CN-122002529-A - SVR indoor positioning method based on DBSCAN feature selection

CN122002529ACN 122002529 ACN122002529 ACN 122002529ACN-122002529-A

Abstract

The SVR indoor positioning method based on DBSCAN feature selection comprises the following steps of S1, obtaining RSSI data of signals received by a mobile terminal and a position fingerprint reference point arranged in a target indoor scene by the aid of the position fingerprint reference point to form position fingerprint data, screening target position fingerprint reference points according to the position fingerprint data, S2, processing the RSSI data of the target fingerprint reference point by means of a DBSCAN cluster analysis algorithm, S3, constructing an SVR indoor positioning prediction model, inputting the data processed in the S2 into the SVR indoor positioning prediction model for offline training, S4, obtaining RSSI data of a mobile terminal to be tested between the target indoor scene and the access point, and inputting the RSSI data into the trained SVR indoor positioning prediction model to obtain the position of the mobile terminal to be tested in the target indoor scene.

Inventors

ZHANG SHUAI
CHEN JIANGUANG
HU CHUAN
LI HUARONG

Assignees

浙江德清知路导航研究院有限公司
重庆交通大学

Dates

Publication Date: 20260508
Application Date: 20251111

Claims (5)

1. A SVR indoor positioning method based on DBSCAN feature selection is characterized by comprising the following steps: S1, acquiring RSSI data of signals received by a mobile terminal and a position fingerprint reference point arranged in a target indoor scene and an access point to form position fingerprint data, and screening the target position fingerprint reference point according to the position fingerprint data; S2, processing RSSI data of a target fingerprint reference point by adopting a DBSCAN cluster analysis algorithm; S3, constructing an SVR indoor positioning prediction model, and inputting the data processed in the step S2 into the SVR indoor positioning prediction model for offline training; S4, acquiring RSSI data between the mobile terminal to be tested and the access point in the target indoor scene, and inputting the RSSI data into the SVR indoor positioning prediction model after training to obtain the position of the mobile terminal to be tested in the target indoor scene.
2. The SVR indoor positioning method based on DBSCAN feature selection of claim 1, wherein the step S1 of obtaining the RSSI of the signal received by the access point by the mobile terminal and the position fingerprint reference point arranged in the target indoor scene specifically comprises the following steps: Determining signal received power of a mobile terminal at any point within a target indoor scene : ; Wherein: indicating that the mobile terminal is at distance At which reference signal received power is to be determined, Representing mobile terminal distance from access point distance The signal received power at which the signal is received, Which represents the path loss index (pathloss) and, For the reference propagation distance to be chosen, Representing the random fading strength of the signal; Based on signal received power Determining RSSI data: 。
3. The SVR indoor positioning method based on DBSCAN feature selection according to claim 2, wherein in the step S1, the screening of the target position fingerprint reference point according to the position fingerprint data specifically comprises the following steps: constructing a data set R of RSSI values received by each access point by each position fingerprint reference point: ; Wherein: indicating that an ith position fingerprint reference point in a target indoor scene receives an RSSI value of a jth access point; Constructing fingerprint data sets of all reference points received by mobile terminal in real time : ; Wherein: Representing the total number of location fingerprint reference points in the target indoor scene, Representing the total number of access points in the target indoor scene, Indicating that the mobile terminal in the target indoor scene receives the RSSI value of the jth access point at the ith position fingerprint reference point, , ; Calculating average value of RSSI values of jth access point received by mobile terminal at fingerprint reference points of all positions : ; Calculating variance : ; Calculating Euclidean distance between RSSI characteristic value of mobile terminal and characteristic value in data set R : ; Distance of European style Normalization processing: ; determining a ratio of a frequency of signals received in a single period from a jth access point to a total number of signal acquisitions , Determining measurement accuracy for a jth access point : ; Wherein: Setting a constant coefficient; Calculating Euclidean distance between actually measured RSSI average value of mobile terminal and corresponding RSSI value of access point in fingerprint database : ; Distance of European style Normalization processing: ; Calculating decision probability of position fingerprint reference point : ; Decision probability of fingerprint reference points of all positions Descending order and screening out decision probability And taking the position fingerprint reference point which is larger than the set value as a target fingerprint reference point.
4. The SVR indoor positioning method based on DBSCAN feature selection according to claim 1, wherein the RSSI data of the target fingerprint reference point is processed by adopting a DBSCAN cluster analysis algorithm: initializing DBSCAN cluster analysis algorithm parameters including minimum points and neighborhood sizes; Wherein the minimum point number is represented by MinPts, and the neighborhood size is represented by MinPts A representation; Determining a clustering center: wherein p and q each represent feature data to be judged, Representing the Euclidean distance between two feature data; and calculating the davison fort Ding Zhishu of the DBSCAN cluster analysis algorithm under different minimum points and neighborhood sizes, and selecting the minimum points and the neighborhood sizes corresponding to the minimum value of the davison fort index as optimal parameters.
5. The SVR indoor positioning method based on DBSCAN feature selection of claim 1 is characterized in that the construction of the SVR indoor positioning prediction model specifically comprises the following steps: determining an input sample of the SVR indoor positioning model, wherein the input sample comprises an RSSI value of a position fingerprint reference point and a position coordinate corresponding to the RSSI value; The optimization model of the SVR indoor positioning model is as follows: ; Wherein: representing a relaxation variable, C representing a penalty coefficient, and m representing the number of support vector machines; The kernel function of the optimization model of the SVR indoor positioning model is as follows: ; Wherein: Representing the center of the kernel function, Representing the kernel parameters; Wherein, the SVR indoor positioning prediction model firstly adopts GS algorithm to punishment coefficient C and nuclear parameter Performing preliminary optimization, and then determining a final penalty coefficient C and a core parameter by adopting a particle swarm optimization algorithm 。

Description

SVR indoor positioning method based on DBSCAN feature selection Technical neighborhood The invention relates to an indoor positioning method, in particular to an SVR indoor positioning method based on DBSCAN feature selection. Background The outdoor position service based on the GNSS can better meet the application requirements of people, however, the indoor satellite signals are interfered by shielding, multipath effect and the like, so that the GNSS can not realize accurate positioning indoors, and reliable position service is provided. Indoor positioning services are strongly demanded, and have great demands in the fields of retail industry, catering industry, logistics industry, manufacturing industry, chemical industry, electric power industry, medical industry and the like. In the prior art, due to the application of the Wi-Fi technology and the indoor positioning technology based on Wi-Fi signals, the Wi-Fi signals in public places are covered in a large area, the distribution range is wide, the distribution density is high, and the indoor positioning method based on Wi-Fi is widely focused. Meanwhile, the Wi-Fi-based indoor positioning method has the characteristics of no need of deploying additional equipment, low positioning cost, strong applicability, no influence of light and the like [35], and is more beneficial to popularization and promotion. The fingerprint matching positioning method based on RSSI in the indoor positioning technology of Wi-Fi signals is one of the common Wi-Fi indoor positioning methods because time synchronization is not needed and positioning can be realized under the non-line-of-sight condition, but the conventional fingerprint matching positioning method based on RSSI has the following defects that the conventional RSSI data has low parameter optimization efficiency and low accuracy when the position prediction model is trained, and more generally, the RSSI data used for training the prediction model is often matched with the actual position to a lower degree, so that the accuracy of a final prediction result is low. Therefore, in order to solve the above-mentioned technical problems, a new technical means is needed. Disclosure of Invention In view of the above, the present invention aims to provide a SVR indoor positioning method based on DBSCAN feature selection, which can accurately match RSSI data for indoor positioning, select reasonable RSSI data for training an indoor positioning prediction model, thereby ensuring the accuracy of a parameter optimization result of the indoor positioning prediction model, and further ensuring the accuracy of a final indoor positioning result. The invention provides a SVR indoor positioning method based on DBSCAN feature selection, which comprises the following steps: S1, acquiring RSSI data of signals received by a mobile terminal and a position fingerprint reference point arranged in a target indoor scene and an access point to form position fingerprint data, and screening the target position fingerprint reference point according to the position fingerprint data; S2, processing RSSI data of a target fingerprint reference point by adopting a DBSCAN cluster analysis algorithm; S3, constructing an SVR indoor positioning prediction model, and inputting the data processed in the step S2 into the SVR indoor positioning prediction model for offline training; S4, acquiring RSSI data between the mobile terminal to be tested and the access point in the target indoor scene, and inputting the RSSI data into the SVR indoor positioning prediction model after training to obtain the position of the mobile terminal to be tested in the target indoor scene. Further, in step S1, acquiring the RSSI of the signal received by the access point by the mobile terminal and the location fingerprint reference point disposed in the target indoor scene specifically includes: Determining signal received power of a mobile terminal at any point within a target indoor scene : ; Wherein: indicating that the mobile terminal is at distance At which reference signal received power is to be determined,Representing mobile terminal distance from access point distanceThe signal received power at which the signal is received,Which represents the path loss index (pathloss) and,For the reference propagation distance to be chosen,Representing the random fading strength of the signal; Based on signal received power Determining RSSI data: 。 further, in the step S1, the screening the target position fingerprint reference point according to the position fingerprint data specifically includes: constructing a data set R of RSSI values received by each access point by each position fingerprint reference point: ; Wherein: indicating that an ith position fingerprint reference point in a target indoor scene receives an RSSI value of a jth access point; Constructing fingerprint data sets of all reference points received by mobile terminal in real time : ; Wherein: Representing the tota