Passive personnel positioning method and device based on multi-region Gaussian probability fusion model

Abstract

The application discloses a passive personnel positioning method and equipment of a multi-region Gaussian probability fusion model. The method sets a passive positioning system with a target true and false identification model. The positioning system comprises M signal transmitters or signal reflectors and N signal receivers installed on the edges of a positioning space. Any signal transmitter or signal reflector and any signal receiver form a line-of-sight link, and a total of line-of-sight links are formed. The position of a person is determined according to the relationship between the signal strength received by the signal receiver and each line-of-sight link. The target true and false identification model of the positioning system is trained, and then the trained positioning system is used to actually position a target. The feature distribution of the true and false targets in a sub-region is calculated according to the feature data corresponding to the true and false targets in the sub-region. The application can accurately identify the false targets in the positioning result, and can reduce the artificial cost and time cost for collecting training data.

Description

Technical Field

[0001] This invention relates to the field of wireless positioning, and in particular to a passive personnel positioning method and device based on a multi-region Gaussian probability fusion model. Background Technology

[0002] Currently, with the emergence of concepts such as smart buildings and the Industrial Internet, indoor positioning technology has developed rapidly. Passive positioning technology is a relatively new technology that has gained attention in recent years. This technology eliminates the need for personnel to carry positioning devices, making it highly attractive to manufacturers in the smart IoT field. However, in multi-person positioning scenarios, this technology is prone to false target localization. To address this issue, supervised learning algorithms are often used as classifiers to distinguish between real and false targets. However, multipath interference is severe in indoor environments, and the shadow link features of false targets in different areas are not entirely consistent. While high-density fingerprint point collection can improve this problem, it significantly increases the manual cost of constructing the training database. Therefore, solving this problem has significant economic value for the practical application of passive positioning technology in smart building systems. Summary of the Invention

[0003] This invention provides a passive personnel positioning method and device based on a multi-region Gaussian probability fusion model to solve the technical problems existing in the prior art.

[0004] The technical solution adopted by this invention to solve the technical problems existing in the prior art is as follows: A passive personnel localization method based on a multi-region Gaussian probability fusion model is proposed. This method establishes a passive localization system comprising: a data processing system; M signal transmitters or signal reflectors installed around the perimeter of the localization space; and N signal receivers. A line-of-sight link is formed between any signal transmitter or signal reflector and any signal receiver, resulting in a total of [number missing]. Line-of-sight links; the data processing system is used to collect signals received by the signal receiver and determine the personnel's position based on the relationship between the signal strength received by the signal receiver and the corresponding line-of-sight links; Let real people located in the positioning space be considered true targets, and targets other than real people that are located be considered false targets. Line-of-sight links occluded by true targets are used to extract true target samples. The data processing system clusters the true target samples, resulting in multiple sets. The region occupied by all true targets in each set is called a training sub-region. Based on the number of true and false targets within each training sub-region, the training sub-regions are classified as follows: training sub-region with many true target samples, training sub-region with few true target samples, training sub-region with many false target samples, and training sub-region with few false target samples. Within the data processing system... The system sets up corresponding Gaussian distribution models for real targets (multiple samples), real targets (few samples), and pseudo-targets (multiple samples), and pseudo-targets (few samples). The parameters of the corresponding Gaussian distribution models are obtained by training real target samples and / or pseudo-target samples within each training sub-region. The data processing system processes the signals from each line-of-sight link acquired in real time, extracts the feature vectors corresponding to each candidate target, substitutes them into the Gaussian distribution model corresponding to the training sub-region where the candidate target is located, and fuses multiple Gaussian distribution models to determine the authenticity of the candidate targets and the location of the real targets.

[0005] Furthermore, the data processing system clusters the true target samples and obtains multiple training sub-regions through the cluster sets. The method for classifying the training sub-regions includes the following steps: Step A1: When there are no people or objects in the positioning space, write the signal strength of the signals received by the signal receivers in the entire positioning space, the signals transmitted by the signal transmitters, or the signals reflected by the signal reflectors, into a matrix. ;set up It is a matrix The element in the c-th row and d-th column, , ; Suppose that P preset targets are standing in different positions in the positioning space, and there are a total of W sets of station distributions. Let e ​​represent the station distribution group number corresponding to the positions of the P preset targets in the positioning space. The signal strength of the corresponding e-th group of stations is collected using a signal receiver, and the backscattered signal strength of the tags received by the signal receivers in the entire positioning space under this state is written into a matrix. ; matrix With matrix Subtracting these two values ​​yields the signal strength variation matrix caused by target occlusion when P preset targets are located in the e-th group of station positions in the positioning space. ; set up: Representation matrix The element in row c and column d; Representation matrix The element in row c and column d; make ; Set a threshold for the change in signal strength of the communication link when there is a preset target obstruction compared to when there is no preset target obstruction. ; Will Greater than the set threshold The communication link is defined as a shadow link; Step A2: Establish a Cartesian coordinate system on the horizontal plane of the positioning space. This Cartesian coordinate system is called coordinate system O. Project the shadow links onto the horizontal plane of the positioning space. Based on the linear equations in two variables of each shadow link projection line in coordinate system O, calculate the coordinates of the intersection points of all shadow link projection lines in coordinate system O: Density clustering algorithm is used to process the intersections of shadow links in coordinate system O, dividing these intersections into multiple intersection set sets. , Q represents the number of intersection sets, and h represents the index of the intersection set; In coordinate system O, the mean value of the intersection points in each intersection point set is regarded as the estimated value of the station coordinates of a candidate target, and the average value of the intersection points in each intersection point set is calculated. Step A3: Assume the station coordinates of P preset targets are known; match the estimated station coordinates of each candidate target with the known station coordinates of the P preset targets; assign the candidate target to the preset target closest to it. Assigned candidate targets are designated as true targets, and unassigned candidate targets are designated as false targets; a feature database of corresponding preset targets is constructed based on the candidate target assignment results; W sets of preset target station coordinate data are collected, and the number of true targets obtained is set to W. The number of false targets is ; The prior probability of a real target appearing in the entire positioning area is ; The prior probability of a false target appearing in the entire positioning area is ; In the formula: This represents the prior probability of the true target appearing in the entire positioning area. This represents the prior probability of a false target appearing in the entire positioning area. Step A4, let the feature corresponding to the q-th true target in the feature database be... ; It is expressed as follows: ; In the formula: q is the true target index in the feature database; ; Let be the coordinates of the q-th true target; The number of shadow links corresponding to the q-th true target; This represents the signal strength value corresponding to the q-th true target; This represents the signal strength change value corresponding to the q-th true target; This represents the average change in signal intensity corresponding to the q-th candidate target; This represents the variance of the signal intensity change corresponding to the q-th candidate target; Represents the mean function; Represents the variance function; Let the feature set corresponding to the true target in the feature database be . , Hierarchical clustering algorithm is used to analyze the feature set. Clustering is performed to obtain L clusters. The region occupied by each cluster is represented by the set of station coordinates of the true targets in that cluster. The region occupied by each cluster is defined as the training sub-region. Let the i-th... The training sub-regions are , This is the training sub-region index. ; The formula for calculating the center coordinates is as follows: ; ; ; In the formula: For the first Coordinates of the center point of each training sub-region; For the first Number of true targets in each training sub-region; For the first The x-axis coordinates of the center point of each training sub-region; For the first The y-axis coordinates of the center point of each training sub-region; w is the first The true target index in each training sub-region; For the first The coordinates of the w-th true target in each training sub-region; For the first The x-axis coordinate of the w-th true target in each training sub-region; For the first The y-axis coordinate of the w-th true target in each training sub-region.

[0006] Furthermore, the method for classifying training sub-regions based on the number of true and false targets within each training sub-region includes the following steps; The number of real targets and fake targets in each training sub-region is counted and sorted from most to least. Regions with the most real targets are identified as real target multi-sample training sub-regions, and regions with the most real targets in the bottom 50% are identified as real target few-sample training sub-regions. Similarly, regions with the most fake targets in the top 50% are identified as fake target multi-sample training sub-regions, and regions with the most fake targets in the bottom 50% are identified as fake target few-sample training sub-regions. The samples within each training sub-region are then labeled according to their classification.

[0007] Furthermore, the true target multi-sample Gaussian distribution model is set up as follows; Let i represent the index of the multi-sample training sub-region of the true target; the number of true targets in the i-th multi-sample training sub-region of the true target is . ; The set of shadow link counts corresponding to the true target in the i-th real target multi-sample training sub-region is ; The set of mean signal intensity changes corresponding to the real targets in the i-th multi-sample training sub-region is (The rest of the set is missing from the original text) ; The set of variances of signal intensity changes corresponding to the real targets in the i-th real target multi-sample training sub-region is: ; , and All three have the same number of elements. ; The true target multi-sample Gaussian distribution model is set as follows: ; In the formula: This represents the number of shadow links corresponding to the true target in the i-th true target multi-sample training sub-region; This represents the mean signal intensity change of the true target in the multi-sample training sub-region of the i-th true target; This represents the variance of the signal intensity change corresponding to the true target in the i-th true target multi-sample training sub-region; The Gaussian distribution function represents the number of shadow links corresponding to the true target in the i-th true target multi-sample training sub-region; Let represent the Gaussian distribution function of the mean signal intensity change of the true target in the multi-sample training sub-region of the i-th true target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the true target in the multi-sample training sub-region of the i-th true target.

[0008] Furthermore, the Gaussian distribution model of the true target with few samples is set up as follows; Let k represent the index of the few-sample training sub-region of the true target; the number of true targets in the k-th few-sample training sub-region of the true target is . ; The set of shadow link counts corresponding to the true target in the few-sample training sub-region of the k-th true target is: ; The set of mean signal intensity changes corresponding to the k-th real target in the few-sample training sub-region is: ; The set of variances of signal intensity changes corresponding to the true targets in the few-sample training sub-region of the k-th true target is: ; , and All three have the same number of elements. ; The initial model for setting up a small sample Gaussian distribution model for the true target is as follows: ; In the formula: This represents the number of shadow links corresponding to the true target in the few-sample training sub-region of the k-th true target; This represents the mean signal intensity change of the true target in the few-sample training sub-region of the k-th true target; This represents the variance of the signal intensity variation corresponding to the true target in the few-sample training sub-region of the k-th true target; The primary Gaussian distribution function represents the number of shadow links corresponding to the true target in the few-sample training sub-region of the k-th true target; The primary Gaussian distribution function represents the mean signal intensity change of the true target in the few-sample training sub-region of the k-th true target; Let represent the primary Gaussian distribution function of the variance of the signal intensity change corresponding to the true target in the few-sample training sub-region of the k-th true target; The distance between the center point of the k-th true target few-sample training sub-region and the center point of the i-th true target many-sample training sub-region is calculated using the following formula: ; In the formula: This represents the distance between the center point of the k-th true target few-sample training sub-region and the center point of the i-th true target many-sample training sub-region; This represents the coordinates of the center point of the i-th real target multi-sample training sub-region; This represents the coordinates of the center point of the k-th true target training sub-region (few samples). By combining the distance between the center point of the k-th true target few-sample training sub-region and the center point of the i-th true target many-sample training sub-region, the initial model of the true target few-sample Gaussian distribution model is optimized, resulting in the following optimized true target few-sample Gaussian distribution model: ; ; ; In the formula: The high-level Gaussian distribution function representing the number of shadow links corresponding to the true target in the few-sample training sub-region of the k-th true target; The high-level Gaussian distribution function represents the mean signal intensity change of the true target in the few-sample training sub-region of the k-th true target; The high-order Gaussian distribution function represents the variance of the signal intensity change corresponding to the true target in the few-sample training sub-region of the k-th true target; The Gaussian distribution function representing the number of shadow links corresponding to the true target in the multi-sample training sub-region of the i-th true target; Let represent the Gaussian distribution function of the mean signal intensity change corresponding to the true target in the multi-sample training sub-region of the i-th true target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the true target in the multi-sample training sub-region of the i-th true target; This represents the floor function.

[0009] Furthermore, the pseudo-target multi-sample Gaussian distribution model is set up as follows; Let j represent the index of the multi-sample training sub-region of the pseudo-target; the number of pseudo-targets in the j-th multi-sample training sub-region of the pseudo-target is . ; The set of shadow link counts corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target is: ; The set of mean signal intensity changes corresponding to the j-th pseudo-target in the multi-sample training sub-region is: ; The set of variances of signal intensity changes corresponding to the pseudo-targets in the multi-sample training sub-region of the j-th pseudo-target is: ; , and All three have the same number of elements. ; The pseudo-target multi-sample Gaussian distribution model is set as follows: ; In the formula: This represents the number of shadow links corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; This represents the mean signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; This represents the variance of the signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; The Gaussian distribution function representing the number of shadow links corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let represent the Gaussian distribution function of the mean signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the j-th pseudo-target in the multi-sample training sub-region.

[0010] Furthermore, the pseudo-target few-sample Gaussian distribution model is set up as follows; Let g represent the index of the few-shot training sub-region of the pseudo-target; the number of pseudo-targets in the g-th few-shot training sub-region of the pseudo-target is . ; The set of shadow link counts corresponding to the pseudo-target in the few-sample training subregion of the g-th pseudo-target is: ; The set of mean signal intensity changes corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target is: ; The set of variances of signal intensity changes corresponding to the pseudo-targets in the few-sample training sub-region of the g-th pseudo-target is: ; , and All three have the same number of elements. ; The initial model for setting up a pseudo-target Gaussian distribution model with few samples is as follows: ; In the formula: This represents the number of shadow links corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; This represents the mean signal intensity change of the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; This represents the variance of the signal intensity change corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The primary Gaussian distribution function represents the number of shadow links corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; Let denot be the primary Gaussian distribution function representing the mean signal intensity change of the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; Let represent the primary Gaussian distribution function of the variance of the signal intensity change corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The distance between the center point of the g-th pseudo-target few-sample training sub-region and the center point of the j-th pseudo-target many-sample training sub-region is calculated using the following formula: ; In the formula: This represents the distance between the center point of the g-th pseudo-target few-sample training sub-region and the center point of the j-th pseudo-target many-sample training sub-region; This represents the coordinates of the center point of the j-th pseudo-target multi-sample training sub-region; This represents the coordinates of the center point of the few-sample training sub-region of the g-th pseudo-target; By combining the distance between the center point of the g-th pseudo-target few-sample training sub-region and the center point of the j-th pseudo-target many-sample training sub-region, the initial model of the pseudo-target few-sample Gaussian distribution model is optimized, resulting in the following optimized pseudo-target few-sample Gaussian distribution model: ; ; ; In the formula: The high-level Gaussian distribution function representing the number of shadow links corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The high-level Gaussian distribution function represents the mean signal intensity change of the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The high-order Gaussian distribution function represents the variance of the signal intensity change corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The Gaussian distribution function representing the number of shadow links corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let be the Gaussian distribution function representing the mean of the signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; This represents the floor function.

[0011] Furthermore, the method for determining personnel location by real-time acquisition and processing of signals from each line-of-sight link in the data processing system includes the following steps: Step B1: Use a signal receiver to collect the received signal strength at time t in real time, and generate a matrix based on the signal strength of the signal transmitted by the transmitter or reflected by the signal reflector received by each signal receiver at time t. ; matrix With matrix Subtracting these two matrices yields the signal intensity change matrix at time t due to target occlusion. ; set up: Representation matrix The element in row c and column d; Representation matrix The element in row c and column d; make ; Will Greater than the set threshold The communication link is defined as a shadow link; Step B2: Project the shadow links obtained at time t onto the horizontal plane of the positioning space. Based on the linear equations in two variables of each shadow link projection line in coordinate system O, calculate the coordinates of the intersection points of all shadow link projection lines at time t in coordinate system O: Step B3: Use density clustering algorithm to process the intersection points of the shadow links in coordinate system O at time t, and divide these intersection points into multiple intersection point sets. , Z represents the number of intersection sets at time t, and u represents the index of the intersection set at time t. In coordinate system O, the mean value of the intersection points in each intersection point set at time t is regarded as the estimated value of the station coordinates of a candidate target, and the average value of the intersection points in each intersection point set at time t is calculated. Step B4 calculates the feature information of each candidate target at time t. The expression for the feature information of the candidate target is as follows: ; In the formula: v represents the candidate target number; ; The feature information of the v-th candidate target; Let v be the coordinates of the v-th candidate target; The number of shadow links corresponding to the v-th candidate target; This represents the signal strength value corresponding to the v-th candidate target; This represents the signal intensity change value corresponding to the v-th candidate target; This represents the average signal intensity change corresponding to the v-th candidate target; This represents the variance of the signal intensity change corresponding to the v-th candidate target; Step B5: Match the estimated position coordinates of any candidate target at time t with the known center point coordinates of the training sub-region; assign the candidate target to the training sub-region whose center point is closest to it. Based on the training sub-region to which any candidate target belongs at time t, the feature information of the candidate target is substituted into the Gaussian distribution model corresponding to the training sub-region to calculate the probability that the candidate target is judged as a real target. and the probability of being identified as a false target ; When the v-th candidate target belongs to the numbered When training a sub-region of a true target using multiple samples, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to the true target: ; In the formula: Indicates the first A Gaussian distribution function of the number of shadow links corresponding to real targets in a multi-sample training sub-region; Indicates the first The Gaussian distribution function of the mean signal intensity change of the real target in a multi-sample training sub-region; Indicates the first The Gaussian distribution function of the variance of the signal intensity change of the real target in the multi-sample training sub-region; When the v-th candidate target belongs to the numbered When training a sub-region with few samples of the true target, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to the true target: ; In the formula: Indicates the first A high-level Gaussian distribution function of the number of shadow links corresponding to real targets in a few sample training sub-regions; Indicates the first The high-level Gaussian distribution function of the mean signal intensity change of the real target in a small sample training sub-region; Indicates the first The high-order Gaussian distribution function of the variance of signal intensity change corresponding to a real target in a small sample training sub-region; When the v-th candidate target belongs to the numbered When training a sub-region using multiple samples of pseudo-targets, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to a pseudo-target: ; In the formula: Indicates the first A Gaussian distribution function of the number of shadow links corresponding to pseudo-targets in a multi-sample training sub-region; Indicates the first The Gaussian distribution function of the mean signal intensity change of the pseudo-target in the multi-sample training sub-region; Indicates the first The Gaussian distribution function of the variance of the signal intensity change of each pseudo-target in the multi-sample training sub-region; When the v-th candidate target belongs to the numbered When training a sub-region with few samples of pseudo-targets, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to a pseudo-target: ; In the formula: Indicates the first A high-level Gaussian distribution function of the number of shadow links corresponding to pseudo-targets in a few-sample training sub-region; Indicates the first The high-level Gaussian distribution function of the mean signal intensity change of pseudo-targets in a small sample training sub-region; Indicates the first The high-order Gaussian distribution function of the variance of signal intensity change corresponding to pseudo-targets in a small sample training sub-region; The authenticity of a candidate target is determined by the probability that it belongs to a true or false target. This would be a false target; Then it is the true target; Step B6: Take the center position of the set of intersection points of each candidate target that is judged as a real target as the coordinates of that target.

[0012] Furthermore, the signal receiver adopts a ZigBee signal receiver, a Bluetooth signal receiver, or an RFID receiver; a ZigBee signal transmitter and a ZigBee signal receiver are used to form a line-of-sight link; a Bluetooth signal transmitter and a Bluetooth signal receiver are used to form a line-of-sight link; an RFID receiver uses an RFID tag as a signal reflector to form a line-of-sight link; the RFID transmitting antenna and the RFID receiver are integrated together, the RFID transmitting antenna transmits radio frequency signals to the RFID tag, and the RFID receiver receives the radio frequency signals reflected by the RFID tag.

[0013] The present invention also provides an apparatus for a passive personnel localization method based on a multi-region Gaussian probability fusion model, comprising a memory and a processor, wherein the memory is used to store a computer program; and the processor is used to execute the computer program and, when executing the computer program, implement the steps of the passive personnel localization method based on the multi-region Gaussian probability fusion model as described above.

[0014] The advantages and positive effects of this invention are: it can accurately identify false targets in the positioning results and significantly reduce the manual and time costs of collecting training data, which can greatly promote the application of passive human positioning technology in smart buildings, smart cities and other fields. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the passive personnel localization method based on a multi-region Gaussian probability fusion model according to the present invention.

[0016] Figure 2 This is a schematic diagram of the shadow link projected onto the horizontal plane of the positioning space by the RFID system.

[0017] Figure 3 This paper compares the cumulative distribution function of the counting error of a passive personnel positioning method based on a multi-region Gaussian probability fusion model and a traditional geometric positioning algorithm when using RFID devices.

[0018] Figure 4 This is a schematic diagram of the shadowed link projected onto a Cartesian coordinate system on the horizontal plane of the positioning space based on a ZigBee or Bluetooth networking system. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0020] In the description of this invention, the terms "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," and "bottom," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and do not require the invention to be constructed and operated in a specific orientation; therefore, they should not be construed as limitations on the invention. The terms "connected" and "linked" used in this invention should be interpreted broadly. For example, they can refer to a fixed connection or a detachable connection; a direct connection or an indirect connection through intermediate components; or an electrical connection or signal transmission. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0021] Please see Figures 1 to 4 A passive personnel positioning method based on a multi-region Gaussian probability fusion model is proposed. This method establishes a passive positioning system comprising: a data processing system; M signal transmitters or signal reflectors installed around the perimeter of the positioning space; and N signal receivers. A line-of-sight link is formed between any one signal transmitter or signal reflector and any one signal receiver, resulting in a total of [number missing]. Line-of-sight links; the data processing system is used to collect signals received by the signal receiver and determine the personnel's position based on the relationship between the signal strength received by the signal receiver and the corresponding line-of-sight links; Let real people located in the positioning space be considered true targets, and targets other than real people that are located be considered false targets. Line-of-sight links occluded by true targets are used to extract true target samples. The data processing system clusters the true target samples, resulting in multiple sets. The region occupied by all true targets in each set is called a training sub-region. Based on the number of true and false targets within each training sub-region, the training sub-regions are classified as follows: training sub-region with many true target samples, training sub-region with few true target samples, training sub-region with many false target samples, and training sub-region with few false target samples. Within the data processing system... The system sets up corresponding Gaussian distribution models for real targets (multiple samples), real targets (few samples), and pseudo-targets (multiple samples), and pseudo-targets (few samples). The parameters of the corresponding Gaussian distribution models are obtained by training real target samples and / or pseudo-target samples within each training sub-region. The data processing system processes the signals from each line-of-sight link acquired in real time, extracts the feature vectors corresponding to each candidate target, substitutes them into the Gaussian distribution model corresponding to the training sub-region where the candidate target is located, and fuses multiple Gaussian distribution models to determine the authenticity of the candidate targets and the location of the real targets.

[0022] Preferably, the data processing system clusters the true target samples and obtains multiple training sub-regions through the cluster set. The method for classifying the training sub-regions may include the following steps: Step A1: When there are no people or objects in the positioning space, write the signal strength of the signals received by the signal receivers in the entire positioning space, the signals transmitted by the signal transmitters, or the signals reflected by the signal reflectors, into a matrix. ;set up It is a matrix The element in the c-th row and d-th column, , ; Suppose that P preset targets are standing in different positions in the positioning space, and there are a total of W sets of station distributions. Let e ​​represent the station distribution group number corresponding to the positions of the P preset targets in the positioning space. The signal strength of the corresponding e-th group of stations is collected using a signal receiver, and the backscattered signal strength of the tags received by the signal receivers in the entire positioning space under this state is written into a matrix. ; matrix With matrix Subtracting these two values ​​yields the signal strength variation matrix caused by target occlusion when P preset targets are located in the e-th group of station positions in the positioning space. ; set up: Representation matrix The element in row c and column d; Representation matrix The element in row c and column d; make ; Set a threshold for the change in signal strength of the communication link when there is a preset target obstruction compared to when there is no preset target obstruction. ; Will Greater than the set threshold The communication link is defined as a shadow link; Step A2: Establish a Cartesian coordinate system on the horizontal plane of the positioning space. This Cartesian coordinate system is called coordinate system O. Project the shadow links onto the horizontal plane of the positioning space. Based on the linear equations in two variables of each shadow link projection line in coordinate system O, calculate the coordinates of the intersection points of all shadow link projection lines in coordinate system O: Density clustering algorithm is used to process the intersections of shadow links in coordinate system O, dividing these intersections into multiple intersection set sets. , Q represents the number of intersection sets, and h represents the index of the intersection set; In coordinate system O, the mean value of the intersection points in each intersection point set is regarded as the estimated value of the station coordinates of a candidate target, and the average value of the intersection points in each intersection point set is calculated. Step A3: Assume the station coordinates of P preset targets are known; match the estimated station coordinates of each candidate target with the known station coordinates of the P preset targets; assign the candidate target to the preset target closest to it. Assigned candidate targets are designated as true targets, and unassigned candidate targets are designated as false targets; a feature database of corresponding preset targets is constructed based on the candidate target assignment results; W sets of preset target station coordinate data are collected, and the number of true targets obtained is set to W. The number of false targets is ; The prior probability of a real target appearing in the entire positioning area is ; The prior probability of a false target appearing in the entire positioning area is ; In the formula: This represents the prior probability of the true target appearing in the entire positioning area. This represents the prior probability of a false target appearing in the entire positioning area. Step A4, let the feature corresponding to the q-th true target in the feature database be... ; It is expressed as follows: ; In the formula: q is the true target index in the feature database; ; Let be the coordinates of the q-th true target; The number of shadow links corresponding to the q-th true target; This represents the signal strength value corresponding to the q-th true target; This represents the signal strength change value corresponding to the q-th true target; This represents the average change in signal intensity corresponding to the q-th candidate target; This represents the variance of the signal intensity change corresponding to the q-th candidate target; Represents the mean function; Represents the variance function; Let the feature set corresponding to the true target in the feature database be . , Hierarchical clustering algorithm is used to analyze the feature set. Clustering is performed to obtain L clusters. The region occupied by each cluster is represented by the set of station coordinates of the true targets in that cluster. The region occupied by each cluster is defined as the training sub-region. Let the i-th... The training sub-regions are , This is the training sub-region index. ; The formula for calculating the center coordinates is as follows: ; ; ; In the formula: For the first Coordinates of the center point of each training sub-region; For the first Number of true targets in each training sub-region; For the first The x-axis coordinates of the center point of each training sub-region; For the first The y-axis coordinates of the center point of each training sub-region; w is the first The true target index in each training sub-region; For the first The coordinates of the w-th true target in each training sub-region; For the first The x-axis coordinate of the w-th true target in each training sub-region; For the first The y-axis coordinate of the w-th true target in each training sub-region.

[0023] Preferably, the method for classifying training sub-regions based on the number of true and false targets within each training sub-region may include the following steps; The number of real targets and fake targets in each training sub-region is counted and sorted from most to least. Regions with the most real targets are identified as real target multi-sample training sub-regions, and regions with the most real targets in the bottom 50% are identified as real target few-sample training sub-regions. Similarly, regions with the most fake targets in the top 50% are identified as fake target multi-sample training sub-regions, and regions with the most fake targets in the bottom 50% are identified as fake target few-sample training sub-regions. The samples within each training sub-region are then labeled according to their classification.

[0024] Preferably, the true target multi-sample Gaussian distribution model can be set up as follows; Let i represent the index of the multi-sample training sub-region of the true target; the number of true targets in the i-th multi-sample training sub-region of the true target is . ; The set of shadow link counts corresponding to the true target in the i-th real target multi-sample training sub-region is ; The set of mean signal intensity changes corresponding to the real targets in the i-th multi-sample training sub-region is (The rest of the set is missing from the original text) ; The set of variances of signal intensity changes corresponding to the real targets in the i-th real target multi-sample training sub-region is: ; , and All three have the same number of elements. ; The true target multi-sample Gaussian distribution model is set as follows: ; In the formula: This represents the number of shadow links corresponding to the true target in the i-th true target multi-sample training sub-region; This represents the mean signal intensity change of the true target in the multi-sample training sub-region of the i-th true target; This represents the variance of the signal intensity change corresponding to the true target in the i-th true target multi-sample training sub-region; The Gaussian distribution function represents the number of shadow links corresponding to the true target in the i-th true target multi-sample training sub-region; Let represent the Gaussian distribution function of the mean signal intensity change of the true target in the multi-sample training sub-region of the i-th true target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the true target in the multi-sample training sub-region of the i-th true target.

[0025] Preferably, the Gaussian distribution model of the true target with few samples can be set as follows; Let k represent the index of the few-sample training sub-region of the true target; the number of true targets in the k-th few-sample training sub-region of the true target is . ; The set of shadow link counts corresponding to the true target in the few-sample training sub-region of the k-th true target is: ; The set of mean signal intensity changes corresponding to the k-th real target in the few-sample training sub-region is: ; The set of variances of signal intensity changes corresponding to the true targets in the few-sample training sub-region of the k-th true target is: ; , and All three have the same number of elements. ; The initial model for setting up a small sample Gaussian distribution model for the true target is as follows: ; In the formula: This represents the number of shadow links corresponding to the true target in the few-sample training sub-region of the k-th true target; This represents the mean signal intensity change of the true target in the few-sample training sub-region of the k-th true target; This represents the variance of the signal intensity variation corresponding to the true target in the few-sample training sub-region of the k-th true target; The primary Gaussian distribution function represents the number of shadow links corresponding to the true target in the few-sample training sub-region of the k-th true target; The primary Gaussian distribution function represents the mean signal intensity change of the true target in the few-sample training sub-region of the k-th true target; Let represent the primary Gaussian distribution function of the variance of the signal intensity change corresponding to the true target in the few-sample training sub-region of the k-th true target; The distance between the center point of the k-th true target few-sample training sub-region and the center point of the i-th true target many-sample training sub-region is calculated using the following formula: ; In the formula: This represents the distance between the center point of the k-th true target few-sample training sub-region and the center point of the i-th true target many-sample training sub-region; This represents the coordinates of the center point of the i-th real target multi-sample training sub-region; This represents the coordinates of the center point of the k-th true target training sub-region (few samples). By combining the distance between the center point of the k-th true target few-sample training sub-region and the center point of the i-th true target many-sample training sub-region, the initial model of the true target few-sample Gaussian distribution model is optimized, resulting in the following optimized true target few-sample Gaussian distribution model: ; ; ; In the formula: The high-level Gaussian distribution function representing the number of shadow links corresponding to the true target in the few-sample training sub-region of the k-th true target; The high-level Gaussian distribution function represents the mean signal intensity change of the true target in the few-sample training sub-region of the k-th true target; The high-order Gaussian distribution function represents the variance of the signal intensity change corresponding to the true target in the few-sample training sub-region of the k-th true target; The Gaussian distribution function representing the number of shadow links corresponding to the true target in the multi-sample training sub-region of the i-th true target; Let represent the Gaussian distribution function of the mean signal intensity change corresponding to the true target in the multi-sample training sub-region of the i-th true target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the true target in the multi-sample training sub-region of the i-th true target; This represents the floor function.

[0026] Preferably, the pseudo-target multi-sample Gaussian distribution model can be set up as follows; Let j represent the index of the multi-sample training sub-region of the pseudo-target; the number of pseudo-targets in the j-th multi-sample training sub-region of the pseudo-target is . ; The set of shadow link counts corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target is: ; The set of mean signal intensity changes corresponding to the j-th pseudo-target in the multi-sample training sub-region is: ; The set of variances of signal intensity changes corresponding to the pseudo-targets in the multi-sample training sub-region of the j-th pseudo-target is: ; , and All three have the same number of elements. ; The pseudo-target multi-sample Gaussian distribution model is set as follows: ; In the formula: This represents the number of shadow links corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; This represents the mean signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; This represents the variance of the signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; The Gaussian distribution function representing the number of shadow links corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let represent the Gaussian distribution function of the mean signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the j-th pseudo-target in the multi-sample training sub-region.

[0027] Preferably, the pseudo-target few-sample Gaussian distribution model can be set as follows; Let g represent the index of the few-shot training sub-region of the pseudo-target; the number of pseudo-targets in the g-th few-shot training sub-region of the pseudo-target is . ; The set of shadow link counts corresponding to the pseudo-target in the few-sample training subregion of the g-th pseudo-target is: ; The set of mean signal intensity changes corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target is: ; The set of variances of signal intensity changes corresponding to the pseudo-targets in the few-sample training sub-region of the g-th pseudo-target is: ; , and All three have the same number of elements. ; The initial model for setting up a pseudo-target Gaussian distribution model with few samples is as follows: ; In the formula: This represents the number of shadow links corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; This represents the mean signal intensity change of the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; This represents the variance of the signal intensity change corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The primary Gaussian distribution function represents the number of shadow links corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; Let denot be the primary Gaussian distribution function representing the mean signal intensity change of the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; Let represent the primary Gaussian distribution function of the variance of the signal intensity change corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The distance between the center point of the g-th pseudo-target few-sample training sub-region and the center point of the j-th pseudo-target many-sample training sub-region is calculated using the following formula: ; In the formula: This represents the distance between the center point of the g-th pseudo-target few-sample training sub-region and the center point of the j-th pseudo-target many-sample training sub-region; This represents the coordinates of the center point of the j-th pseudo-target multi-sample training sub-region; This represents the coordinates of the center point of the few-sample training sub-region of the g-th pseudo-target; By combining the distance between the center point of the g-th pseudo-target few-sample training sub-region and the center point of the j-th pseudo-target many-sample training sub-region, the initial model of the pseudo-target few-sample Gaussian distribution model is optimized, resulting in the following optimized pseudo-target few-sample Gaussian distribution model: ; ; ; In the formula: The high-level Gaussian distribution function representing the number of shadow links corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The high-level Gaussian distribution function represents the mean signal intensity change of the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The high-order Gaussian distribution function represents the variance of the signal intensity change corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The Gaussian distribution function representing the number of shadow links corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let be the Gaussian distribution function representing the mean of the signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; This represents the floor function.

[0028] Preferably, the method for determining the location of personnel by real-time acquisition and processing of signals from each line-of-sight link by the data processing system may include the following steps: Step B1: Use a signal receiver to collect the received signal strength at time t in real time, and generate a matrix based on the signal strength of the signal transmitted by the transmitter or reflected by the signal reflector received by each signal receiver at time t. ; matrix With matrix Subtracting these two matrices yields the signal intensity change matrix at time t due to target occlusion. ; set up: Representation matrix The element in row c and column d; Representation matrix The element in row c and column d; make ; Will Greater than the set threshold The communication link is defined as a shadow link; Step B2: Project the shadow links obtained at time t onto the horizontal plane of the positioning space. Based on the linear equations in two variables of each shadow link projection line in coordinate system O, calculate the coordinates of the intersection points of all shadow link projection lines at time t in coordinate system O: Step B3: Use density clustering algorithm to process the intersection points of the shadow links in coordinate system O at time t, and divide these intersection points into multiple intersection point sets. , Z represents the number of intersection sets at time t, and u represents the index of the intersection set at time t. In coordinate system O, the mean value of the intersection points in each intersection point set at time t is regarded as the estimated value of the station coordinates of a candidate target, and the average value of the intersection points in each intersection point set at time t is calculated. Step B4 calculates the feature information of each candidate target at time t. The expression for the feature information of the candidate target is as follows: ; In the formula: v represents the candidate target number; ; The feature information of the v-th candidate target; Let v be the coordinates of the v-th candidate target; The number of shadow links corresponding to the v-th candidate target; This represents the signal strength value corresponding to the v-th candidate target; This represents the signal intensity change value corresponding to the v-th candidate target; This represents the average signal intensity change corresponding to the v-th candidate target; This represents the variance of the signal intensity change corresponding to the v-th candidate target; Step B5: Match the estimated position coordinates of any candidate target at time t with the known center point coordinates of the training sub-region; assign the candidate target to the training sub-region whose center point is closest to it. Based on the training sub-region to which any candidate target belongs at time t, the feature information of the candidate target is substituted into the Gaussian distribution model corresponding to the training sub-region to calculate the probability that the candidate target is judged as a real target. and the probability of being identified as a false target ; When the v-th candidate target belongs to the numbered When training a sub-region of a true target using multiple samples, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to the true target: ; In the formula: Indicates the first A Gaussian distribution function of the number of shadow links corresponding to real targets in a multi-sample training sub-region; Indicates the first The Gaussian distribution function of the mean signal intensity change of the real target in a multi-sample training sub-region; Indicates the first The Gaussian distribution function of the variance of the signal intensity change of the real target in the multi-sample training sub-region; When the v-th candidate target belongs to the numbered When training a sub-region with few samples of the true target, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to the true target: ; In the formula: Indicates the first A high-level Gaussian distribution function of the number of shadow links corresponding to real targets in a few sample training sub-regions; Indicates the first The high-level Gaussian distribution function of the mean signal intensity change of the real target in a small sample training sub-region; Indicates the first The high-order Gaussian distribution function of the variance of signal intensity change corresponding to a real target in a small sample training sub-region; When the v-th candidate target belongs to the numbered When training a sub-region using multiple samples of pseudo-targets, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to a pseudo-target: ; In the formula: Indicates the first A Gaussian distribution function of the number of shadow links corresponding to pseudo-targets in a multi-sample training sub-region; Indicates the first The Gaussian distribution function of the mean signal intensity change of the pseudo-target in the multi-sample training sub-region; Indicates the first The Gaussian distribution function of the variance of the signal intensity change of each pseudo-target in the multi-sample training sub-region; When the v-th candidate target belongs to the numbered When training a sub-region with few samples of pseudo-targets, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to a pseudo-target: ; In the formula: Indicates the first A high-level Gaussian distribution function of the number of shadow links corresponding to pseudo-targets in a few-sample training sub-region; Indicates the first The high-level Gaussian distribution function of the mean signal intensity change of pseudo-targets in a small sample training sub-region; Indicates the first The high-order Gaussian distribution function of the variance of signal intensity change corresponding to pseudo-targets in a small sample training sub-region; The authenticity of a candidate target is determined by the probability that it belongs to a true or false target. This would be a false target; Then it is the true target; Step B6: Take the center position of the set of intersection points of each candidate target that is judged as a real target as the coordinates of that target.

[0029] Preferably, the signal receiver can be a ZigBee signal receiver, a Bluetooth signal receiver, or an RFID receiver; a ZigBee signal transmitter and a ZigBee signal receiver are used to form a line-of-sight link; a Bluetooth signal transmitter and a Bluetooth signal receiver are used to form a line-of-sight link; an RFID receiver uses an RFID tag as a signal reflector to form a line-of-sight link; an RFID transmitting antenna and an RFID receiver are integrated together, the RFID transmitting antenna transmits radio frequency signals to the RFID tag, and the RFID receiver receives the radio frequency signals reflected by the RFID tag.

[0030] Preferably, the positioning system may include: an RFID receiver, N RFID antennas, and M RFID tags; each RFID antenna is connected to the RFID receiver port via a feeder, and the RFID receiver controls each RFID antenna to send and receive signals; the N RFID antennas and M RFID tags are installed around the perimeter of the positioning space, and a line-of-sight link is formed between any RFID antenna and any RFID tag, forming a total [missing information - likely a specific configuration]. Line-of-sight links; the relationship between the strength of the backscattered signal from the tag received by the RFID antenna and the corresponding line-of-sight links is used to determine the location of personnel. The deployment scenario for positioning is shown in the attached diagram. Figure 2 The cumulative distribution function of the counting error of the number of people in the model simulation is as follows: Figure 3 As shown.

[0031] RFID is an abbreviation for Radio Frequency Identification. An RFID receiver, also known as an RFID reader, is a non-contact identification device that automatically identifies electronic tags and collects data through radio frequency signals. As a core component of an RFID system, it is responsible for wireless communication with the tags to read or write data. The RFID antenna is a key component of the RFID system, mainly used for wirelessly transmitting radio frequency signals between the reader and the electronic tag, enabling data reception, transmission, and energy transfer. An RFID electronic tag is a core component of radio frequency identification technology, also known as an RFID tag, inductive tag, or electronic tag. The working principle of an RFID tag is based on electromagnetic induction or microwave communication. When a tag enters the reader's magnetic field range, passive tags obtain energy through induced current and send data, while active tags actively transmit signals.

[0032] Preferably, the positioning system may include: a ZigBee signal transmitter and a ZigBee signal receiver, and the positioning scenario deployment is as shown in the appendix. Figure 4 As shown.

[0033] ZigBee is a short-range, low-power, low-data-rate wireless communication technology primarily used for device interconnection and data transmission in the Internet of Things (IoT) field. It supports self-organizing network topologies, including star, tree, and mesh structures, to achieve high reliability and network self-healing capabilities. A ZigBee transmitter typically refers to a device or module in a ZigBee network responsible for transmitting wireless signals. A ZigBee receiver typically refers to a device or component in a ZigBee network responsible for receiving wireless data.

[0034] Preferably, the positioning system may include: a Bluetooth signal transmitter and a Bluetooth signal receiver, and the positioning scenario deployment is as shown in the appendix. Figure 4 As shown.

[0035] The present invention also provides an apparatus for a passive personnel localization method based on a multi-region Gaussian probability fusion model, comprising a memory and a processor, wherein the memory is used to store a computer program; and the processor is used to execute the computer program and, when executing the computer program, implement the steps of the passive personnel localization method based on the multi-region Gaussian probability fusion model as described above.

[0036] Working principle of this invention: A line-of-sight (LoS) link is a straight-line transmission path between two communication devices, where no obstacles or interference affect signal propagation. Such links are typically used in open spaces or flat areas, such as in broadcasting, satellite communications, laser communications, and radio communications. The effective transmission distance of a line-of-sight link depends on the transmit power, receiver sensitivity, frequency band used, and attenuation characteristics of the transmission medium. A line-of-sight link (LoS link) is a wireless communication path between communication devices that propagates in a straight line without physical obstacles. It offers high transmission quality but has limited range and is commonly used in satellite communications, microwave relays, and drones.

Claims

1. A passive personnel localization method based on a multi-region Gaussian probability fusion model, characterized in that, This method employs a passive positioning system, comprising: a data processing system; M signal transmitters or signal reflectors installed around the perimeter of the positioning space; and N signal receivers. A line-of-sight link is formed between any one signal transmitter or signal reflector and any one signal receiver, resulting in a total of [number missing]. Line-of-sight links; the data processing system is used to collect signals received by the signal receiver and determine the personnel's position based on the relationship between the signal strength received by the signal receiver and the corresponding line-of-sight links; Let real people located in the positioning space be considered true targets, and targets other than real people that are located be considered false targets. Line-of-sight links occluded by true targets are used to extract true target samples. The data processing system clusters the true target samples, resulting in multiple sets. The region occupied by all true targets in each set is called a training sub-region. Based on the number of true and false targets within each training sub-region, the training sub-regions are classified as follows: training sub-region with many true target samples, training sub-region with few true target samples, training sub-region with many false target samples, and training sub-region with few false target samples. Within the data processing system... The system sets up corresponding Gaussian distribution models for real targets (multiple samples), real targets (few samples), and pseudo-targets (multiple samples), and pseudo-targets (few samples). The parameters of the corresponding Gaussian distribution models are obtained by training real target samples and / or pseudo-target samples within each training sub-region. The data processing system processes the signals from each line-of-sight link acquired in real time, extracts the feature vectors corresponding to each candidate target, substitutes them into the Gaussian distribution model corresponding to the training sub-region where the candidate target is located, and fuses multiple Gaussian distribution models to determine the authenticity of the candidate targets and the location of the real targets.

2. The passive personnel localization method based on the multi-region Gaussian probability fusion model according to claim 1, characterized in that, The data processing system clusters real target samples and obtains multiple training sub-regions from the cluster sets. The method for classifying the training sub-regions includes the following steps: Step A1: When there are no people or objects in the positioning space, write the signal strength of the signals received by the signal receivers in the entire positioning space, the signals transmitted by the signal transmitters, or the signals reflected by the signal reflectors, into a matrix. ;set up It is a matrix The element in the c-th row and d-th column, , ; Suppose that P preset targets are standing in different positions in the positioning space, and there are a total of W sets of station distributions. Let e ​​represent the station distribution group number corresponding to the positions of the P preset targets in the positioning space. The signal strength of the corresponding e-th group of stations is collected using a signal receiver, and the backscattered signal strength of the tags received by the signal receivers in the entire positioning space under this state is written into a matrix. ; matrix With matrix Subtracting these two values ​​yields the signal strength variation matrix caused by target occlusion when P preset targets are located in the e-th group of station positions in the positioning space. ; set up: Representation matrix The element in row c and column d; Representation matrix The element in row c and column d; make ; Set a threshold for the change in signal strength of the communication link when there is a preset target obstruction compared to when there is no preset target obstruction. ; Will Greater than the set threshold The communication link is defined as a shadow link; Step A2: Establish a Cartesian coordinate system on the horizontal plane of the positioning space. This Cartesian coordinate system is called coordinate system O. Project the shadow links onto the horizontal plane of the positioning space. Based on the linear equations in two variables of each shadow link projection line in coordinate system O, calculate the coordinates of the intersection points of all shadow link projection lines in coordinate system O: Density clustering algorithm is used to process the intersections of shadow links in coordinate system O, dividing these intersections into multiple intersection set sets. , Q represents the number of intersection sets, and h represents the index of the intersection set; In coordinate system O, the mean value of the intersection points in each intersection point set is regarded as the estimated value of the station coordinates of a candidate target, and the average value of the intersection points in each intersection point set is calculated. Step A3: Assume the station coordinates of P preset targets are known; match the estimated station coordinates of each candidate target with the known station coordinates of the P preset targets; assign the candidate target to the preset target closest to it. Assigned candidate targets are designated as true targets, and unassigned candidate targets are designated as false targets; a feature database of corresponding preset targets is constructed based on the candidate target assignment results; W sets of preset target station coordinate data are collected, and the number of true targets obtained is set to W. The number of false targets is ; The prior probability of a real target appearing in the entire positioning area is ; The prior probability of a false target appearing in the entire positioning area is ; In the formula: This represents the prior probability of the true target appearing in the entire positioning area. This represents the prior probability of a false target appearing in the entire positioning area. Step A4, let the feature corresponding to the q-th true target in the feature database be... ; It is expressed as follows: ； In the formula: q is the true target index in the feature database; ; Let be the coordinates of the q-th true target; The number of shadow links corresponding to the q-th true target; This represents the signal strength value corresponding to the q-th true target; This represents the signal strength change value corresponding to the q-th true target; This represents the average change in signal intensity corresponding to the q-th candidate target; This represents the variance of the signal intensity change corresponding to the q-th candidate target; Represents the mean function; Represents the variance function; Let the feature set corresponding to the true target in the feature database be . , Hierarchical clustering algorithm is used to analyze the feature set. Clustering is performed to obtain L clusters. The region occupied by each cluster is represented by the set of station coordinates of the true targets in that cluster. The region occupied by each cluster is defined as the training sub-region. Let the i-th... The training sub-regions are , This is the training sub-region index. ; The formula for calculating the center coordinates is as follows: ； ； ； In the formula: For the first Coordinates of the center point of each training sub-region; For the first Number of true targets in each training sub-region; For the first The x-axis coordinates of the center point of each training sub-region; For the first The y-axis coordinates of the center point of each training sub-region; w is the first The true target index in each training sub-region; For the first The coordinates of the w-th true target in each training sub-region; For the first The x-axis coordinate of the w-th true target in each training sub-region; For the first The y-axis coordinate of the w-th true target in each training sub-region.

3. The passive personnel localization method based on the multi-region Gaussian probability fusion model according to claim 2, characterized in that, The method for classifying training sub-regions based on the number of real and false targets within each training sub-region includes the following steps; The number of real targets and fake targets in each training sub-region is counted and sorted from most to least. Regions with the most real targets are identified as real target multi-sample training sub-regions, and regions with the most real targets in the bottom 50% are identified as real target few-sample training sub-regions. Similarly, regions with the most fake targets in the top 50% are identified as fake target multi-sample training sub-regions, and regions with the most fake targets in the bottom 50% are identified as fake target few-sample training sub-regions. The samples within each training sub-region are then labeled according to their classification.

4. The passive personnel localization method based on the multi-region Gaussian probability fusion model according to claim 2, characterized in that, Set up the true target multi-sample Gaussian distribution model as follows; set up : i represents the index of the multi-sample training sub-region of the true target; the number of true targets in the i-th multi-sample training sub-region of the true target is . ; The set of shadow link counts corresponding to the true target in the i-th real target multi-sample training sub-region is ; The set of mean signal intensity changes corresponding to the real targets in the i-th multi-sample training sub-region is (The rest of the set is missing from the original text) ; The set of variances of signal intensity changes corresponding to the real targets in the i-th real target multi-sample training sub-region is: ; , and All three have the same number of elements. ; The true target multi-sample Gaussian distribution model is set as follows: ； In the formula: This represents the number of shadow links corresponding to the true target in the i-th true target multi-sample training sub-region; This represents the mean signal intensity change of the true target in the multi-sample training sub-region of the i-th true target; This represents the variance of the signal intensity change corresponding to the true target in the i-th true target multi-sample training sub-region; The Gaussian distribution function represents the number of shadow links corresponding to the true target in the i-th true target multi-sample training sub-region; Let represent the Gaussian distribution function of the mean signal intensity change of the true target in the multi-sample training sub-region of the i-th true target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the true target in the multi-sample training sub-region of the i-th true target.

5. The passive personnel localization method based on the multi-region Gaussian probability fusion model according to claim 4, characterized in that, Set up a Gaussian distribution model for a small number of samples of the true target as follows; Let k represent the index of the few-sample training sub-region of the true target; the number of true targets in the k-th few-sample training sub-region of the true target is . ; The set of shadow link counts corresponding to the true target in the few-sample training sub-region of the k-th true target is: ; The set of mean signal intensity changes corresponding to the k-th real target in the few-sample training sub-region is: ; The set of variances of signal intensity changes corresponding to the true targets in the few-sample training sub-region of the k-th true target is: ; , and All three have the same number of elements. ; The initial model for setting up a small sample Gaussian distribution model for the true target is as follows: ； In the formula: This represents the number of shadow links corresponding to the true target in the few-sample training sub-region of the k-th true target; This represents the mean signal intensity change of the true target in the few-sample training sub-region of the k-th true target; This represents the variance of the signal intensity variation corresponding to the true target in the few-sample training sub-region of the k-th true target; The primary Gaussian distribution function represents the number of shadow links corresponding to the true target in the few-sample training sub-region of the k-th true target; The primary Gaussian distribution function represents the mean signal intensity change of the true target in the few-sample training sub-region of the k-th true target; Let represent the primary Gaussian distribution function of the variance of the signal intensity change corresponding to the true target in the few-sample training sub-region of the k-th true target; The distance between the center point of the k-th true target few-sample training sub-region and the center point of the i-th true target many-sample training sub-region is calculated using the following formula: ； In the formula: This represents the distance between the center point of the k-th true target few-sample training sub-region and the center point of the i-th true target many-sample training sub-region; This represents the coordinates of the center point of the i-th real target multi-sample training sub-region; This represents the coordinates of the center point of the k-th true target training sub-region (few samples). By combining the distance between the center point of the k-th true target few-sample training sub-region and the center point of the i-th true target many-sample training sub-region, the initial model of the true target few-sample Gaussian distribution model is optimized, resulting in the following optimized true target few-sample Gaussian distribution model: ； ； ； In the formula: The high-level Gaussian distribution function representing the number of shadow links corresponding to the true target in the few-sample training sub-region of the k-th true target; The high-level Gaussian distribution function represents the mean signal intensity change of the true target in the few-sample training sub-region of the k-th true target; The high-order Gaussian distribution function represents the variance of the signal intensity change corresponding to the true target in the few-sample training sub-region of the k-th true target; The Gaussian distribution function representing the number of shadow links corresponding to the true target in the multi-sample training sub-region of the i-th true target; Let represent the Gaussian distribution function of the mean signal intensity change corresponding to the true target in the multi-sample training sub-region of the i-th true target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the true target in the multi-sample training sub-region of the i-th true target; This represents the floor function.

6. The passive personnel localization method based on the multi-region Gaussian probability fusion model according to claim 5, characterized in that, Set up the pseudo-target multi-sample Gaussian distribution model as follows; Let j represent the index of the multi-sample training sub-region of the pseudo-target; the number of pseudo-targets in the j-th multi-sample training sub-region of the pseudo-target is . ; The set of shadow link counts corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target is: ; The set of mean signal intensity changes corresponding to the j-th pseudo-target in the multi-sample training sub-region is: ; The set of variances of signal intensity changes corresponding to the pseudo-targets in the multi-sample training sub-region of the j-th pseudo-target is: ; , and All three have the same number of elements. ; The pseudo-target multi-sample Gaussian distribution model is set as follows: ； In the formula: This represents the number of shadow links corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; This represents the mean signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; This represents the variance of the signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; The Gaussian distribution function representing the number of shadow links corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let represent the Gaussian distribution function of the mean signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the j-th pseudo-target in the multi-sample training sub-region.

7. The passive personnel localization method based on the multi-region Gaussian probability fusion model according to claim 6, characterized in that, Set up the pseudo-target small sample Gaussian distribution model as follows; Let g represent the index of the few-shot training sub-region of the pseudo-target; the number of pseudo-targets in the g-th few-shot training sub-region of the pseudo-target is . ; The set of shadow link counts corresponding to the pseudo-target in the few-sample training subregion of the g-th pseudo-target is: ; The set of mean signal intensity changes corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target is: ; The set of variances of signal intensity changes corresponding to the pseudo-targets in the few-sample training sub-region of the g-th pseudo-target is: ; , and All three have the same number of elements. ; The initial model for setting up a pseudo-target Gaussian distribution model with few samples is as follows: ； In the formula: This represents the number of shadow links corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; This represents the mean signal intensity change of the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; This represents the variance of the signal intensity change corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The primary Gaussian distribution function represents the number of shadow links corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; Let denot be the primary Gaussian distribution function representing the mean signal intensity change of the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; Let represent the primary Gaussian distribution function of the variance of the signal intensity change corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The distance between the center point of the g-th pseudo-target few-sample training sub-region and the center point of the j-th pseudo-target many-sample training sub-region is calculated using the following formula: ； In the formula: This represents the distance between the center point of the g-th pseudo-target few-sample training sub-region and the center point of the j-th pseudo-target many-sample training sub-region; This represents the coordinates of the center point of the j-th pseudo-target multi-sample training sub-region; This represents the coordinates of the center point of the few-sample training sub-region of the g-th pseudo-target; By combining the distance between the center point of the g-th pseudo-target few-sample training sub-region and the center point of the j-th pseudo-target many-sample training sub-region, the initial model of the pseudo-target few-sample Gaussian distribution model is optimized, resulting in the following optimized pseudo-target few-sample Gaussian distribution model: ； ； ； In the formula: The high-level Gaussian distribution function representing the number of shadow links corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The high-level Gaussian distribution function represents the mean signal intensity change of the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The high-order Gaussian distribution function represents the variance of the signal intensity change corresponding to the pseudo-target in the few-sample training sub-region of the g-th pseudo-target; The Gaussian distribution function representing the number of shadow links corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let be the Gaussian distribution function representing the mean of the signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; Let represent the Gaussian distribution function of the variance of the signal intensity change corresponding to the pseudo-target in the multi-sample training sub-region of the j-th pseudo-target; This represents the floor function.

8. The passive personnel localization method based on the multi-region Gaussian probability fusion model according to claim 7, characterized in that, The data processing system acquires and processes signals from each line-of-sight link in real time to determine personnel location. The method involves the following steps: Step B1: Use a signal receiver to collect the received signal strength at time t in real time, and generate a matrix based on the signal strength of the signal transmitted by the transmitter or reflected by the signal reflector received by each signal receiver at time t. ; matrix With matrix Subtracting these two matrices yields the signal intensity change matrix at time t due to target occlusion. ; set up: Representation matrix The element in row c and column d; Representation matrix The element in row c and column d; make ; Will Greater than the set threshold The communication link is defined as a shadow link; Step B2: Project the shadow links obtained at time t onto the horizontal plane of the positioning space. Based on the linear equations in two variables of each shadow link projection line in coordinate system O, calculate the coordinates of the intersection points of all shadow link projection lines at time t in coordinate system O: Step B3: Use density clustering algorithm to process the intersection points of the shadow links in coordinate system O at time t, and divide these intersection points into multiple intersection point sets. , Z represents the number of intersection sets at time t, and u represents the index of the intersection set at time t. In coordinate system O, the mean value of the intersection points in each intersection point set at time t is regarded as the estimated value of the station coordinates of a candidate target, and the average value of the intersection points in each intersection point set at time t is calculated. Step B4 calculates the feature information of each candidate target at time t. The expression for the feature information of the candidate target is as follows: ； In the formula: v represents the candidate target number; ; The feature information of the v-th candidate target; Let v be the coordinates of the v-th candidate target; The number of shadow links corresponding to the v-th candidate target; This represents the signal strength value corresponding to the v-th candidate target; This represents the signal intensity change value corresponding to the v-th candidate target; This represents the average signal intensity change corresponding to the v-th candidate target; This represents the variance of the signal intensity change corresponding to the v-th candidate target; Step B5: Match the estimated position coordinates of any candidate target at time t with the known center point coordinates of the training sub-region; assign the candidate target to the training sub-region whose center point is closest to it. Based on the training sub-region to which any candidate target belongs at time t, the feature information of the candidate target is substituted into the Gaussian distribution model corresponding to the training sub-region to calculate the probability that the candidate target is judged as a real target. and the probability of being identified as a false target ; When the v-th candidate target belongs to the numbered When training a sub-region of a true target using multiple samples, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to the true target: ； In the formula: Indicates the first A Gaussian distribution function of the number of shadow links corresponding to real targets in a multi-sample training sub-region; Indicates the first The Gaussian distribution function of the mean signal intensity change of the real target in a multi-sample training sub-region; Indicates the first The Gaussian distribution function of the variance of the signal intensity change of the real target in the multi-sample training sub-region; When the v-th candidate target belongs to the numbered When training a sub-region with few samples of the true target, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to the true target: ； In the formula: Indicates the first A high-level Gaussian distribution function of the number of shadow links corresponding to real targets in a few sample training sub-regions; Indicates the first The high-level Gaussian distribution function of the mean signal intensity change of the real target in a small sample training sub-region; Indicates the first The high-order Gaussian distribution function of the variance of signal intensity change corresponding to a real target in a small sample training sub-region; When the v-th candidate target belongs to the numbered When training a sub-region using multiple samples of pseudo-targets, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to a pseudo-target: ； In the formula: Indicates the first A Gaussian distribution function of the number of shadow links corresponding to pseudo-targets in a multi-sample training sub-region; Indicates the first The Gaussian distribution function of the mean signal intensity change of the pseudo-target in the multi-sample training sub-region; Indicates the first The Gaussian distribution function of the variance of the signal intensity change of each pseudo-target in the multi-sample training sub-region; When the v-th candidate target belongs to the numbered When training a sub-region with few samples of pseudo-targets, substitute the feature information of the candidate target into the following formula to calculate the probability that the candidate target belongs to a pseudo-target: ； In the formula: Indicates the first A high-level Gaussian distribution function of the number of shadow links corresponding to pseudo-targets in a few-sample training sub-region; Indicates the first The high-level Gaussian distribution function of the mean signal intensity change of pseudo-targets in a small sample training sub-region; Indicates the first The high-order Gaussian distribution function of the variance of signal intensity change corresponding to pseudo-targets in a small sample training sub-region; The authenticity of a candidate target is determined by the probability that it belongs to a true or false target. This would be a false target; Then it is the true target; Step B6: Take the center position of the set of intersection points of each candidate target that is judged as a real target as the coordinates of that target.

9. The passive personnel localization method based on the multi-region Gaussian probability fusion model according to claim 1, characterized in that, The signal receiver uses a ZigBee signal receiver, a Bluetooth signal receiver, or an RFID receiver; a ZigBee signal transmitter and a ZigBee signal receiver are used together to form a line-of-sight link; a Bluetooth signal transmitter and a Bluetooth signal receiver are used together to form a line-of-sight link; an RFID receiver uses an RFID tag as a signal reflector and is used together to form a line-of-sight link; the RFID transmitting antenna and the RFID receiver are integrated together, the RFID transmitting antenna transmits radio frequency signals to the RFID tag, and the RFID receiver receives the radio frequency signals reflected by the RFID tag.

10. A device for a passive personnel localization method based on a multi-region Gaussian probability fusion model, comprising a memory and a processor, characterized in that, The memory is used to store a computer program; the processor is used to execute the computer program and, when executing the computer program, implement the steps of the passive personnel localization method of the multi-region Gaussian probability fusion model as described in any one of claims 1 to 9.

Images