A high-precision positioning method and system based on optimal combination of regional Ap

Abstract

The present invention provides a high-precision positioning method and system based on the optimal combination of regional Aps. The minimum circumscribed rectangular area of ​​the target area is first determined, recorded as the maximum partition area. In the offline stage, each valid Ap periodically receives the RSSI of all Rps, and the received RSSI data is pre-processed by using the X-score outlier removal method to form a total offline fingerprint database; the number of Aps contained in each sub-fingerprint database is selected, and the sub-fingerprint database is extracted from the total offline fingerprint database. The maximum partition area is divided by using the MAGS algorithm, and the optimal Ap group of each divided sub-area is determined by the RAC algorithm, thereby completing the positioning of the divided sub-area. The present invention reduces unnecessary Ap deployment, enhances robustness to environmental changes, and greatly reduces time, manpower, and material expenditures while improving positioning accuracy.

Description

Technical Field

[0001] The present invention relates to the field of positioning technology, and in particular to a high-precision positioning method and system based on an optimal combination of regions Ap. Background Art

[0002] With the rapid development of the Internet of Things, the demand for precise positioning services from various users is increasing day by day, which has promoted the development of positioning technology to a certain extent. However, in indoor environments, there are environmental interference factors such as concrete structures, which greatly increases the difficulty for users to receive satellite signals and make it difficult to accurately locate based on satellite signals.

[0003] To address this problem, researchers have proposed indoor positioning methods based on various wireless technologies, such as accurate positioning through radio frequency identification, ZigBee, Bluetooth, Wi-Fi, etc. Among them, the positioning method based on Wi-Fi fingerprint has become a hot topic in indoor accurate positioning research due to the relatively low cost of infrastructure and user equipment. However, it should be noted that the indoor positioning process of Wi-Fi fingerprint is divided into two stages: offline and online. In the offline stage, the received signal strength of multiple signal transmitters (Ap) is collected at each reference point (Rp), and together with the corresponding location information, a real fingerprint database is formed; and when the user's fingerprint is received in the online stage, the positioning algorithm is used to retrieve fingerprint information with a high degree of matching, so the positioning accuracy is highly dependent on the establishment of the fingerprint database in the offline stage. In order to obtain higher positioning accuracy, a large amount of reference point fingerprint information data needs to be processed and analyzed, which will cost a lot of time and manpower and material costs. Summary of the invention

[0004] Purpose of the invention: The first purpose of the present invention is to provide a low-cost, high-efficiency, high-precision positioning method based on the optimal combination of area Ap, and the second purpose is to provide a high-precision positioning system based on the optimal combination of area Ap.

[0005] Technical solution: A high-precision positioning method based on the optimal combination of regional Ap, comprising the following steps:

[0006] (1) Determine the smallest circumscribed rectangular area of ​​the target area and record it as the maximum partition area. There are a total of N Aps that can be detected in the maximum partition area;

[0007] (2) In the offline phase, each Ap periodically receives the RSSI of all Rps and preprocesses the received RSSI data to form a total offline fingerprint database;

[0008] (3) Select the number of Aps contained in each sub-fingerprint database, and extract the sub-fingerprint database from the total offline fingerprint database obtained in step (2), and the number of Aps ranges from 3 to N;

[0009] (4) In the offline stage, the MAGS algorithm is used to divide the largest partition area into regions. The RAC algorithm is used to determine the combination of several optimal Aps under each partition sub-region, which is recorded as the optimal Ap group. Several partition sub-regions and the optimal Ap combination corresponding to each partition sub-region are obtained. The sub-fingerprint database corresponding to the optimal Ap combination is extracted, and the positioning of all partition sub-regions is completed by the WKNN positioning algorithm.

[0010] Specifically, in step (4), the MAGS algorithm includes:

[0011] (41) According to the area of ​​the largest partition area, all Rp are evenly distributed and arranged, and the area of ​​the largest single Rp circumscribed circle is set as the minimum positioning partition area, and the largest partition area is divided into 4 triangular partition sub-areas along the diagonal line;

[0012] (42) The combination of the three best Aps under each sub-region is determined by the RAC algorithm, which is recorded as the best Ap group. The sub-region is positioned using the sub-fingerprint database corresponding to the best Ap group. The sub-region whose positioning accuracy reaches the set threshold is no longer further divided. The sub-region and the corresponding best Ap combination are recorded, and the corresponding sub-fingerprint database is extracted. The positioning of the sub-region is completed by the WKNN positioning algorithm. The sub-region whose positioning accuracy does not reach the set threshold is further divided into 4 triangular sub-regions of equal area.

[0013] (43) Repeat step (42) until the positioning accuracy reaches the set threshold, record the sub-regions and the corresponding optimal Ap combination, extract the corresponding sub-fingerprint database, and complete the positioning of the sub-regions through the WKNN positioning algorithm, or the final sub-region reaches the minimum positioning partition area but still cannot reach the positioning accuracy set threshold;

[0014] (44) For the divided sub-areas that cannot reach the set threshold of positioning accuracy through division, increase the number of Aps in the best Ap group, repeat steps (42) and (43) until the positioning Ap group contains N Aps and the final divided sub-area reaches the minimum positioning divided area, record the divided sub-areas whose positioning accuracy reaches the set threshold and the corresponding best Ap combination, extract the corresponding sub-fingerprint database, and complete the positioning of the divided sub-area through the WKNN positioning algorithm;

[0015] (45) After step (44) is completed, the sub-regions whose positioning accuracy still fails to reach the set threshold are marked as areas where positioning cannot be completed. A new Ap is introduced, and steps (42)-(44) are repeated until the positioning accuracy of all sub-regions reaches the set threshold. The corresponding sub-fingerprint database is extracted, and the positioning of all sub-regions is completed through the WKNN positioning algorithm.

[0016] Preferably, in step (2), the received RSSI data is preprocessed using an X-score outlier removal method:

[0017] (21) In the offline phase, each Rp samples the RSSI emitted by the N detectable Aps in the maximum partition area at a fixed time interval, with m+2 samples as a sampling cycle, and collects data for several cycles, where m is a positive integer;

[0018] (22) For one of the cycles n, first calculate the mean of the RSSI data sampled each time by the i-th Rp in cycle n, then use the mean to calculate the discrete value of the RSSI data of the i-th Rp in cycle n, and use the mean and discrete value to calculate the X score of each sample of the i-th Rp in cycle n, remove the two data with the lowest X score from the m+2 data under each Ap in cycle n, and take the mean of the remaining m data to complete the preprocessing;

[0019] (23) Repeat step (22) to obtain the RSSI data of all Rp preprocessed to form the total offline fingerprint database.

[0020] Specifically, the calculation formula for the discrete degree value is:

[0021]

[0022] Where: Represents the discrete degree value of the RSSI data of the hth Ap collected by the i-th Rp in period n; It represents the RSSI data of the hth Ap at the kth sampling time collected by the i-th Rp in period n. It represents the mean value of RSSI data of h-th Ap collected by i-th Rp in period n; m+2 is the number of sampling times in one period.

[0023] Specifically, the calculation formula for the X score is:

[0024]

[0025] Where: is the X score of the RSSI data of the hth Ap at the kth sampling time collected by the i-th Rp in period n, It represents the RSSI data of the hth Ap at the kth sampling time collected by the i-th Rp in period n. It represents the mean value of the RSSI data of the hth Ap collected by the i-th Rp in period n. It represents the discrete degree value of the RSSI data of the hth Ap collected by the i-th Rp in period n.

[0026] Specifically, the RAC algorithm includes: for one divided sub-region, finding the stability index ST corresponding to all Aps and the information index IH corresponding to all Aps, and then finding the reliability index RAC corresponding to all Aps:

[0027]

[0028] Where: RAC is the reliability index, ST is the stability index, and IH is the information index;

[0029] When it is necessary to divide the k best Ap combinations in the sub-area, k Aps with the largest reliability index RAC are selected from all Aps as the best Ap group.

[0030] Specifically, the calculation formula of the stability index ST is:

[0031] ,

[0032]

[0033] Where: is the stability index ST of the sth Ap in the partitioned sub-region m, is the total amount of Rp, is the loc operator, which means sorting the subsequent array from small to large and selecting the value of the zth element. If z is not an integer, take half of the sum of the elements corresponding to the two closest integer ordinals. X is the Xth sampling period. It is the RSSI data of the sth Ap in the nth sampling period at the i-th Rp in the divided sub-area m.

[0034] Specifically, the information index IH formula is:

[0035]

[0036] Where: represents the information value of the sth Ap in the partitioned sub-region m, is the mean value of the RSSI data of the sth Ap at the kth Rp in the partitioned sub-region m over X sampling periods, is the mean value of the RSSI data of the sth Ap at the lth Rp in the partitioned sub-region m over X sampling periods, is the total amount of Rp.

[0037] The present invention also provides a high-precision positioning system based on the optimal combination of regional Ap, comprising the following modules:

[0038] Maximum partition area establishment module: used to determine the smallest circumscribed rectangular area of ​​the target area and record it as the maximum partition area. There are a total of N Aps that can be detected in the maximum partition area.

[0039] Rp signal data receiving module: used for each Ap to periodically receive the RSSI of all Rps in the offline stage, and pre-process the received RSSI data to form a total offline fingerprint database;

[0040] Sub-fingerprint database establishment module: select the number of Aps contained in each sub-fingerprint database, extract the sub-fingerprint database from the total offline fingerprint database obtained in step (2), and the number of Aps ranges from 3 to N;

[0041] Regional Ap optimal combination module: In the offline stage, the MAGS algorithm is used to divide the largest partition area into regions, and the RAC algorithm is used to determine the combination of several optimal Aps under each partition sub-region, which is recorded as the optimal Ap group, and several partition sub-regions and the optimal Ap combination corresponding to each partition sub-region are obtained.

[0042] Beneficial effects: Compared with the prior art, the significant effects of the present invention are: firstly, the sampled RSSI data is accurately processed by the X-score outlier removal method, which effectively eliminates environmental noise and measurement errors, improves data stability and positioning accuracy, and then optimizes the selection of Ap combinations and regional division strategies through the MAGS algorithm, reducing unnecessary Ap deployment, thereby achieving the purpose of reducing overall energy consumption and costs, and enhancing robustness to environmental changes. This method can intelligently select the optimal Ap combination and dynamically adjust the number of Aps, ensuring the high efficiency of positioning and the universality of application scenarios, while improving positioning accuracy, greatly reducing time, manpower, and material expenditures. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the method of embodiment 1 of the present invention.

[0044] Figure 2 This is a schematic diagram of the maximum partition area in Example 1 of the present invention.

[0045] Figure 3 It is a schematic diagram of the optimal Ap combination for dividing sub-regions in embodiment 1 of the present invention.

[0046] Figure 4 It is a schematic diagram of sub-region segmentation according to Embodiment 1 of the present invention.

[0047] Figure 5 It is a schematic diagram of the stability index, information content index and reliability index of Example 1 of the present invention. DETAILED DESCRIPTION

[0048] A preferred embodiment of the present invention is further described below in conjunction with the accompanying drawings.

[0049] Example 1

[0050] See also Figure 1 As shown, this embodiment provides a high-precision positioning method based on the optimal combination of regions Ap, including the following steps:

[0051] (1) Determine the smallest circumscribed rectangular area of ​​the target area and record it as the maximum partition area. There are a total of N Aps (signal transmitting devices) that can be detected in the maximum partition area.

[0052] In actual scenarios, the target area often appears as an irregular closed area on the plane. The circumscribed minimum rectangular area of ​​the area is the maximum partition area. Each Rp (reference point) is evenly arranged in the maximum partition area. According to the actual funding supply of the project and the actual price of each Rp, in this embodiment, a total of m*n Rps are set in the maximum partition area.

[0053] (2) In the offline phase, each Ap periodically receives the RSSI of all Rps and preprocesses the received RSSI data to form a total offline fingerprint database.

[0054] The present invention adopts the X-score outlier removal method to pre-process the received RSSI data, including the following steps:

[0055] (21) In the offline phase, each Rp samples the RSSI emitted by the N detectable Aps in the largest partition area at a fixed time interval, with m+2 samples as a sampling cycle, and collects data for several cycles, where m is a positive integer.

[0056] (22) For one of the cycles n, first calculate the mean of the RSSI data sampled each time by the ith Rp in cycle n, then use the mean to calculate the discrete value of the RSSI data of the ith Rp in cycle n, and use the mean and discrete value to calculate the X score of each sample of the ith Rp in cycle n. Eliminate the two data with the lowest X score from the m+2 data under each Ap in cycle n, and take the mean of the remaining m data to complete the preprocessing.

[0057] The mean is calculated as follows:

[0058]

[0059] The calculation formula for the discrete degree value is:

[0060]

[0061] Where: It represents the average value of RSSI data collected by the i-th Rp in cycle n on all Aps, Represents the discrete degree value of the RSSI data of the hth Ap collected by the i-th Rp in period n; It represents the RSSI data of the hth Ap at the kth sampling time collected by the i-th Rp in period n. It represents the mean value of RSSI data of h-th Ap collected by i-th Rp in period n; m+2 is the number of sampling times in one period.

[0062] The formula for calculating the X-score is:

[0063]

[0064] Where: is the X score of the RSSI data of the hth Ap at the kth sampling time collected by the i-th Rp in period n, It represents the RSSI data of the hth Ap at the kth sampling time collected by the i-th Rp in period n. It represents the mean value of the RSSI data of the hth Ap collected by the i-th Rp in period n. It represents the discrete degree value of the RSSI data of the hth Ap collected by the i-th Rp in period n.

[0065] (23) Repeat step (22) to obtain the RSSI data of all Rp preprocessed to form the total offline fingerprint database Dataset_X, which is specifically:

[0066]

[0067] in, Represents the RSSI data corresponding to the Rp point in the mth row and nth column after being processed by the X-score outlier removal method.

[0068] When extracting q Ap combinations, since there are N Aps in the valid Ap group in the maximum partition area, according to the permutations and combinations, Ap combinations, for each of which, extract the corresponding sub-fingerprint database from the total offline fingerprint database Dataset_X according to the corresponding Ap combination. For example, when q is 3, the Ap combination can be (2,4,5). Extract the RSSI values ​​of the second, fourth, and fifth Ap under all Rp from the total fingerprint database of X scores to form a new sub-fingerprint database Subdata(2,3,5). Subdata(2,3,5) is specifically:

[0069]

[0070] (4) In the offline stage, the MAGS (minimum Ap group area segmentation) algorithm is used to divide the largest partition area. The RAC algorithm is used to determine the combination of several optimal Aps under each partition sub-area, which is recorded as the optimal Ap group. Several partition sub-areas and the optimal Ap combination corresponding to each partition sub-area are obtained. The sub-fingerprint database corresponding to the optimal Ap combination is extracted, and the positioning of all partition sub-areas is completed by the WKNN positioning algorithm.

[0071] The MAGS algorithm specifically includes the following steps:

[0072] (41) According to the area of ​​the largest partition area, all Rp are evenly distributed and arranged, and the area of ​​the largest single Rp circumscribed circle is set as the minimum positioning partition area S_area, and the largest partition area is divided into 4 triangular partition sub-areas along the diagonal line.

[0073] (42) The RAC algorithm is used to determine the combination of the three best Aps in each sub-region, which is recorded as the best Ap group. The sub-fingerprint database corresponding to the best Ap group is used to locate the sub-region. The sub-region whose positioning accuracy reaches the set threshold Pos_acc will not be further divided. The sub-region and the corresponding best Ap combination are recorded, and the corresponding sub-fingerprint database is extracted. The positioning of the sub-region is completed by the WKNN positioning algorithm. The sub-region whose positioning accuracy does not reach the set threshold Pos_acc is further divided into 4 triangular sub-regions of equal area.

[0074] (43) Repeat step (42) until the positioning accuracy reaches the set threshold Pos_acc, record the sub-region and the corresponding optimal Ap combination, extract the corresponding sub-fingerprint database, and complete the positioning of the sub-region through the WKNN positioning algorithm, or the final sub-region reaches the minimum positioning partition area S_area but the positioning accuracy still does not reach the set threshold Pos_acc.

[0075] (44) For the divided sub-areas that cannot reach the positioning accuracy set threshold Pos_acc through division, increase the number of Aps in the best Ap group, repeat steps (42) and (43) until the positioning Ap group contains N Aps and the final divided sub-area reaches the minimum positioning divided area S_area, record the divided sub-areas whose positioning accuracy reaches the set threshold Pos_acc and the corresponding best Ap combination, extract the corresponding sub-fingerprint database, and complete the positioning of the divided sub-area through the WKNN positioning algorithm.

[0076] (45) After step (44) is completed, the sub-regions whose positioning accuracy still fails to reach the set threshold Pos_acc are marked as areas where positioning cannot be completed. A new Ap is introduced, and steps (42)-(44) are repeated until the positioning accuracy of all sub-regions reaches the set threshold Pos_acc. The corresponding sub-fingerprint database is extracted, and the positioning of all sub-regions is completed through the WKNN positioning algorithm.

[0077] The above RAC algorithm includes: for a certain sub-region, finding the stability index ST corresponding to all Aps and the information index IH corresponding to all Aps in the sub-region, and then finding the reliability index RAC corresponding to all Aps:

[0078]

[0079] Where: RAC is the reliability index, ST is the stability index, and IH is the information index;

[0080] The larger the reliability index RAC value is, the more suitable the Ap is to be selected into the best Ap group in this divided sub-area. When a combination of k best Aps in the divided sub-area is required, k Aps with the largest reliability index RAC are selected from all Aps as the best Ap group.

[0081] The calculation formula of stability index ST is:

[0082] ,

[0083]

[0084] Where: is the stability index ST of the sth Ap in the partitioned sub-region m, is the total amount of Rp, is the loc operator, which means to sort the subsequent array from small to large and select the value of the zth element. If z is not an integer, take half of the sum of the elements corresponding to the two closest integer ordinals. X is the Xth sampling period. It is the RSSI data of the sth Ap in the nth sampling period at the i-th Rp in the divided sub-area m.

[0085] The information index IH formula is:

[0086]

[0087] Where: represents the information value of the sth Ap in the partitioned sub-region m, is the mean value of the RSSI data of the sth Ap at the kth Rp in the partitioned sub-region m over X sampling periods, is the mean value of the RSSI data of the sth Ap at the lth Rp in the partitioned sub-region m over X sampling periods, is the total amount of Rp.

[0088] Through the MAGS algorithm, the minimum required Ap combination can be obtained for each Rp that needs to be located while meeting the positioning accuracy requirements, thereby maximizing the savings in positioning energy consumption. At the same time, the sub-areas that need to add Aps to meet the positioning requirements can be screened out, and new Aps can be established in a targeted manner, thereby saving costs.

[0089] The following is an explanation of this solution through a specific application scenario:

[0090] Please refer to Figure 2 As shown in the figure, the irregular closed area is the target area. The number of valid Aps within the target area is 8. According to the above step (1), the maximum division area of ​​the target area is obtained, which is a square area of ​​50m×50m. Rp is evenly set at intervals of 5m. According to the above step (2), 12 samples are taken as one sampling cycle, and 10 cycles of RSSI data are collected. The sampling interval is 0.5s. The total offline fingerprint database Dataset_X is obtained by the X-score outlier removal method. According to the above step (3), Ap combinations consisting of 3 to 8 Aps are extracted from the total offline fingerprint database Dataset_X. Ap combinations, for each of which Ap combination, a corresponding sub-fingerprint database can be extracted from the total offline fingerprint database Dataset_X according to the corresponding Ap combination.

[0091] According to the method described in step (4) above, the area of ​​the largest single Rp circumscribed circle is set to the minimum positioning division area S_area = 78.54m 2 , the positioning accuracy threshold Pos_acc=0.5m. Please refer to Figure 3 As shown in the figure, according to the MAGS algorithm, the maximum partition area is divided into four identical triangular partition sub-areas along the diagonal line. According to the RAC algorithm, the best Ap combination corresponding to each partition sub-area is obtained. The four-divided area is A 1 , A 2 , A 3 , A 4 , the corresponding optimal Ap combination is (Ap 1 ,Ap 8 ,Ap 7 )、(Ap 1 ,Ap 3 ,Ap 8 )、(Ap 2 ,Ap 4 ,Ap 5 )、(Ap6 ,Ap 4 ,Ap 5 ), where region A 4 If the positioning accuracy threshold Pos_acc cannot be reached under the combination of 3 Aps, the sub-area needs to be divided using the triangle quartering method. Please refer to Figure 4 As shown in the figure, the final regional division result of this example is Figure 4 The shaded area a 31.3 Take as an example, the reliability index RAC in the divided sub-area is calculated, and the data corresponding to each Ap are shown in the following Table 1.

[0092] Table 1

[0093] Ap1 Ap2 Ap3 Ap4 Ap5 Ap6 Ap7 Ap8 ST 10.52 14.72 12.58 17.32 19.11 11.36 13.82 16.33 IH 1.15 1.52 0.78 1.33 1.67 1.72 0.92 1.86 RAC 32.89 60.82 26.67 62.62 86.75 53.11 34.56 82.56

[0094] Please refer to Figure 5 As shown in the figure, the RAC algorithm comprehensively considers the two indicators of stability and information volume, making the data change trend more readable. The reliability indicators are sorted from high to low, and the best Ap combination to achieve positioning accuracy in the divided sub-area is selected as (Ap 6 ,Ap 2 ,Ap 4 ,Ap 8 ,Ap 5 ), thereby extracting the corresponding sub-fingerprint database and completing the positioning of the area through the WKNN positioning algorithm. According to the WKNN positioning algorithm, the positioning accuracy of area a31.3 is 0.36m, which is higher than the positioning accuracy setting threshold of 0.5m.

[0095] Example 2

[0096] This embodiment provides a high-precision positioning system based on the optimal combination of regions Ap corresponding to the high-precision positioning method based on the optimal combination of regions Ap recorded in Embodiment 1, including the following modules:

[0097] Maximum partition area establishment module: used to determine the smallest circumscribed rectangular area of ​​the target area and record it as the maximum partition area. There are a total of N Aps that can be detected in the maximum partition area.

[0098] Rp signal data receiving module: used for each Ap to periodically receive the RSSI of all Rps in the offline stage, and pre-process the received RSSI data to form a total offline fingerprint database;

[0099] Sub-fingerprint database establishment module: select the number of Aps contained in each sub-fingerprint database, extract the sub-fingerprint database from the total offline fingerprint database obtained in step (2), and the number of Aps ranges from 3 to N;

[0100] Regional Ap optimal combination module: In the offline stage, the MAGS algorithm is used to divide the largest partition area into regions, and the RAC algorithm is used to determine the combination of several optimal Aps under each partition sub-region, which is recorded as the optimal Ap group, and several partition sub-regions and the optimal Ap combination corresponding to each partition sub-region are obtained.

Claims

1. A high-precision positioning method based on the optimal combination of regional Ap, characterized in that: The following steps are involved: (1) Determine the smallest circumscribed rectangular area of ​​the target area and record it as the maximum partition area. The number of detectable signal transmitting devices Ap in the maximum partition area is N; (2) In the offline phase, each Ap periodically receives the received signal strength indication RSSI of all reference points Rp, and pre-processes the received RSSI data to form a total offline fingerprint database; The received RSSI data is preprocessed using the X-score outlier removal method: (21) In the offline phase, each Rp samples the RSSI emitted by the N detectable Aps in the maximum partition area at a fixed time interval, with m+2 samples as a sampling cycle, and collects data for several cycles, where m is a positive integer; (22) For one of the cycles n, first calculate the mean of the RSSI data sampled each time by the i-th Rp in cycle n, then use the mean to calculate the discrete value of the RSSI data of the i-th Rp in cycle n, and use the mean and discrete value to calculate the X score of each sample of the i-th Rp in cycle n, remove the two data with the lowest X score from the m+2 data under each Ap in cycle n, and take the mean of the remaining m data to complete the preprocessing; (23) Repeat step (22) to obtain the RSSI data of all Rp preprocessed to form a total offline fingerprint database; The calculation formula of the discrete degree value is: ， Where: Represents the discrete degree value of the RSSI data of the hth Ap collected by the i-th Rp in period n; It represents the RSSI data of the hth Ap at the kth sampling time collected by the i-th Rp in period n. represents the mean value of the RSSI data of the hth Ap collected by the i-th Rp in period n; m+2 is the number of sampling times in one period; The calculation formula of the X score is: ， Where: is the X score of the RSSI data of the hth Ap at the kth sampling time collected by the i-th Rp in period n, It represents the RSSI data of the hth Ap at the kth sampling time collected by the i-th Rp in period n. It represents the mean value of the RSSI data of the hth Ap collected by the i-th Rp in period n. Represents the discrete degree value of the RSSI data of the hth Ap collected by the i-th Rp in period n; (3) Select the number of Aps contained in each sub-fingerprint database, and extract the sub-fingerprint database from the total offline fingerprint database obtained in step (2), and the number of Aps ranges from 3 to N; (4) In the offline stage, the MAGS algorithm is used to divide the largest partition area into regions. The RAC algorithm is used to determine the combination of several optimal Aps under each partition sub-region, which is recorded as the optimal Ap group. Several partition sub-regions and the optimal Ap combination corresponding to each partition sub-region are obtained. The sub-fingerprint database corresponding to the optimal Ap combination is extracted, and the positioning of all partition sub-regions is completed by the Weighted k-NearestNeighbors positioning algorithm. The specific MAGS algorithm is: (41) According to the area of ​​the largest partition area, all Rp are evenly distributed and arranged, and the area of ​​the largest single Rp circumscribed circle is set as the minimum positioning partition area, and the largest partition area is divided into 4 triangular partition sub-areas along the diagonal line; (42) The three best Ap combinations under each sub-region are determined by the RAC algorithm, which are recorded as the best Ap group. The sub-regions are located using the sub-fingerprint database corresponding to the best Ap group. The sub-regions whose positioning accuracy reaches the set threshold are no longer further divided. The sub-regions and the corresponding best Ap combination are recorded, and the corresponding sub-fingerprint database is extracted. The positioning of the sub-regions is completed by the Weighted k-Nearest Neighbors positioning algorithm. The sub-regions whose positioning accuracy does not reach the set threshold are further divided into four triangular sub-regions of equal area. (43) Repeat step (42) until the positioning accuracy reaches the set threshold, record the sub-regions and the corresponding best Ap combination, extract the corresponding sub-fingerprint database, and complete the positioning of the sub-region through the Weighted k-Nearest Neighbors positioning algorithm, or the final sub-region reaches the minimum positioning partition area but still cannot reach the positioning accuracy set threshold; (44) For the divided sub-areas that cannot reach the set threshold of positioning accuracy through division, increase the number of Aps in the best Ap group, repeat steps (42) and (43) until the positioning Ap group contains N Aps and the final divided sub-area reaches the minimum positioning divided area, record the divided sub-areas whose positioning accuracy reaches the set threshold and the corresponding best Ap combination, extract the corresponding sub-fingerprint database, and complete the positioning of the divided sub-area through the Weighted k-Nearest Neighbors positioning algorithm; The RAC algorithm is specifically as follows: for one divided sub-region, the stability index ST corresponding to all Aps and the information index IH corresponding to all Aps are obtained, and then the reliability index RAC corresponding to all Aps is obtained: ， Where: RAC is the reliability index, ST is the stability index, and IH is the information index; When it is necessary to divide the k best Ap combinations in the sub-area, the k Aps with the largest reliability index RAC are selected from all Aps as the best Ap group; The calculation formula of the stability index ST is: ， ， Where: is the stability index ST of the sth Ap in the partitioned sub-region m, is the total amount of Rp, is the loc operator, which means to sort the subsequent array from small to large and select the value of the zth element. If z is not an integer, take half of the sum of the elements corresponding to the two closest integer ordinals. X is the Xth sampling period. RSSI data of the sth Ap in the nth sampling period at the ith Rp in the divided sub-area m; The information index IH formula is: ， Where: represents the information value of the sth Ap in the partitioned sub-region m, is the mean value of the RSSI data of the sth Ap at the kth Rp in the partitioned sub-region m over X sampling periods, is the mean value of the RSSI data of the sth Ap at the lth Rp in the partitioned sub-region m over X sampling periods, is the total amount of Rp.

2. The high-precision positioning method according to claim 1, characterized in that: The following steps are also included: (45) After step (44) is completed, the sub-regions whose positioning accuracy still fails to reach the set threshold are marked as areas where positioning cannot be completed. A new Ap is introduced and steps (42)-(44) are repeated until the positioning accuracy of all sub-regions reaches the set threshold. The corresponding sub-fingerprint database is extracted and the positioning of all sub-regions is completed using the Weighted k-Nearest Neighbors positioning algorithm.

3. A high-precision positioning system constructed based on the high-precision positioning method based on the optimal combination of regional Ap according to any one of claims 1-2, characterized in that: Includes the following modules: Maximum partition area establishment module: used to determine the smallest circumscribed rectangular area of ​​the target area and record it as the maximum partition area. There are a total of N signal transmitting devices Ap that can be detected in the maximum partition area; Rp signal data receiving module: used for periodically receiving the received signal strength indication RSSI of all reference points Rp at each Ap in the offline stage, and preprocessing the received RSSI data to form a total offline fingerprint database; Sub-fingerprint database establishment module: select the number of Aps contained in each sub-fingerprint database, extract the sub-fingerprint database from the total offline fingerprint database obtained in step (2), and the number of Aps ranges from 3 to N; Regional Ap optimal combination module: In the offline stage, the MAGS algorithm is used to divide the largest partition area into regions, and the RAC algorithm is used to determine the combination of several optimal Aps under each partition sub-region, which is recorded as the optimal Ap group. Several partition sub-regions and the optimal Ap combination corresponding to each partition sub-region are obtained, and the sub-fingerprint database corresponding to the optimal Ap combination is extracted. The Weighted k-Nearest Neighbors positioning algorithm is used to complete the positioning of all partition sub-regions.

Images