[0032]

[0033] Among them, the kernel function is selected as follows:

[0034]

[0035] Among them, α l is the Lagrangian parameter; δ is the support vector kernel function parameter; α is the input Lagrangian parameter; y is the label class of the input reference point RSSI value; X l is the RSSI value of the lth reference point; X is the input;

[0036] S213, determining the support vector kernel function parameter δ and the penalty factor C;

[0037] S214, determine K classification hyperplanes, and output the largest hyperplane as the region to which the reference point belongs:

[0038]

[0039] Furthermore, in step S3, the process of assigning a higher weight to the generated particles that belong to the same region includes the following sub-steps:

[0040] S301, based on the normal distribution of specific parameters, generate particle RSSI signal data according to the RSSI signal data of the historical positioning point of the terminal to be positioned, and tThe RSSI data received at the location and its membership degree for each area and the RSSI value of the generated particle and its membership degree for each area are input into K classification decision models respectively to judge whether the terminal to be located is in l t The RSSI data received at the location and the area category to which the RSSI signal data of the generated particle belongs;

[0041] S302: assign weights to the coordinates of the generated particles that are consistent with the category of the terminal to be located. M is the number of generated particles; weights are assigned to the coordinates of the generated particles that are inconsistent with the category of the terminal to be located.

[0042] S303, using the Pearson correlation coefficient between the RSSI value of the generated particle and the RSSI value of the reference point in the region to which it belongs, the coordinates of the generated particle are estimated. The coordinate formula is:

[0043]

[0044] in, is the nth RSSI characteristic value of the i-th generated particle, is the nth RSSI eigenvalue of the zth reference point, Z is the number of reference points in the region to which the particle belongs, and N is the number of RSSI eigenvalues of the reference points and the generated particles. is the standard deviation of the RSSI signal characteristics of the i-th generated particle, is the standard deviation of the RSSI signal characteristic of the zth generated particle, (x z ,y z ,z z ) is the coordinate of the reference point, (x i ,y i ,z i ) is the coordinate of the i-th generated particle.

[0045] Furthermore, in step S301, based on the normal distribution of specific parameters, the process of generating particle RSSI signal data according to the RSSI signal data of the historical positioning points of the terminal to be positioned includes the following sub-steps:

[0046] Based on the signal data received from the reference point by each base station, the corresponding population density in the area divided by each classification decision model is determined:

[0047]

[0048]

[0049] Where, is the fluctuation degree of the signal data feedback received by the v-th base station in the hospital from the reference points in a specific area, Z is the number of reference points in the specific area of the hospital that send signal data to the v-th base station, RSSI_v i is the signal characteristic value of the i-th reference point sending signal data to the v-th base station in a specific area, is the average value of the signal data received by the v-th base station in the specific area of the hospital, is the density of people in the area, and N is the number of base stations in the hospital;

[0050] According to the population density, a normal distribution with specific parameters is set:

[0051]

[0052] Among them, X is the RSSI signal data of the generated particle, RSSI t-1 The terminal to be located is t-1 RSSI signal data sent by the location to the base station, The average value of the signal data received by the base station in the area where the terminal to be located belongs;

[0053] Based on the normal distribution of specific parameters, the historical positioning points of the terminal to be located are t-1 The signal data is set as a random seed to generate random numbers for particle RSSI signal data.

[0054] Furthermore, in step S3, based on the movement trend changes described by the historical positioning points of the terminal to be located, based on the movement patterns of the human body, and based on the area to which the terminal to be located belongs, an optimal sampling area with a large posterior probability density is constructed, and a higher weight is assigned to the generated particles in the optimal sampling area. The process includes the following sub-steps:

[0055] S311, obtaining the historical positioning point location l of the terminal to be positioned t-3 (x t-3 ,y t-3 ,z t-3 ), l t-2 (x t-2 ,y t-2 ,z t-2 ) and l t-1 (x t-1 ,y t-1 ,z t-1 ), judge the vertical movement of the terminal to be positioned, and calculate the historical positioning point l of the terminal to be positioned. t-1 ,l t-2 ,l t-3 Construct the optimal sampling area:

[0056] (1) When z t-3 =zt-2 =z t-1 When the terminal to be positioned is on plane z=z t-1 Upward movement, no vertical displacement;

[0057] Use the quadratic function to fit the terminal to be located in the plane z=z t-1 The trajectory of action on a:

[0058] y a =a0+a1x+a2x 2 ;

[0059] Set the historical positioning point of the terminal to be positioned on the plane z=z t-1 Position on l t-3 ,l t-2 ,l t-1 Substitute into the equation to find the parameters a0, a1, a2;

[0060] Use linear function to solve line b:

[0061] y b =a′0+a′1x;

[0062] Set the historical positioning point of the terminal to be positioned on the plane z=z t-1 Position on l t-2 ,l t-1 Substitute into the equation to find the parameters a′0, a′1;

[0063] The location of the historical positioning point of the terminal to be positioned l t-1 As the center of the circle, with l t-1 With l t-2 distance As the radius, we get the sphere c:

[0064] (xx t-1 ) 2 +(yy t-1 ) 2 +(zz t-1 ) 2 =Δd 2 ;

[0065] Simultaneously solve the quadratic function a, the sphere c, and the plane z=z t-1 , determine the intersection point A2 where a and c intersect; combine the line b, the sphere c, and the plane z=z t-1 , determine the intersection point B2 where b and c intersect; move A2 and B2 to Plane and Plane projection, we get A′2 and B′2 and A″2 and B″2 respectively; where v max The fastest ascent speed of the human body, The historical positioning point l of the terminal to be positionedt-2 With l t-1 The time interval for successful positioning;

[0066] From point A'2, point B'2, point A"2, point B"2, point l' t-1 , point l″ t-1 Get plane A′2A″2l′ t-1 l″ t-1 , plane B′2B″2l′ t-1 l″ t-1 , plane A′2B′2l′ t-1 With plane A″2B″2l″ t-1 The plane equation of

[0067] From the plane A′2A″2l′ t-1 l″ t-1 , plane B′2B″2l′ t-1 l″ t-1 , plane A′2B′2l′ t-1 With plane A″2B″2l″ t-1 Construct the optimal sampling area A′1B′1l′ t-1 A″2B″2l″ t-1 ;

[0068] (2) When z t-1 ≠z t-2 ≠z t-2 When the terminal to be located is in the state of going upstairs or downstairs, there is displacement in the vertical direction;

[0069] Use a three-dimensional spiral to fit the trajectory of the terminal to be located:

[0070]

[0071] Among them, The historical positioning point l of the terminal to be positioned t-2 With l t-1 The time interval for successful positioning is determined by the historical position of the terminal to be positioned l t-1 (x t-1 ,y t-1 ,z t-1 ) obtain the parameters r, ω, b of the three-dimensional helix;

[0072] The historical location of the terminal to be located l t-3 ,l t-2 ,l t-1 Projection onto plane z=z t-1 , get the projection point position l′ t-3 (x t-3 ,yt-3,z t-1 ),l′ t-2 (xt-2 ,y t-2 ,z t-1 ),l′ t-1 (x t-1 ,yt-1,z t-1 );

[0073] Use a quadratic function to fit z = z t-1 Movement trajectory on the projection surface:

[0074] y a =a0+a1x+a2x 2 ;

[0075] The projection point is on the plane z=z t-1 The projection point position l′ on t-3 ,l′ t-2 ,l′ t-1 Substitute into the equation to find the parameters a0, a1, a2;

[0076] Solve z=z using linear function t-1 Line b on the projection plane:

[0077] y b =a′0+a′1x;

[0078] The projection point is on the plane z=z t-1 The projection point position l′ on t-2 ,l′ t-1 Substitute into the equation to find the parameters a′0, a′1;

[0079] The location of the historical positioning point of the terminal to be located l t-1 As the center of the circle, with l t-1 With l t-2 distance As the radius, we get the sphere c:

[0080] (xx t-1 ) 2 +(yy t-1 ) 2 +(zz t-1 ) 2 =Δd 2 ;

[0081] Combine sphere c and the three-dimensional spiral to determine the intersection point A2 between the three-dimensional spiral and circle c; combine sphere c and curve a to determine the intersection point A3 between curve a and circle c; combine sphere c and line b to determine the intersection point B2 between line b and circle c; project B2 onto the z-plane of A2's z-coordinate to obtain B′2;

[0082] From point A2, point B′2, point A3, point B2, point l t-1, find the plane A2A3l t-1 , plane B′2B2l t-1 , plane A2B′2l t-1 With plane A3B2l t-1 The plane equation of

[0083] By plane A2A3l t-1 , plane B′2B2l t-1 , plane A2B′2l t-1 With plane A3B2l t-1 Constitute the optimal sampling area A3B2A2B′2l t-1 ;

[0084] S312: Determine whether the generated particle is within the optimal sampling area and normalize the weights of samples inside and outside the area:

[0085]

[0086] Among them, M1 is the number of generated particles in the optimal sampling area, M2 is the number of generated particles in other areas, and the value of q is an integer greater than 1. represents the weight of the i-th particle at time t.

[0087] Furthermore, in step S3, based on the resampling idea, the particle positions are weighted according to the particle weights to obtain the positioning information of the terminal to be positioned, which includes the following sub-steps:

[0088] When the number of historical positioning points of the terminal to be located T satisfies 0≤T≤2, the optimal sampling area cannot be formed, and the weight of the particle in the area consistent with the terminal to be located is The weight of the particle that is not in the same area as the terminal to be located

[0089] When the number of historical positioning points of the terminal to be positioned is greater than 2, the optimal sampling area is constructed; based on the known particle positions, it is determined whether the generated particles are within the optimal sampling area; the weight of the particles in the optimal sampling area is Weights of particles outside the optimal sampling area

[0090] When the number of weighted particles is less than or equal to When , particles with weights heavier than the preset weight threshold are retained, and particles with weights lighter than the preset weight threshold are resampled;

[0091] The coordinates of the terminal to be located (x t ,y t ,z t ):

[0092]

[0093] Where, The terminal to be located is t Position The position coordinates of the i-th generated particle.

[0094] Furthermore, in step S312, the process of determining whether the generated particles are within the optimal sampling area includes the following sub-steps:

[0095] Determine the target location l t Generated Whether the particle is located in the hemisphere on the A2 side of the sphere c:

[0096]

[0097]

[0098]

[0099] Where i = 1, 2, ..., M;

[0100] If the normal vectors of two planes are known and Then the judgment vector The formula for whether it is between two planes:

[0101]

[0102]

[0103] Assuming the coordinates of the three points are known, construct the normal vector plane equation

[0104] In the optimal sampling area A′1B′1l′ t-1 A″2B″2l″ t-1 In the figure, let plane A′2A″2l′ t-1 l″ t-1 The normal vector is Plane B′2B″2l′ t-1 l″ t-1 The normal vector is Plane A′2B′2l′ t-1 The normal vector is and plane A″2B″2l″ t-1 The normal vector is

[0105] In the optimal sampling area A3B2A2B′2l t-1 In the figure, let plane A2A3l t-1 The normal vector is Plane B′2B2lt-1 The normal vector is Plane A2B′2l t-1 The normal vector is and plane A3B2l t-1 The normal vector is

[0106] Determine the target location l t Generated Whether the particle is located in the optimal sampling area:

[0107]

[0108]

[0109]

[0110] When z t-3 =z t-2 =z t-1 When p=1,2;

[0111] When z t-1 ≠z t-2 ≠z t-2 When p=3,4.

[0112] Beneficial effects:

[0113] First, the multi-scenario adaptive B5G three-dimensional positioning method for smart medical care of the present invention obtains the signal data of the positioning point currently collected by the terminal to be positioned in the hospital; based on the signal data and category labels of the reference points in the hospital during the offline phase, the LBP-SVM algorithm is used to train a classification decision model; based on the classification decision model, an improved particle filter algorithm is used to perform position calculation to obtain the positioning information of the terminal to be positioned; and the positioning information is fed back to the terminal to be positioned. Based on the zoning of the hospital, the terminals to be positioned are classified, and the particle filter is used to convert the moving state into the stationary state, thereby improving the positioning accuracy and, to a certain extent, being more adaptable to complex environments.

[0114] Second, the multi-scenario adaptive B5G three-dimensional positioning method for smart medical care of the present invention, compared with the traditional positioning method, is a multi-scenario adaptive B5G three-dimensional positioning algorithm for smart medical care that evenly deploys 5G base stations in the hospital, improves the effective utilization rate of site resources, reduces construction and maintenance costs, and facilitates the implementation of base station construction projects. The signal characteristic values are extracted to build a high-quality fingerprint library, and the walking trajectories of mobile personnel are used to perform physical modeling according to the physical environment in the hospital to plan the optimal sampling area. It only needs to match and weight the signal data in the optimal sampling area to perform high-precision three-dimensional positioning.

[0115] Third, the multi-scenario adaptive B5G three-dimensional positioning method for smart medical care of the present invention is based on the fingerprint database constructed by signal position-features, and based on the obtained information data of the terminal to be located, generates particles that conform to the normal distribution and weights them, further significantly improving the accuracy of indoor positioning position solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0116] Figure 1 This is a flow chart of the multi-scenario adapted B5G three-dimensional positioning method for smart medical care according to an embodiment of the present invention.

[0117] Figure 2 4 is a flow chart of position calculation in the online stage of an embodiment of the present invention.

[0118] Figure 3 This is a flow chart of the optimal sampling area division algorithm according to an embodiment of the present invention.

[0119] Figure 4 Schematic diagram of the optimal sampling area in the online stage of an embodiment of the present invention. DETAILED DESCRIPTION

[0120] The following examples may enable those skilled in the art to more fully understand the present invention, but are not intended to limit the present invention in any way.

[0121] In one embodiment, Figure 1 As shown in the figure, a multi-scenario adaptive B5G three-dimensional positioning algorithm for smart medical care is provided, which is applied to the base station currently connected to the terminal to be positioned in the hospital as the execution subject, including the following steps:

[0122] Step 1: Obtain signal data of the positioning point currently collected by the terminal to be positioned in the hospital.

[0123] In step 2, based on the signal data and category labels of the reference points in the hospital during the offline phase, the LBP-SVM algorithm is used to train a classification decision model.

[0124] Specifically, based on the signal data and category labels of reference points within the hospital during the offline phase, the LBP-SVM algorithm is used to train a classification decision model. The specific steps include: obtaining reference point signal data and reference point locations, and labeling the reference points according to regions; performing LBP feature extraction on the reference point signal data within the region to obtain the LBP data features of the reference point in the region to which it belongs, and calculating the reference point's membership with respect to each region; and inputting the RSSI signal data of the reference points within the region and the reference point's membership with respect to each region into the SVM to obtain the classification decision model. The LBP algorithm is used for RSSI feature extraction. Due to the complex and dynamic environment of large shopping malls and the high flow of people, 5G signals are susceptible to interference during propagation. Furthermore, when using the terminal to be located for signal acquisition, it is not possible to always face the corresponding access point base station during signal acquisition. When acquiring signals with the back facing the access point base station, the received signal will pass through the human body, causing varying degrees of attenuation, resulting in significant instability in the acquired signal strength. Furthermore, various metal objects can also interfere with the signal value. Therefore, whether in the offline stage or the online stage, the selection of signal feature values is very important. Therefore, the LBP algorithm is used to extract RSSI features.

[0125] Exemplarily, a method for performing LBP feature extraction on RSSI signal data of a reference point in a region to obtain LBP data features of the reference point in the region to which it belongs, and calculating the membership of the reference point with respect to each region includes:

[0126] Obtain reference point signal data and reference point positions, and add labels to reference points according to regions.

[0127] To satisfy rotation invariance, several adjacent RSSI values centered on the RSSI value are obtained based on a circular LBP operator, and the several RSSI values are sequentially compared with the central RSSI value.

[0128] The adjacent RSSI values whose RSSI values are greater than the central RSSI value are marked as 1, and the adjacent RSSI values whose RSSI values are less than the central RSSI value are marked as 0; the marks of the adjacent RSSI values are extracted in sequence according to a preset order to form a binary digital string, and the binary digital string is used as the LBP feature of the RSSI value.

[0129] Each RSSI value in the area is traversed according to the method of obtaining the LBP feature of a certain RSSI value to obtain the LBP feature of the RSSI value of each reference point in the area.

[0130] Use the log-likelihood ratio similarity to calculate the membership of each reference point in the region:

[0131]

[0132]

[0133]

[0134] s l =2*(matrixEntropy-rowEntropy-columEntropy)

[0135] in, is the nth RSSI eigenvalue of the zth reference point in the area, Z is the number of reference points in the area, and N is the number of RSSI eigenvalues of the reference points.

[0136] Exemplarily, the RSSI signal data of the reference points in the region and the membership of the reference points with respect to each region are input into the SVM to obtain the classification decision model, which includes the following steps:

[0137] The reference point RSSI signal data is used as the training set of the support vector machine to obtain the regional classification decision model.

[0138] The input is the reference point RSSI and the reference point membership s for each area l Assume that one is a positive sample, that is, the reference point belonging to the region, and the rest are negative samples. To maximize the point-to-plane distance of any sample, the support vector-to-plane distance formula can be optimized as follows:

[0139]

[0140]

[0141] Among them, the parameter ω k , b k are the normal vector and intercept of the hyperplane of the kth class respectively; Indicates mapping the input of the linearly inseparable low-dimensional space to a linearly separable high-dimensional space, y l The label class of the RSSI value of the lth reference point, RSSI l is the RSSI value of the lth reference point, θ l is the relaxation factor, and C is the penalty factor.

[0142] Since the dual problem is easier to solve, the dual algorithm of support vector machine under linear separable conditions is used to transform the optimization problem into the dual variable through Lagrange duality, that is, the optimal solution of the original problem is obtained by solving the dual problem equivalent to the original problem.

[0143] Introducing the kernel function K(X l,X), generalize the linear classification problem to the nonlinear classification problem, and obtain the optimal solution of the original problem by solving the dual problem:

[0144]

[0145] st0≤α l ≤s l C,l=1,2,…,L

[0146]

[0147] Among them, the kernel function is selected as follows:

[0148]

[0149] Among them, α l is the Lagrangian parameter.

[0150] Determine the support vector kernel function parameter δ and the penalty factor C in order to control the generalization ability of the support vector machine.

[0151] Determine K classification hyperplanes, and the hyperplane with the largest output is the region to which the reference point belongs.

[0152]

[0153] Step 3: Based on the classification decision model, an improved particle filter algorithm is used to perform position calculation to obtain the positioning information of the terminal to be positioned.

[0154] Step 4: Feedback the positioning information to the terminal to be positioned.

[0155] Specifically, if Figure 2 As shown, based on a normal distribution with specific parameters, particle RSSI signal data is generated based on the RSSI signal data of the historical positioning points of the terminal to be located. Based on the classification decision model, it is determined whether the category of the region to which the terminal to be located belongs is consistent with the category of the region to which the generated particles belong. A higher weight is assigned to the generated particles whose region is consistent. Based on the changes in the movement trend described by the historical positioning points of the terminal to be located, according to the laws of human movement, and based on the region to which the terminal to be located belongs, an optimal sampling region with a high posterior probability density is constructed, and a higher weight is assigned to the generated particles within the optimal sampling region. Based on the resampling concept, the particle positions are weighted according to the particle weights to obtain the positioning information of the terminal to be located.

[0156] Exemplarily, the process of generating particle RSSI signal data based on the normal distribution of specific parameters and the RSSI signal data of the historical positioning points of the terminal to be located, determining whether the category of the area to which the terminal to be located belongs is consistent with the category of the area to which the generated particles belong based on the classification decision model, and assigning a higher weight to the generated particles that belong to the same area includes:

[0157] Based on the normal distribution of specific parameters, the particle RSSI signal data is generated according to the RSSI signal data of the historical positioning point of the terminal to be positioned, and the terminal to be positioned is t The RSSI data received at the location and its membership degree for each area and the RSSI value of the generated particle and its membership degree for each area are input into K classification decision models respectively to judge whether the terminal to be located is in l t The RSSI data received at the location and the RSSI signal data of the generated particle belong to the area category.

[0158] For the generated particles that are consistent with the category of the terminal to be located, the coordinates of the generated particles are given weights. For the generated particles that are inconsistent with the category of the terminal to be located, the coordinates of the generated particles are given weights.

[0159] The coordinates of the generated particles are estimated using the Pearson correlation coefficient between the RSSI value of the generated particles and the RSSI value of the reference point in the area to which they belong. The formula is:

[0160]

[0161] in, is the nth RSSI characteristic value of the i-th generated particle, is the nth RSSI eigenvalue of the zth reference point, Z is the number of reference points in the region to which the particle belongs, and N is the number of RSSI eigenvalues of the reference points and the generated particles. is the standard deviation of the RSSI signal characteristics of the i-th generated particle, is the standard deviation of the RSSI signal characteristic of the zth generated particle, (x z ,y z ,z z ) is the coordinate of the reference point, (x i ,y i ,z i ) is the generated particle coordinates.

[0162] Exemplarily, based on a normal distribution of specific parameters, the process of generating particle RSSI signal data according to RSSI signal data of historical positioning points of the terminal to be positioned includes:

[0163] Analyze the signal data fed back by the reference point received by each base station to determine the population density corresponding to the area divided by each classification decision model:

[0164]

[0165]

[0166] Where, is the fluctuation degree of the signal data feedback received by the v-th base station in the hospital from the reference points in a specific area, Z is the number of reference points in the specific area of the hospital that send signal data to the v-th base station, RSSI_v i is the signal characteristic value of the i-th reference point sending signal data to the v-th base station in a specific area, is the average value of the signal data received by the v-th base station in the specific area of the hospital, is the density of people in the area, and N is the number of base stations in the hospital;

[0167] Since the mobile state of the terminal to be located will be affected by the surrounding environment, a normal distribution of specific parameters is set according to the density of people:

[0168]

[0169] Among them, X is the RSSI signal data of the generated particle, RSSI t-1 The terminal to be located is t-1 RSSI signal data sent by the location to the base station, It is the average value of the signal data received by the base station in the area where the terminal to be located belongs.

[0170] Based on the normal distribution of specific parameters, the historical positioning points of the terminal to be located are t-1 The signal data is set as a random seed to generate random numbers for particle RSSI signal data.

[0171] For example, Figure 3 As shown, according to the movement trend change described by the historical positioning points of the terminal to be located, according to the movement law of the human body, and according to the area to which the terminal to be located belongs, the optimal sampling area with large posterior probability density is constructed, as shown in FIG. Figure 4 As shown in Figure 2, the process of giving higher weights to generated particles within the optimal sampling area includes the following steps:

[0172] According to the movement trend changes described by the historical positioning points of the terminal to be located and the movement rules of the human body, the corresponding optimal sampling area is constructed.

[0173] Since the historical positioning point position of the terminal to be positioned is l t-3 (x t-3 ,yt-3 ,z t-3 ), l t-2 (x t-2 ,y t-2 ,z t-2 ), l t-1 (x t-1 ,y t-1 ,z t-1 ) has been learned, and the vertical movement of the terminal to be positioned is judged, and the historical positioning point l of the terminal to be positioned is used. t-1 ,l t-2 ,l t-3 Construct the optimal sampling area.

[0174] (1) When z t-3 =z t-2 =z t-1 When the terminal to be positioned is on plane z=z t-1 There is no vertical movement.

[0175] Use the quadratic function to fit the terminal to be located in the plane z=z t-1 The trajectory of action on a:

[0176] y a =a0+a1x+a2x 2 .

[0177] Set the historical positioning point of the terminal to be positioned on the plane z=z t-1 Position on l t-3 ,l t-2 ,l t-1 Substitute into the equation to find the parameters a0, a1, and a2.

[0178] Use linear function to solve line b:

[0179] y b =a′0+a′1x.

[0180] Set the historical positioning point of the terminal to be positioned on the plane z=z t-1 Position on l t-2 ,l t-1 Substitute into the equation and find the parameters a′0, a′1.

[0181] The location of the historical positioning point of the terminal to be positioned l t-1 As the center of the circle, with l t-1 With l t-2 distance As the radius, we get the sphere c:

[0182] (xx t-1 ) 2 +(yy t-1 ) 2+(zz t-1 ) 2 =Δd 2 .

[0183] Simultaneously solve the quadratic function a, the sphere c, and the plane z=z t-1 , it is easy to determine the intersection point A2 where a and c intersect. Simultaneously, the line b, the sphere c, and the plane z=z t-1 , it is easy to determine the intersection point B2 where b and c intersect. Plane and Plane projection, we get A′2 and B′2 and A″2 and B″2 respectively. Among them, v max The fastest ascent speed of the human body, The historical positioning point l of the terminal to be positioned t-2 With l t-1 The interval between successful positioning.

[0184] From point A'2, point B'2, point A"2, point B"2, point l' t-1 , point l″ t-1 We can get plane A′2A″2l′ t-1 l″ t-1 , plane B′2B″2l′ t-1 l″ t-1 , plane A′2B′2l′ t-1 With plane A″2B″2l″ t-1 The plane equation of . By plane A′2A″2l′ t-1 l″ t-1 , plane B′2B″2l′ t-1 l″ t-1 , plane A′2B′2l′ t-1 With plane A″2B″2l″ t-1 Construct the optimal sampling area A′1B′1l′ t-1 A″2B″2l″ t-1 .

[0185] (2) When z t-1 ≠z t-2 ≠z t-2 When the terminal to be positioned is in the state of going upstairs or downstairs, there is displacement in the vertical direction.

[0186] Use a three-dimensional spiral to fit the trajectory of the terminal to be located:

[0187]

[0188] Among them, The historical positioning point l of the terminal to be positioned t-2 With l t-1 The time interval for successful positioning can be determined by the historical position of the terminal to be positioned lt-1 (x t-1 ,y t-1 ,z t-1 ) to obtain the parameters r, ω, and b of the three-dimensional helix.

[0189] The historical location of the terminal to be located l t-3 ,l t-2 ,l t-1 Projection onto plane z=z t-1 , that is, obtain the projection point position l′ t-3 (x t-3 ,y t-3 ,z t-1 ),l′ t-2 (x t-2 ,y t-2 ,z t-1 ),l′ t-1 (x t-1 ,y t-1 ,z t-1 ).

[0190] Use a quadratic function to fit z = z t-1 Movement trajectory on the projection surface:

[0191] y a =a0+a1x+a2x 2 ;

[0192] The projection point is on the plane z=z t-1 The projection point position l′ on t-3 ,l′ t-2 ,l′ t-1 Substitute into the equation to find the parameters a0, a1, a2;

[0193] Solve z=z using linear function t-1 Line b on the projection plane:

[0194] y b =a′0+a′1x;

[0195] The projection point is on the plane z=z t-1 The projection point position l′ on t-2 ,l′ t-1 Substitute into the equation to find the parameters a′0, a′1;

[0196] The location of the historical positioning point of the terminal to be located l t-1 As the center of the circle, with l t-1 With l t-2 distance As the radius, we get the sphere c:

[0197] (xx t-1 ) 2+(yy t-1 ) 2 +(zz t-1 ) 2 =Δd 2 .

[0198] By combining the sphere c and the three-dimensional spiral, it is easy to determine the intersection A2 of the three-dimensional spiral and the circle c; by combining the sphere c and the curve a, it is easy to determine the intersection A3 of the curve a and the circle c; by combining the sphere c and the straight line b, it is easy to determine the intersection B2 of the straight line b and the circle c; project B2 onto the z plane where the z coordinate of A2 is located, and get B′2.

[0199] From point A2, point B′2, point A3, point B2, point l t-1 , we can get plane A2A3l t-1 , plane B′2B2l t-1 , plane A2B′2l t-1 With plane A3B2l t-1 The plane equation. By plane A2A3l t-1 , plane B′2B2l t-1 , plane A2B′2l t-1 With plane A3B2l t-1 Constitute the optimal sampling area A3B2A2B′2l t-1 .

[0200] According to the continuity of the curve motion, the unknown target has a higher probability of being located in the optimal sampling area in time, and the posterior probability density of the optimal sampling area is larger. On this basis, a higher weight is given to the particle position in the optimal sampling area.

[0201] Normalize the weights of samples inside and outside the region, indicating that the greater the posterior probability of the particle in the optimal sampling region, the greater the weight. The formulas are expressed as follows:

[0202]

[0203]

[0204] Among them, M1 is the number of generated particles in the optimal sampling area, M2 is the number of generated particles in other areas, and the value of q is an integer greater than 1. represents the weight of the i-th particle at time t.

[0205] Exemplarily, based on the resampling concept, the method of weighting the particle positions according to the particle weights to obtain the positioning information of the terminal to be positioned includes:

[0206] When the number of historical positioning points of the terminal to be positioned is 0≤T≤2, the optimal sampling area cannot be formed.

[0207] At this time, the particle in the same area as the terminal to be located has a weight Inconsistent particle,weights

[0208] When the number of historical positioning points of the terminal to be positioned T>2, the optimal sampling area is constructed.

[0209] Given the particle position, determine whether the generated particle is within the optimal sampling area.

[0210] At this time, the particles in the optimal sampling area have weights Particles outside the optimal sampling area, weight

[0211] When the number of particles with heavy weights ≤ When , the particles with heavy weights are retained and the particles with light weights are resampled.

[0212] The coordinates of the terminal to be located are estimated by generating particles weightedly, namely:

[0213]

[0214] Exemplarily, the process of determining whether the generated particles are within the optimal sampling area includes:

[0215] Determine the target location l t Generated i=1,2,…,M particles are located in the hemisphere on the A2 side of sphere c:

[0216]

[0217]

[0218]

[0219] If the normal vectors of two planes are known and Then the judgment vector The formula for whether it is between two planes:

[0220]

[0221]

[0222] Assuming the coordinates of the three points are known, the normal vector plane equation can be constructed

[0223] In the optimal sampling area A′1B′1l′ t-1 A″2B″2l″ t-1 In the figure, let plane A′2A″2l′ t-1 l″t-1 The normal vector is Plane B′2B″2l′ t-1 l″ t-1 The normal vector is Plane A′2B′2l′ t-1 The normal vector is and plane A″2B″2l″ t-1 The normal vector is

[0224] In the optimal sampling area A3B2A2B′2l t-1 In the figure, let plane A2A3l t-1 The normal vector is Plane B′2B2l t-1 The normal vector is Plane A2B′2l t-1 The normal vector is and plane A3B2l t-1 The normal vector is

[0225] Determine the target location l t Generated Whether the particle i=1,2,…,M is located in the optimal sampling area:

[0226]

[0227]

[0228]

[0229] When z t-3 =z t-2 =z t-1 When k=1,2;

[0230] When z t-1 ≠z t-2 ≠z t-2 When k=3,4.

[0231] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0232] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0233] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A multi-scenario adaptive B5G three-dimensional positioning method for smart medical care, characterized by: The positioning method comprises the following steps: S1, obtaining the signal data of the positioning point currently collected by the terminal to be positioned in the hospital; S2: Obtain the signal data and positions of reference points in the hospital during the offline phase, and label the reference points according to regions. Perform LBP feature extraction on the signal data of the reference points in the region to obtain the LBP data features of the reference points in the region to which they belong, and calculate the membership of the reference points with respect to each region. Input the RSSI signal data of the reference points in the region and the membership of the reference points with respect to each region into the SVM, and train K classification decision models, where K is the number of regions in the hospital. S3: Based on the normal distribution of specific parameters, particle RSSI signal data is generated based on the RSSI signal data of the historical positioning points of the terminal to be located. Based on the classification decision model, it is determined whether the category of the area to which the terminal to be located belongs is consistent with the category of the area to which the generated particles belong, and a higher weight is assigned to the generated particles that belong to the same area. Based on the movement trend changes described by the historical positioning points of the terminal to be located, according to the movement patterns of the human body, and according to the area to which the terminal to be located belongs, an optimal sampling area with a large posterior probability density is constructed, and a higher weight is assigned to the generated particles within the optimal sampling area. Based on the resampling concept, the particle positions are weighted according to the particle weight to obtain the positioning information of the terminal to be located. S4, feeding back the obtained positioning information to the terminal to be positioned; In step S2, the process of obtaining the LBP data features of the reference point in the region to which it belongs and calculating the membership of the reference point with respect to each region includes the following sub-steps: S201, select one of the reference point RSSI values and use it as the center RSSI value; S202, obtaining several adjacent RSSI values centered on the selected RSSI value based on a circular LBP operator, and sequentially comparing the obtained several adjacent RSSI values with the central RSSI value; S203, marking adjacent RSSI values whose RSSI values are greater than the central RSSI value as 1, and marking adjacent RSSI values whose RSSI values are less than the central RSSI value as 0; extracting the marks of the adjacent RSSI values in a predetermined order to form a binary digital string, and using the binary digital string as the LBP feature of the central RSSI value; S204, repeating steps S201 to S203, traversing each RSSI value in the area, and obtaining the LBP feature of the RSSI value of each reference point in the area; S205, using the log-likelihood ratio similarity to calculate the membership degree of each reference point in the region according to the following formula: s l =2*(matrixEntropy-rowEntropy-columEntropy) in, is the nth RSSI eigenvalue of the zth reference point in the region, Z is the number of reference points in the region, N is the number of RSSI eigenvalues of the reference points; rowEntropy is the row entropy; columEntropy is the column entropy; matrixEntropy is the matrix entropy; s l is the log-likelihood ratio similarity.

2. The multi-scenario adaptive B5G three-dimensional positioning method for smart medical care according to claim 1 is characterized in that: In step S2, the RSSI signal data of the reference points in the region and the membership of the reference points to each region are input into the SVM. The process of training and obtaining K classification decision models includes the following sub-steps: S211, using the reference point RSSI signal data as a training set for a support vector machine to obtain a regional classification decision model; the input of the classification decision model is the reference point RSSI and the membership degree s of the reference point to each region. l ; Assume that the reference points belonging to this area are positive samples, and the rest are negative samples; the optimization problem of the support vector to plane distance formula is as follows: Among them, the parameter ω k , b k are the normal vector and intercept of the hyperplane of the kth class respectively; Indicates mapping the input of the linearly inseparable low-dimensional space to a linearly separable high-dimensional space, y l The label class of the RSSI value of the lth reference point, RSSI l is the RSSI value of the lth reference point, θ l is the relaxation factor, C is the penalty factor; l=1,2,…,L, L is the total number of reference points; the superscript T indicates transposition; S212, introduce the kernel function K(X l ,X), generalize the linear classification problem to the nonlinear classification problem, and obtain the optimal solution of the original optimization problem by solving the dual problem equivalent to the optimization problem: s.t.0≤α l ≤s l C Among them, the kernel function is selected as follows: Among them, α l is the Lagrangian parameter; δ is the support vector kernel function parameter; α is the input Lagrangian parameter; y is the label class of the input reference point RSSI value; X l is the RSSI value of the lth reference point; X is the input; S213, determining the support vector kernel function parameter δ and the penalty factor C; S214, determine K classification hyperplanes, and output the largest hyperplane as the region to which the reference point belongs:

3. The multi-scenario adaptive B5G three-dimensional positioning method for smart medical care according to claim 1 is characterized in that: In step S3, the process of assigning a higher weight to particles generated in the same region includes the following sub-steps: S301, based on the normal distribution of specific parameters, generate particle RSSI signal data according to the RSSI signal data of the historical positioning point of the terminal to be positioned, and t The RSSI data received at the location and its membership degree for each area and the RSSI value of the generated particle and its membership degree for each area are input into K classification decision models respectively to judge whether the terminal to be located is in l t The RSSI data received at the location and the area category to which the RSSI signal data of the generated particle belongs; S302: assign weights to the coordinates of the generated particles that are consistent with the category of the terminal to be located. M is the number of generated particles; Assign weights to the coordinates of the generated particles that are inconsistent with the category of the terminal to be located S303, using the Pearson correlation coefficient between the RSSI value of the generated particle and the RSSI value of the reference point in the region to which it belongs, the coordinates of the generated particle are estimated. The coordinate formula is: in, is the nth RSSI characteristic value of the i-th generated particle, is the nth RSSI eigenvalue of the zth reference point, Z is the number of reference points in the region to which the particle belongs, and N is the number of RSSI eigenvalues of the reference points and the generated particles. is the standard deviation of the RSSI signal characteristics of the i-th generated particle, is the standard deviation of the RSSI signal characteristic of the zth generated particle, (x z ,y z ,z z ) is the coordinate of the reference point, (x i ,y i ,z i ) is the coordinate of the i-th generated particle.

4. The multi-scenario adaptive B5G three-dimensional positioning method for smart medical care according to claim 3 is characterized in that: In step S301, based on the normal distribution of specific parameters, the process of generating particle RSSI signal data according to the RSSI signal data of the historical positioning points of the terminal to be positioned includes the following sub-steps: Based on the signal data received from the reference point by each base station, the corresponding population density in the area divided by each classification decision model is determined: Where, is the fluctuation degree of the signal data feedback received by the v-th base station in the hospital from the reference points in a specific area, Z is the number of reference points in the specific area of the hospital that send signal data to the v-th base station, RSSI_v i is the signal characteristic value of the i-th reference point sending signal data to the v-th base station in a specific area, is the average value of the signal data received by the v-th base station in the specific area of the hospital, is the density of people in the area, and H is the number of base stations in the hospital; According to the population density, a normal distribution with specific parameters is set: Among them, X is the RSSI signal data of the generated particle, RSSI t-1 The terminal to be located is t-1 RSSI signal data sent by the location to the base station, The average value of the signal data received by the base station in the area where the terminal to be located belongs; Based on the normal distribution of specific parameters, the historical positioning points of the terminal to be located are t-1 The signal data is set as a random seed to generate random numbers for particle RSSI signal data.

5. The multi-scenario adaptive B5G three-dimensional positioning method for smart medical care according to claim 1 is characterized in that: In step S3, based on the movement trend changes described by the historical positioning points of the terminal to be located, the movement patterns of the human body, and the area to which the terminal to be located belongs, an optimal sampling area with a large posterior probability density is constructed. The process of assigning a higher weight to the generated particles in the optimal sampling area includes the following sub-steps: S311, obtaining the historical positioning point location l of the terminal to be positioned t-3 (x t-3 ,y t-3 ,z t-3 ), l t-3 (x t-3 ,y t-2 ,z t-2 ) and l t-1 (x t-1 ,y t-1 ,z t-1 ), judge the vertical movement of the terminal to be positioned, and calculate the historical positioning point l of the terminal to be positioned. t-1 ,l t-2 ,l t-3 Construct the optimal sampling area: (1) When z t-3 =z t-2 =z t-1 When the terminal to be positioned is on plane z=z t-1 Upward movement, no vertical displacement; Use the quadratic function to fit the terminal to be located in the plane z=z t-1 The trajectory of action on a: y a =a0+a1x+a2x 2 ; Set the historical positioning point of the terminal to be positioned on the plane z=z t-1 Position on l t-3 ,l t-2 ,l t-1 Substitute into the equation to find the parameters a0, a1, a2; Use linear function to solve line b: the b =a′0+a′1x: Set the historical positioning point of the terminal to be positioned on the plane z=z t-1 Position on l t-2 ,l t-1 Substitute into the equation to find the parameters a′0, a′1; The location of the historical positioning point of the terminal to be positioned l t-1 As the center of the circle, with l t-1 With l t-2 distance As the radius, we get the sphere c: (x-x t-1 ) 2 +(y-y t-1 ) 2 +(z-z t-1 ) 2 =Δd 2 ; Simultaneously solve the quadratic function a, the sphere c, and the plane z=z t-1 , determine the intersection point A2 where a and c intersect; combine the line b, the sphere c, and the plane z=z t-1 , determine the intersection point B2 where b and c intersect; move A2 and B2 to Plane and Projecting on the plane, we get A′2 and B′2 as well as A″2 and B″2 respectively; Among them, v max The fastest ascent speed of the human body, The historical positioning point l of the terminal to be positioned t-2 With l t-1 The time interval for successful positioning; From point A'2, point B'2, point A"2, point B"2, point l' t-1 , point l″ t-1 Get plane A′2A″2l′ t-1 l″ t-1 , plane B′2B″2l′ t- 1l″ t-1 , plane A′2B′2l′ t-1 With plane A″2B″2l″ t-1 The plane equation of From the plane A′2A″2l′ t-1 l″ t-1 , plane B′2B"2l′ t-1 l″ t-1 , plane A′2B′2l′ t-1 With plane A″2B″2l″ t-1 Construct the optimal sampling area A′1B′1l′ t-1 A″2B″2l″ t-1 ; (2) When z t-1 ≠z t-2 ≠z t-2 When the terminal to be located is in the state of going upstairs or downstairs, there is displacement in the vertical direction; Use a three-dimensional spiral to fit the trajectory of the terminal to be located: Among them, The historical positioning point l of the terminal to be positioned t-2 With l t-1 The time interval for successful positioning is determined by the historical position of the terminal to be positioned l t-1 (x t-1 ,y t-1 ,z t-1 ) get the parameters r, ω, b of the three-dimensional helix; The historical location of the terminal to be located l t-3 ,l t-2 ,l t-1 Projection onto plane z=z t-1 , get the projection point position l′ t-3 (x t-3 ,y t-3 ,z t-1 ),l′ t-2 (x t-2 ,y t-2 ,z t-1 ),l′ t-1 (x t-1 ,y t-1 ,z t-1 ); Use a quadratic function to fit z = z t-1 Movement trajectory on the projection surface: y a =a0+a1x+a2x 2 ; The projection point is on the plane z=z t-1 The projection point position l′ on t-3 ,l′ t-2 ,l′ t-1 Substitute into the equation to find the parameters a0, a1, a2; Solve z=z using linear function t-1 Line b on the projection plane: the b =a′0+a′1x: The projection point is on the plane z=z t-1 The projection point position l′ on t-2 ,l′ t-1 Substitute into the equation to find the parameters a′0, a′1; The location of the historical positioning point of the terminal to be located l t-1 As the center of the circle, with l t-1 With l t-2 distance As the radius, we get the sphere c: (x-x t-1 ) 2 +(y-y t-1 ) 2 +(z-z t-1 ) 2 =Δd 2 ; Combine sphere c and the three-dimensional spiral to determine the intersection point A2 between the three-dimensional spiral and circle c; combine sphere c and curve a to determine the intersection point A3 between curve a and circle c; combine sphere c and line b to determine the intersection point B2 between line b and circle c; project B2 onto the z-plane of A2's z-coordinate to obtain B′2; From point A2, point B′2, point A3, point B2, point l t-1 , find the plane A2A3l t-1 , plane B′2B2l t-1 , plane A2B′2l t-1 With plane A3B2l t-1 The plane equation of By plane A2A3l t-1 , plane B′2B2l t-1 , plane A2B′2l t-1 With plane A3B2l t-1 Constitute the optimal sampling area A3B2A2B′2l t-1 ; S312: Determine whether the generated particle is within the optimal sampling area and normalize the weights of samples inside and outside the area: Among them, M1 is the number of generated particles in the optimal sampling area, M2 is the number of generated particles in other areas, and the value of q is an integer greater than 1. represents the weight of the i-th particle at time t.

6. The multi-scenario adaptive B5G three-dimensional positioning method for smart medical care according to claim 1 is characterized in that: In step S3, based on the resampling idea, the particle positions are weighted according to the particle weights to obtain the positioning information of the terminal to be located, which includes the following sub-steps: When the number of historical positioning points of the terminal to be located T satisfies 0≤T≤2, the optimal sampling area cannot be formed, and the weight of the particle in the area consistent with the terminal to be located is The weight of the particle that is not in the same area as the terminal to be located When the number of historical positioning points of the terminal to be positioned is greater than 2, the optimal sampling area is constructed; based on the known particle positions, it is determined whether the generated particles are within the optimal sampling area; the weight of the particles in the optimal sampling area is Weights of particles outside the optimal sampling area When the number of particles with heavy weights is less than or equal to When , particles with weights heavier than the preset weight threshold are retained, and particles with weights lighter than the preset weight threshold are resampled; The coordinates of the terminal to be located (x t ,y t ,z t ): Where, The terminal to be located is t Position The position coordinates of the i-th generated particle.

7. The multi-scenario adaptive B5G three-dimensional positioning method for smart medical care according to claim 5 is characterized in that: In step S312, the process of determining whether the generated particles are within the optimal sampling area includes the following sub-steps: Determine the target location l t Generated Whether the particle is located in the hemisphere on the A2 side of the sphere c: Where i = 1, 2, ..., M; If the normal vectors of two planes are known and Then the judgment vector The formula for whether it is between two planes: Assuming the coordinates of the three points are known, construct the normal vector plane equation In the optimal sampling area A′1B′1l" t-1 A"2B"2l" t-1 In the figure, let plane A′2A″2l′ t-1 l″ t-1 The normal vector is Plane B′2B″2l′ t-1 l" t-1 The normal vector is Plane A′2B′2l′ t-1 The normal vector is and plane A"2B"2l" t-1 The normal vector is In the optimal sampling area A3B2A2B′2l t-1 In the figure, let plane A2A3l t-1 The normal vector is Plane B′2B2l t-1 The normal vector is Plane A2B′2l t-1 The normal vector is and plane A3B2l t-1 The normal vector is Determine the target location l t Generated Whether the particle is located in the optimal sampling area: When z t-3 =z t-2 =z t-1 When p=1,2; When z t-1 ≠z t-2 ≠z t-2 When p=3,4.

Images