An RSS Fingerprint Reconstruction Method Applied to Visible Light Positioning

Abstract

The present invention relates to the technical field of indoor visible light positioning, and in particular to an RSS fingerprint reconstruction method applied to visible light positioning, including: constructing a visible light RSS data acquisition method for eliminating the influence of natural light, and using the real-time RSS difference between high level and low level to eliminate the influence of natural light; constructing a visible light RSS fingerprint reconstruction method without coefficient calibration, and using the RSS matrix of the training position to reconstruct the RSS vector of each interpolation position; combining the RSS matrix of the training position with the RSS matrix of the interpolation position to form a dense visible light RSS database, and completing the reconstruction and update of the visible light RSS fingerprint. The technology of the present invention greatly reduces the labor cost and time cost of constructing the RSS database on the premise of achieving a positioning accuracy similar to that of the actual dense RSS database. The proposed RSS fingerprint reconstruction method of the present invention does not require coefficient calibration, is not affected by natural light, and can be well applied to the RSS fingerprint reconstruction of visible light positioning.

Description

Technical Field

[0001] The present invention relates to the technical field of indoor visible light positioning, and specifically to a method for reconstructing RSS fingerprints applied to visible light positioning. Background Art

[0002] With a large number of mobile devices accessing the Internet of Things and their applications in indoor scenarios, the research and development of indoor positioning systems have become increasingly important. As an emerging indoor positioning technology, a visible light positioning system can use LED light sources as the transmitters of positioning signals and PDs as the receivers of positioning signals. It has attracted much attention from researchers due to its low cost, high positioning accuracy, energy conservation, and no electromagnetic radiation.

[0003] Visible light positioning based on RSS fingerprints can achieve centimeter-level positioning accuracy and can be widely applied to the high-precision positioning of indoor mobile robots. However, the density of RSS fingerprints is proportional to the positioning accuracy. The higher the density of RSS fingerprints, the higher the positioning accuracy, but the higher the labor cost and time cost of collecting RSS fingerprints, especially in the case where the RSS database needs to be frequently updated due to environmental changes. And the method of reconstructing RSS data can effectively reduce the labor and time costs related to building the RSS database. Therefore, in view of the above problems, a method for reconstructing RSS fingerprints applied to visible light positioning is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for reconstructing RSS fingerprints applied to visible light positioning to solve the problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] A method for reconstructing RSS fingerprints applied to visible light positioning includes the following steps:

[0007] Step S1: Construct a method for collecting visible light RSS data that eliminates the influence of natural light. By collecting the RSS difference between high and low levels in real time, eliminate the influence of natural light on RSS data;

[0008] Step S2: Construct a method for reconstructing visible light RSS fingerprints without coefficient calibration. Based on the RSS matrix at the training positions, reconstruct the RSS vector at each interpolation position through an interpolation algorithm;

[0009] Step S3: Combine the RSS matrix at the training positions with the RSS matrix at the interpolation positions to form a dense visible light RSS database, and complete the reconstruction and update of RSS fingerprints.

[0010] As a preferred solution, the specific implementation of step S1 includes:

[0011] M LED emission sources use time-division multiple access (TDMA) technology to send pulse-width modulation (PWM) positioning signals. The positioning time slot of each LED includes:

[0012] The first time slot represents the start signal;

[0013] The second time slot differentiates the IDs of different LEDs by frequency;

[0014] The third time slot uses a pulse-width modulation signal with a duty cycle of 50%. The low-level RSS value and high-level RSS value of the m-th LED received at the i-th training position are measured respectively, and the influence of ambient light is eliminated by the difference between the two;

[0015] In the two-stage amplification circuit of the PD receiver, the secondary amplifier adaptively adjusts the amplification factor through a programmable digital potentiometer to ensure that the output voltage falls within the high and low level judgment intervals.

[0016] As a preferred solution, the specific implementation of step S2 includes:

[0017] Step S2-1: Divide the positioning area into a grid structure and construct a three-dimensional coordinate matrix of L training positions;

[0018] Step S2-2: Construct a training position RSS matrix corresponding to the coordinate matrix, where each training position contains an RSS vector for receiving M LEDs;

[0019] Step S2-3: Calculate the Euclidean distances between the training positions and the LED emission sources, and between the interpolation positions and the LED emission sources respectively;

[0020] Step S2-4: Sort the distance vectors between the interpolation position and all training positions in ascending order, and select the indices of the two nearest neighbor training positions;

[0021] Step S2-5: Based on the LOS Lambert channel gain model, calculate the optical power of the training positions and convert it into RSS values through the LED emission power, Lambert order, PD receiving area, filtering gain, condenser lens gain, the Lambert order power of the cosine of the LED radiation angle, the cosine of the PD incident angle, and the square of the distance; the RSS values of the interpolation positions are calculated using the same model;

[0022] When reconstructing the RSS value of the interpolation position, the following operations are performed:

[0023] Based on the RSS values of the two nearest neighbor training positions, weight them according to the logarithmic difference ratio of the distances between the interpolation position and the two training positions;

[0024] Superimpose the RSS compensation value caused by the change in the PD incident angle. This compensation value is determined by the following method:

[0025] Based on the logarithmic difference ratio of the distances between the interpolation position and the two training positions, the logarithms of the cosines of the incident angles at the two training positions are weighted;

[0026] The logarithm of the cosine of the incident angle at the interpolation position itself is superimposed.

[0027] As a preferred solution, in the Lambert channel model, the LED direction angle is vertically downward, and the cosine value of its radiation angle is the height difference between the LED and the receiving position divided by the distance between the two;

[0028] The interpolation position and the training position are at the same horizontal height;

[0029] The incident angle of the PD is measured in real time by an acceleration sensor.

[0030] As a preferred solution, in step S3:

[0031] The dense RSS database generates a (L + N)×M-dimensional dense matrix by merging the L×M-dimensional training position RSS matrix and the N×M-dimensional interpolation position RSS matrix row by row.

[0032] As a preferred solution, the amplification factor adjustment formula of the secondary amplifier is:

[0033] The target voltage minus the ambient light interference compensation value, and then divided by the output voltage of the primary amplifier, and the calculation result is used as the gain of the secondary amplifier.

[0034] It can be seen from the technical solution provided by the present invention above that a method for RSS fingerprint reconstruction applied to visible light positioning provided by the present invention has the beneficial effects that:

[0035] Effectively eliminate natural light interference: By the unique method of collecting the RSS difference between the high level and the low level in real time, the influence of natural light on the RSS data can be offset to a great extent; In actual application scenarios, the intensity of natural light changes constantly and interference is widespread. The method of the present invention ensures the accuracy and stability of the RSS data collected under different natural light conditions, provides a reliable data basis for subsequent positioning calculations, and there is no need to repeatedly collect data, saving time and resources; For example, in an indoor environment with both LED lighting and natural light irradiation, RSS data not affected by natural light can be stably obtained;

[0036] RSS Fingerprint Reconstruction without Coefficient Calibration: Based on the RSS matrix of training positions, an interpolation algorithm is used to reconstruct the RSS vector at each interpolated position; this process avoids complex coefficient calibration steps, reduces the complexity and computational amount of the system, and can efficiently estimate the RSS value at the interpolated position according to the existing RSS information of training positions, improving the efficiency and accuracy of RSS fingerprint reconstruction; taking the positioning area of a large warehouse as an example, with fewer training positions, the RSS vectors at a large number of interpolated positions can be accurately reconstructed using the interpolation algorithm;

[0037] Constructing a Dense RSS Database: The RSS matrix of training positions is combined with the RSS matrix of interpolated positions to form a dense visible light RSS database, greatly enriching the RSS fingerprint information in the positioning area; a denser database can more accurately reflect the signal characteristics at different positions, providing more comprehensive and detailed data support for the visible light positioning system and significantly improving the positioning accuracy of the positioning system; in a complex indoor environment, the realization of centimeter-level positioning depends on this dense and accurate RSS database, enabling more accurate positioning and tracking of personnel or equipment;

[0038] Improving the Performance of the Positioning System: After the method of the present invention is applied to the visible light positioning system, centimeter-level positioning can be achieved in combination with the k-nearest neighbor algorithm, which has great application value in scenarios with extremely high requirements for positioning accuracy, such as medical surgery navigation, equipment positioning in industrial precision manufacturing, etc., effectively promoting the practical application and development of visible light positioning technology in scenarios with high-precision requirements. Brief Description of the Drawings

[0039] Figure 1 It is a schematic diagram of the steps of an RSS fingerprint reconstruction method for visible light positioning according to the present invention. Detailed Embodiment

[0040] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0041] In order to better understand the above technical solutions, the above technical solutions will be described in detail below in combination with the accompanying drawings of the specification and the specific embodiments.

[0042] As Figure 1 shown, an embodiment of the present invention provides an RSS fingerprint reconstruction method for visible light positioning, including the following steps:

[0043] Step S1: Construct a visible light RSS data acquisition method for eliminating the influence of natural light, and eliminate the influence of natural light on RSS data by collecting the RSS difference between high level and low level in real time;

[0044] Step S2: Construct a visible light RSS fingerprint reconstruction method without coefficient calibration. Based on the RSS matrix of the training positions, reconstruct the RSS vector of each interpolation position through an interpolation algorithm;

[0045] Step S3: Combine the RSS matrix of the training positions with the RSS matrix of the interpolation positions to form a dense visible light RSS database, completing the reconstruction and update of the RSS fingerprint.

[0046] In this embodiment, the function of Step S1 is to construct a visible light RSS data acquisition method that can effectively eliminate the influence of natural light, thereby providing an accurate and reliable RSS data basis for subsequent visible light positioning related operations; the specific steps are as follows:

[0047] Step S1-1: LED emission source signal transmission planning:

[0048] There are M LED emission sources in the system, and time division multiple access (TDMA) technology is used to send pulse width modulation (PWM) positioning signals;

[0049] Plan the positioning time slots for each LED:

[0050] The first time slot is used to send a start signal, and the function of this signal is to enable the receiving end to synchronize, clarify the start of a new positioning signal cycle, so as to accurately receive and process the signals of different LEDs subsequently;

[0051] The second time slot distinguishes different LED IDs through different frequencies; when the receiving end receives a specific frequency signal during this time slot, it can identify the corresponding LED identity, which is convenient for subsequent targeted processing of different LED signals;

[0052] The third time slot uses a PWM signal with a duty cycle of 50%. This duty cycle setting makes the signal evenly distributed between high and low levels, which is beneficial for subsequent measurement and calculation;

[0053] Step S1-2: RSS value measurement and elimination of environmental light influence:

[0054] During the third time slot, for the i-th training position, measure the low-level RSS value of receiving the m-th LED respectively and the high-level received signal strength value of receiving the -th LED at the -th training position ;

[0055] Since the intensity of natural light changes at different times and interferes with the received RSS value, by calculating the RSS difference between high and low levels , it can effectively eliminate the influence of natural light on RSS data; since the interference degree of natural light on high and low levels is basically the same, taking the difference can cancel most of the natural light interference;

[0056] Step S1-3: Adjustment of the PD receiver amplification circuit:

[0057] The PD receiver is equipped with a two-stage amplification circuit;

[0058] The second-stage amplifier adaptively adjusts the amplification factor with the help of a programmable digital potentiometer; in the actual environment, the received signal strength is affected by various factors, such as the distance from the LED emission source, environmental occlusion, etc.; the second-stage amplifier needs to automatically adjust the amplification factor according to the actual strength of the received signal through the programmable digital potentiometer to ensure that the output voltage falls within the high and low level judgment interval; only when the output voltage is within this interval can the receiving end accurately identify the high and low level states of the signal, so as to accurately measure and , providing accurate data support for subsequent elimination of the influence of ambient light and the entire visible light RSS fingerprint reconstruction method.

[0059] In this embodiment, the function of step S2 is to construct a visible light RSS fingerprint reconstruction method without coefficient calibration. Based on the RSS matrix of the training positions, the RSS vector of each interpolation position is reconstructed through an interpolation algorithm to obtain denser and more accurate RSS fingerprint information, preparing for the subsequent formation of a dense visible light RSS database; the following are the detailed steps:

[0060] Step S2-1: Divide the positioning area and construct the training position coordinate matrix:

[0061] Divide the positioning area into a grid structure and construct the coordinate matrix of L training positions , where, is the coordinate matrix of the training positions; is the three-dimensional coordinate of the th training position, , ;

[0062] Step S2-2: Construct the RSS matrix corresponding to the coordinate matrix:

[0063] Construct the RSS matrix corresponding to the coordinate matrix , where, is the RSS matrix of the training positions; is the RSS vector of the th training position receiving LEDs, , through this step, the RSS values received from different LEDs at each training position are integrated into a matrix, facilitating subsequent data processing and analysis;

[0064] Step S2-3, Calculate the physical distance:

[0065] Calculate the training position and the LED emission source The physical distance between , and the interpolation position and The emission source The physical distance between , where is the physical distance between the i-th training position and the m-th LED emission source, represents the 2-norm; is the physical distance between the j-th interpolation position and the m-th LED emission source;

[0066] Step S2-4, Select the nearest neighbor training positions:

[0067] For the interpolation position and the physical distance vector of all training positions perform an ascending order sort, and select the indices and of the two nearest neighbor training positions. These two nearest neighbor training positions play a key role in the subsequent interpolation calculation because they are the closest to the interpolation position, and their RSS values have the greatest impact on the RSS value of the interpolation position;

[0068] Step S2-5, Calculate the optical power and RSS value based on the Lambert channel gain model:

[0069] Calculate the optical power and RSS value of the training position: Based on the LOS Lambert channel gain model of visible light communication, calculate the optical power received by the i-th training position from the m-th LED

[0070] , where is the LED emission power; is the Lambert order; is the PD receiving area; is the filtering gain; is the condenser lens gain; is the LED radiation angle between the m-th LED and the i-th training position; is the PD incident angle between the m-th LED and the i-th training position; is the physical distance between the i-th training position and the m-th LED emission source;

[0071] The RSS value of the training location is obtained after taking the logarithm:

[0072] , where is the RSS value of the i-th training location receiving the m-th LED;

[0073] Calculate the RSS value of the interpolation location: The RSS value of the interpolation location satisfies:

[0074] , where is the RSS value of the j-th interpolation location receiving the m-th LED; is the LED radiation angle between the m-th LED and the j-th interpolation location; is the PD incident angle between the m-th LED and the j-th interpolation location; is the physical distance between the j-th interpolation location and the m-th LED emission source;

[0075] Step S2-6, reconstruct the RSS value of the interpolation location:

[0076] Reconstruct the RSS value of the interpolation location through the difference operation of eliminating the model parameters:

[0077] ;

[0078] where represents the RSS compensation value caused by the change of the PD incident angle, and its value is

[0079] ; is the RSS value of the nearest neighbor training location receiving the m-th LED; is the RSS value of the nearest neighbor training location receiving the m-th LED; is the physical distance between the nearest neighbor training location and the m-th LED; is the physical distance between the nearest neighbor training location and the m-th LED, is the physical distance between the interpolation location and the m-th LED. This formula utilizes the RSS values of two nearest neighbor training locations and their distance relationships with the interpolation location, and estimates the RSS value of the interpolation location through linear interpolation, avoiding the complex coefficient calibration process.

[0080] In this embodiment, in the LOS Lambert channel gain model, the direction angle of the LED is vertically downward, that is, , , and The interpolation position and the training position are on the same horizontal plane, that is The incident angle received by the PD , and are measured by the acceleration sensor.

[0081] In this embodiment, the function of step S3 is to merge the RSS matrix of the training position obtained in the previous steps with the RSS matrix of the interpolation position to form a dense visible light RSS database, thereby completing the reconstruction and update of the RSS fingerprint and providing more comprehensive and accurate data support for the subsequent visible light positioning system; the following are the detailed steps:

[0082] Step S3-1: Identify the matrices to be merged:

[0083] Confirm the training position matrix , which records the RSS values received at each training position from different LED emission sources;

[0084] Identify the interpolation position matrix ; this matrix is the RSS value of each interpolation position reconstructed by the interpolation algorithm in step S2;

[0085] Among them, is the RSS matrix of the training position, is the RSS matrix of the interpolation position, is the merged dense RSS database, is the number of training positions, is the number of interpolation positions, is the number of LED emission sources;

[0086] Step S3-2: Perform the matrix merging operation:

[0087] Merge the training position matrix and the interpolation position matrix to generate a new matrix ;

[0088] As a preferred solution, in step S3, the dense RSS database is generated by merging the training position matrix and the interpolation position matrix , generating

[0089] . In this merged matrix, the previous part is the RSS value of the training position, and the latter part is the RSS value of the interpolation position. The two together form a matrix with a larger scale and more dense information;

[0090] Step S3-3: Complete the reconstruction and update of the RSS fingerprint:

[0091] The merged matrix A dense visible light RSS database is formed; due to the addition of interpolation positions, the RSS fingerprint information in the database becomes richer and more comprehensive, completing the reconstruction of the RSS fingerprint; at the same time, the updated database contains RSS information at more positions, can more accurately reflect the signal characteristics within the entire positioning area, and provides a more reliable data basis for subsequent visible light positioning operations based on this database.

[0092] In this embodiment, the secondary amplifier of the PD receiver adjusts the amplification factor through the following feedback mechanism:

[0093] , where is the amplification factor of the secondary amplifier of the PD receiver, is the preset target voltage, is the ambient light interference compensation value, is the output voltage of the primary amplifier.

[0094] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for RSS fingerprint reconstruction applied to visible light positioning, characterized in that: It includes the following steps: Step S1: Construct a visible light RSS data acquisition method to eliminate the influence of natural light. By collecting the RSS difference between high level and low level in real time, the influence of natural light on RSS data is eliminated; Step S2: Construct a visible light RSS fingerprint reconstruction method without coefficient calibration. Based on the RSS matrix of the training positions, the RSS vector of each interpolation position is reconstructed through an interpolation algorithm. The specific implementation includes: Step S2-1: Divide the positioning area into a grid structure and construct a three-dimensional coordinate matrix of L training positions; Step S2-2: Construct a training position RSS matrix corresponding to the coordinate matrix, where each training position contains an RSS vector for receiving M LEDs; Step S2-3: Calculate the Euclidean distances between the training positions and the LED emission sources, and between the interpolation positions and the LED emission sources respectively; Step S2-4: Sort the distance vectors between the interpolation position and all training positions in ascending order, and select the indices of the two nearest neighbor training positions; Step S2-5: Based on the LOS Lambert channel gain model, calculate the optical power of the training positions and convert it into RSS values through the LED emission power, Lambert order, PD receiving area, filtering gain, condenser lens gain, the Lambert order power of the cosine of the LED radiation angle, the cosine of the PD incident angle, and the square of the distance; the RSS values of the interpolation positions are calculated using the same model; When reconstructing the RSS values of the interpolation positions, the following operations are performed: Based on the RSS values of the two nearest neighbor training positions, weight them according to the logarithmic difference ratio of the distances between the interpolation position and the two training positions; Superimpose the RSS compensation value caused by the change of the PD incident angle. This compensation value is determined by the following method: Based on the logarithmic difference ratio of the distances between the interpolation position and the two training positions, weight the logarithms of the cosines of the incident angles of the two training positions; Superimpose the logarithm of the cosine of the incident angle of the interpolation position itself; Step S3: Merge the RSS matrix of the training positions and the RSS matrix of the interpolation positions to form a dense visible light RSS database, and complete the reconstruction and update of the RSS fingerprint.

2. The RSS fingerprint reconstruction method for visible light positioning according to claim 1, wherein: The specific implementation of step S1 includes: M LED emission sources use time division multiple access TDMA technology to send pulse width modulation PWM positioning signals. The positioning time slot of each LED includes: The first time slot represents the start signal; The second time slot distinguishes the IDs of different LEDs by frequency; The third time slot uses a pulse width modulation signal with a duty cycle of 50%. Measure the low level RSS value and high level RSS value of the i-th training position receiving the m-th LED respectively, and eliminate the influence of ambient light through the difference between the two; In the two-stage amplifier circuit of the PD receiver, the secondary amplifier adaptively adjusts the amplification factor through a programmable digital potentiometer to ensure that the output voltage falls within the high and low level judgment range.

3. A visible light positioning RSS fingerprint reconstruction method according to claim 1, wherein: In the Lambert channel model, the LED direction angle is perpendicular downward, and the cosine value of its radiation angle is the height difference between the LED and the receiving position divided by the distance between the two; The interpolation position and the training position are at the same horizontal height; The PD incident angle is measured in real time by an acceleration sensor.

4. A method for reconstructing RSS fingerprints applied to visible light positioning according to claim 1, characterized in that: In the step S3: The dense RSS database generates a (L + N)×M - dimensional dense matrix by combining the L×M - dimensional training location RSS matrix and the N×M - dimensional interpolation location RSS matrix row - by - row.

5. A method for reconstructing RSS fingerprints applied to visible light positioning according to claim 2, characterized in that: The amplification factor adjustment formula of the secondary amplifier is: Subtract the ambient light interference compensation value from the target voltage, and then divide the result by the output voltage of the primary amplifier. The calculated result is used as the gain of the secondary amplifier.

Images