An indoor fusion positioning method based on error correction

Abstract

The present invention discloses an indoor fusion positioning method based on error correction. The method includes: deploying indoor wireless access points and dividing indoor areas; deploying reference points and collecting RSS data in the offline stage; setting calibration reference points by combining multi-point positioning, boundary judgment, and kmeans clustering; in the online stage, obtaining the RSS data of the node to be measured, and performing preliminary positioning on the node to be measured by using multi-point positioning and area judgment; and performing positioning correction on the node to be measured in the online stage by using an algorithm that combines fingerprint positioning and position compensation. The invention improves the overall positioning accuracy by correcting the positioning results in the area close to the indoor edge, reduces the positioning complexity, and improves the positioning efficiency.

Description

Technical Field

[0001] The present invention belongs to the field of indoor positioning, and particularly relates to an indoor fusion positioning method based on error correction Background Art

[0002] As is well known, with the rapid development of mobile Internet-related technologies, wireless positioning technology has quietly integrated into people's daily lives. The growing demand for location-based services on mobile devices has made wireless positioning technology a current research hotspot. Therefore, wireless positioning technology has developed extremely rapidly and has been widely and quickly applied in various fields. The development of the Global Positioning System (GPS), which can provide outdoor positioning, has become mature and is widely used in many fields such as military and transportation. However, the limitations of GPS are also very obvious. Its signal will have serious attenuation when passing through obstacles such as walls into the indoor. Due to the influence of multipath effects and non-line-of-sight propagation, the corresponding relationship between signal characteristics and location becomes more complex. Therefore, it is particularly important to find a positioning method suitable for the indoor environment

[0003] Many indoor positioning studies use the Received Signal Strength Indication (RSSI) of wireless APs as the feature vector of the node to be located, mainly because RSSI is relatively easy to obtain through software. According to whether it is necessary to estimate the distance between the node to be located and the anchor node during the positioning process, the RSSI-based positioning method can be divided into a distance-based method and a distance-free method. The latter is usually called fingerprint positioning. Since it is not necessary to collect RSSI fingerprint information offline, the RSSI-based multi-point positioning method has become a hot spot in indoor positioning research in recent years and is widely used because of its simple algorithm, low computational complexity, and good accuracy after simple processing of RSSI. However, the main problem of RSSI-based indoor positioning is the influence of the environment on RSSI measurement. The measured RSSI changes over time and is usually unreliable

[0004] Current indoor positioning technologies include Bluetooth technology, UWB technology, infrared technology, geomagnetic technology, Zig-Bee technology, etc. However, these technologies all have more or less disadvantages and it is difficult to maintain good accuracy in the edge areas of the indoor environment. Indoor positioning based on multilateration is a relatively common positioning method. Its principle is to determine the attenuation of the signal by receiving the signal strength and then calculate the distance. However, there is a slight difference between the distance determined by this method and the actual distance. Fingerprint positioning for indoor positioning is a method of database comparison and is more suitable for situations where the signal error is large. With the development of communication technology, people's demand for wireless communication speed is increasing. However, most indoor positioning has poor positioning effects in edge areas such as near walls. At the same time, compared with fusion positioning, the accuracy of a single indoor positioning method is always slightly worse.

[0005] In summary, most of the existing indoor positioning technologies have problems such as being unable to correct edge areas such as indoor walls and the performance of their single indoor positioning methods being inferior to that of fusion positioning. Summary of the Invention

[0006] Object of the Invention: The present invention proposes an indoor fusion positioning method based on error correction, which improves the overall positioning accuracy through a fusion positioning algorithm and the correction of edge areas where the positioning effect of ordinary single indoor positioning algorithms is poor in the indoor environment.

[0007] Summary of the Invention: An indoor fusion positioning method based on error correction provided by the present invention specifically includes the following steps:

[0008] Step 1: Deploy indoor wireless access points AP and divide the indoor area into several non-overlapping areas with the same area size. Each area contains and only contains 1 AP.

[0009] Step 2: Select reference points and collect RSS data in the offline stage.

[0010] Step 3: Use a method combining multi-point positioning, boundary judgment, and kmeans clustering to set calibration reference points.

[0011] Step 4: In the online stage, obtain the RSS data of the node to be measured and perform preliminary positioning on the node to be measured using multi-point positioning and area judgment.

[0012] Step 5: Use an algorithm that combines fingerprint positioning and position compensation to correct the preliminary positioning result of the node to be measured and complete the positioning of the node to be measured.

[0013] Further, in Step 1, 4 APs are deployed at the four corners of the square indoor area. Denote the two-dimensional coordinates of the nth AP as (x n , y n), the nth area is Z n , n = 1, 2, 3, 4.

[0014] Furthermore, the implementation process of the second step is as follows:

[0015] Divide the indoor area into M×M equal-sized grids, set the grid points as reference nodes and collect RSS data:

[0016]

[0017] where RSS ij is the RSS data collected at the reference node in the i-th row and j-th column, is the RSS collected at the reference node in the i-th row and j-th column at the nth AP.

[0018] Furthermore, the third step includes the following steps:

[0019] Step 3.1: Estimate the distance from the reference node to each AP:

[0020]

[0021] where represents the distance from the reference node in the i-th row and j-th column to the nth AP, rss 0 represents the average RSS value at a distance of one meter from the AP, and α represents the path loss exponent;

[0022] Step 3.2: Calculate the initial positioning result (x ij , y ij ) of the reference node in the i-th row and j-th column through the multi-point positioning method. The multi-point positioning equation set is as follows:

[0023]

[0024] Step 3.3: Find the reference nodes that are prone to going out of bounds through boundary determination and divide the area prone to going out of bounds. The specific steps are as follows:

[0025] Perform the multi-point positioning in Step 3.2 on each reference node N times, and respectively judge whether the multi-point positioning of each time exceeds the boundary of the indoor area;

[0026] Record the number of times that the multi-point positioning result of the reference node in the i-th row and j-th column exceeds the boundary of the indoor area as Num ij , if then mark this reference node as a node prone to going out of bounds; otherwise, do not mark it; where q is a set threshold;

[0027] Step 3.4: Cluster the nodes prone to going out of bounds in Step 3.3 to set the calibration reference points. The specific steps are as follows:

[0028] Perform kmeans clustering based on two-dimensional coordinates on the nodes prone to going out of bounds, and use the coordinates of the k cluster centers obtained by clustering as the coordinates of the k calibration reference points respectively. Collect RSS data for each calibration reference point:

[0029]

[0030] where RSS u is the RSS data collected at the u-th calibration reference point, u = 1, 2... k, is the RSS value of the u-th calibration reference point at the n-th AP.

[0031] Furthermore, the fourth step includes the following steps:

[0032] Step 4.1: Obtain the RSS data RSS of the node to be measured t :

[0033]

[0034] Step 4.2: Estimate the distance from the node to be measured to each AP;

[0035] Step 4.3: Calculate the initial positioning result of the node to be measured through the multi-point positioning method;

[0036] Step 4.4: Judge whether the initial positioning result of the node to be measured exceeds the boundary of the indoor area. If it does not go out of bounds, output the initial positioning result as the final positioning result; otherwise, execute the fifth step.

[0037] Furthermore, the distance from the node to be measured to the n-th AP is:

[0038] Furthermore, through the following multi-point positioning equation system, calculate the initial positioning result (x t , y t ):

[0039]

[0040] Furthermore, the fifth step includes the following steps:

[0041] Judge the indoor area to which the node to be measured belongs, denoted as Z st ; Take the coordinates of the calibration reference point with the smallest Euclidean distance from the node to be measured as the second positioning result of the node to be measured;

[0042] According to the RSS data at the second positioning result of the node to be measured, judge the indoor area where the second positioning result of the node to be measured is located, denoted as Z nd ; If Z st is the same as Z ndIf they are the same, the second positioning result is output as the final positioning result; otherwise, secondary correction is performed, and the positioning result after secondary correction is output as the final positioning result:

[0043] During secondary correction, the two elements in the RSS data of the node to be measured corresponding to the larger distance from the node to be measured to the AP are removed, and the remaining two elements are retained to obtain RSS'. t :

[0044]

[0045] Synchronize the RSS data of the k calibration reference points to obtain RSS'. u :

[0046]

[0047] Based on RSS' t and RSS' u , the coordinates of the calibration reference point with the smallest Euclidean distance from the node to be measured are used as the final positioning result of the node to be measured.

[0048] Furthermore, the method for determining the indoor area to which a certain point belongs is to use the area where the AP closest to the point is located as the area where the point is located.

[0049] Furthermore, the Euclidean distance between the node to be measured and the u-th calibration reference point is:

[0050]

[0051] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention can improve the overall accuracy of indoor positioning by correcting the positioning results in the edge areas with poor positioning effects near the walls in the room and through the fusion positioning of multi-point positioning and fingerprint positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a flowchart of the present invention;

[0053] Figure 2 is a schematic diagram of the indoor plane;

[0054] Figure 3 is a schematic diagram of the indoor area division;

[0055] Figure 4 is a schematic diagram of the calibration reference point setting;

[0056] Figure 5 is a flowchart of the online positioning algorithm;

[0057] Figure 6 is a graph of the relationship between the number of calibration reference points and the positioning accuracy. Detailed implementation mode

[0058] The technical solution of the invention will be described in detail below with reference to the accompanying drawings:

[0059] The present invention provides an indoor fusion positioning method based on error correction. As Figure 1 shown, this method first deploys indoor wireless access points and divides the indoor area; then performs RSS parameter collection in the offline stage; uses a method combining multi-point positioning, boundary judgment, and kmeans clustering to set calibration reference points; obtains the RSS data of the node to be measured, and finally uses an algorithm combining multi-point positioning, area judgment, fingerprint positioning, and position compensation to accurately position the node to be measured in the online stage.

[0060] The present invention mainly includes three contents: one is to deploy indoor wireless access points and divide the indoor area; the second is to perform RSS parameter collection in the offline stage; use a method combining multi-point positioning, boundary judgment, and kmeans clustering to set calibration reference points; the third is to use an algorithm combining multi-point positioning, area judgment, fingerprint positioning, and position compensation to accurately position the node to be measured in the online stage.

[0061] In one embodiment, 4 APs are deployed at the four corners of a 20m * 20m square room as Figure 2 shown. The two-dimensional coordinates of the nth AP are recorded as (x n , y n ); as Figure 3 shown, the indoor area is divided into 4 areas with the same size and non-overlapping. Each area contains and only contains 1 AP, and the nth area is recorded as Z n .

[0062] In one embodiment, the indoor area of 20m * 20m is divided into equal-sized grids of 0.25m * 0.25m. The grid points are set as reference nodes and RSS data is collected:

[0063]

[0064] where RSS ij is the RSS data collected at the reference node in the i-th row and j-th column, and is the RSS collected at the reference node in the i-th row and j-th column at the nth AP.

[0065] In one embodiment, a method combining multi-point positioning, boundary judgment, and kmeans clustering is used to set calibration reference points, which specifically includes:

[0066] Step 3.1: Estimate the distance from the reference node to each AP:

[0067]

[0068] Among them represents the distance from the reference node at the i-th row and j-th column to the n-th AP, rss 0 represents the average RSS value at a distance of one meter from the AP, and α represents the path loss exponent.

[0069] Step 3.2: Calculate the initial positioning result (x ij , y ij ) of the reference node at the i-th row and j-th column through the multi-point positioning method. The multi-point positioning equation set is as follows:

[0070]

[0071] The solution expression of the above equation set is:

[0072] x = (A T A) -1 A T B

[0073] Among them

[0074]

[0075]

[0076] Step 3.3: Find out the reference nodes that are prone to going out of bounds through boundary determination, and divide the area prone to going out of bounds. The specific steps are as follows:

[0077] Perform multi-point positioning in Step 3.2 on each reference node 1000 times, and then judge whether each reference node goes out of bounds according to whether the positioning result of each time exceeds the boundary of the indoor area;

[0078] Record the number of times that the multi-point positioning result of the reference node at the i-th row and j-th column exceeds the boundary of the indoor area as Num ij , if then mark this reference node as a node prone to going out of bounds; otherwise, do not mark it; where q is a set threshold.

[0079] Step 3.4: Cluster the nodes prone to going out of bounds in Step 3.3 to set calibration reference points. The specific steps are as follows:

[0080] As Figure 4 shown (where Δ represents the reference node), perform kmeans clustering based on two-dimensional coordinates on the nodes prone to going out of bounds, and use the coordinates of the k cluster centers obtained by clustering as the coordinates of the k calibration reference points, and collect RSS data for each calibration reference point:

[0081]

[0082] where RSS u is the RSS data collected at the u-th calibration reference point, u = 1, 2... k, is the RSS of the u-th calibration reference point at the n-th AP.

[0083] In one embodiment, taking the lower left corner of a 20m * 20m square room as shown in Figure 2 as the origin, and denoting the coordinates of its upper right corner as (X r , Y r ), the criterion for determining whether the initial positioning result (x ij , y ij ) of the reference node exceeds the indoor area boundary is:

[0084]

[0085] In one embodiment, as shown in Figure 5 , obtain the RSS data of the node to be measured, and finally adopt an algorithm that combines multi-point positioning, area judgment, fingerprint positioning, and position compensation to perform precise positioning on the node to be measured in the online stage, specifically including:

[0086] Step Four: In the online stage, obtain the RSS data of the node to be measured, and perform preliminary positioning on the node to be measured using multi-point positioning and area judgment:

[0087] Step 4.1: Obtain the RSS data RSS t :

[0088]

[0089] Step 4.2: Estimate the distance from the node to be measured to each AP;

[0090] Step 4.3: Calculate the initial positioning result of the node to be measured through the multi-point positioning method;

[0091] Step 4.4: Determine whether the initial positioning result of the node to be measured exceeds the indoor area boundary. If it does not exceed the boundary, output the initial positioning result as the final positioning result; if it exceeds the boundary, execute Step Five.

[0092] Step Five: Adopt an algorithm that combines fingerprint positioning and position compensation to correct the preliminary positioning result of the node to be measured and complete the positioning of the node to be measured:

[0093] Determine the indoor area to which the node to be measured belongs, denoted as Z st ; Take the coordinates of the calibration reference point with the smallest Euclidean distance from the node to be measured as the second positioning result of the node to be measured; where the Euclidean distance between the node to be measured and the u-th calibration reference point is:

[0094]

[0095] According to the RSS data at the second positioning result of the node to be measured, determine the indoor area where the second positioning result of the node to be measured is located, denoted as Z nd ; If Z st is the same as Z nd , then output the second positioning result as the final positioning result; if Z st is different from Z nd , then perform secondary correction and output the positioning result after secondary correction as the final positioning result.

[0096] During secondary correction, eliminate the two elements with larger distances from the node to be measured to the AP in the RSS data of the node to be measured, and retain the remaining two elements to obtain the new RSS data RSS' t .

[0097]

[0098] Perform synchronous processing on the RSS data of k calibration reference points to obtain RSS' u .

[0099]

[0100] Then, based on RSS' t and RSS' u , use the coordinates of the calibration reference point with the smallest Euclidean distance from the node to be measured as the final positioning result of the node to be measured.

[0101] In one embodiment, the distance from the node to be measured to the nth AP is:[[]]END]]

[0102] In one embodiment, the initial positioning result (x t , y t ) of the node to be measured is calculated through the following multi-point positioning equation system:[[]]END]]

[0103]

[0104] In one embodiment, the method for determining the indoor area to which a certain point belongs is to use the area where the AP closest to the point is located as the area where the point is located.

[0105] Specifically, in one embodiment, as Figure 6 shown, when the number k of calibration reference points takes k = 24, the positioning effect of this algorithm is the best.

[0106] As described above, it is only the specific implementation manner in the present invention, but the protection scope of the present invention is not limited thereto. Any transformation or replacement that can be understood and conceived by those who are familiar with the technology within the technical scope disclosed by the present invention should be covered within the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An indoor fusion positioning method based on error correction, characterized in that, it includes the following steps: Step 1: Deploy indoor wireless access points (APs), and divide the indoor area into several non-overlapping regions with the same area size, and each region contains and only contains 1 AP; Step 2: Select reference points and collect RSS data in the offline stage; Step 3: Adopt a method combining multi-point positioning, boundary judgment and kmeans clustering to set calibration reference points; Step 4: In the online stage, obtain the RSS data of the node t to be measured, and perform preliminary positioning on the node t to be measured by using multi-point positioning and area judgment; Step 5: Adopt an algorithm that combines fingerprint positioning and position compensation to correct the preliminary positioning result of the node t to be measured, and complete the positioning of the node t to be measured; In the first step, 4 APs are deployed at the four corners of the square indoor area. Denote the two-dimensional coordinates of the nth AP as (x n , y n ), and the nth area is Z n , where n = 1, 2, 3, 4; The implementation process of the above Step 2 is as follows: Divide the indoor area into an M×M grid of equal size, set the grid points as reference nodes and collect RSS data: where RSS ij is the RSS data collected at the reference node in the i-th row and j-th column, is the RSS collected at the reference node in the i-th row and j-th column at the n-th AP; The above Step 3 includes the following steps: Step 3.1: Estimate the distance from the reference node to each AP; wherein represents the distance from the reference node at the \(i\)-th row and \(j\)-th column to the \(n\)-th AP, and \(rss\) 0 represents the average RSS value at a distance of one meter from the AP, and \(\alpha\) represents the path loss exponent; Step 3.2: Calculate the initial positioning result (x ij , y ij ) of the reference node in the i-th row and j-th column by the multi-point positioning method. The multi-point positioning equation set is as follows: Step 3.3: Find out the reference nodes that are prone to going out of bounds through boundary determination, and divide the area prone to going out of bounds. The specific steps are as follows: Perform multi-point positioning in Step 3.2 on each reference node N times, and judge whether each multi-point positioning exceeds the boundary of the indoor area; Let the number of times the multi-point positioning result of the reference node in the \(i\)-th row and \(j\)-th column exceeds the boundary of the indoor area be Num ij , if then mark this reference node as an easily out-of-bounds node; otherwise, do not mark it; where \(q\) is a set threshold value; Step 3.4: Cluster the nodes prone to going out of bounds in Step 3.3 to set calibration reference points. The specific steps are as follows: Perform kmeans clustering based on two-dimensional coordinates on the nodes prone to going out of bounds, and use the coordinates of the k cluster centers obtained by clustering as the coordinates of the k calibration reference points respectively, and collect RSS data for each calibration reference point; where RSS u is the RSS data collected at the u-th calibration reference point, where u = 1, 2... k, is the RSS value of the u-th calibration reference point at the n-th AP; The above Step 5 includes the following steps: Determine the indoor area to which the node t to be measured belongs, denoted as Z st ; Use the coordinates of the calibration reference point with the smallest Euclidean distance from the node t to be measured as the second positioning result of the node t to be measured; According to the RSS data at the second positioning result of the node t to be measured, determine the indoor area where the second positioning result of the node t to be measured is located, denoted as Z nd ; If Z st is the same as Z nd , then output the second positioning result as the final positioning result; otherwise, perform secondary correction and output the positioning result after secondary correction as the final positioning result: During the second correction, eliminate the two elements with relatively large distances from the measured node t to the AP in the RSS data of the measured node t, and retain the remaining two elements to obtain the RSS t ': Synchronize the RSS data of k calibration reference points to obtain RSS u ': Based on RSS t ' and RSS u ', the coordinates of the calibration reference point with the smallest Euclidean distance to the node t to be measured are used as the final positioning result of the node t to be measured.

2. The indoor fusion positioning method based on error correction according to claim 1, characterized in that, the above Step 4 includes the following steps: Step 4.1: Obtain the RSS data RSS of the node t to be measured t :[[]] Step 4.2: Estimate the distance from the node t to be measured to each AP; Step 4.3: Calculate the initial positioning result of the node t to be measured by using the multi-point positioning method; Step 4.4: Judge whether the initial positioning result of the node t to be measured exceeds the boundary of the indoor area. If it does not go out of bounds, output the initial positioning result as the final positioning result; otherwise, execute Step 5.

3. The indoor fusion positioning method based on error correction according to claim 2, characterized in that, The distance from the node t to be measured to the nth AP is as follows:

4. The indoor fusion positioning method based on error correction according to claim 2, characterized in that, The initial positioning result (x t , y t ) of the node t to be measured is calculated through the following multi-point positioning equations:

5. The indoor fusion positioning method based on error correction according to claim 1, characterized in that, The method for judging the indoor area to which a certain point belongs is to use the area where the AP closest to the point is located as the area where the point is located.

6. The indoor fusion positioning method based on error correction according to claim 1, characterized in that, The Euclidean distance between the node t to be measured and the u-th calibration reference point is:

Images