Visible light imaging communication train positioning light source adaptive acquisition method

Abstract

The invention discloses a visible light imaging communication train positioning light source adaptive acquisition method, which comprises the following steps: firstly, analyzing the position relationship between an LED positioning light source and a camera, and determining the minimum value and the maximum value of the azimuth angle of the camera to obtain the critical condition of the azimuth angle of the camera; then, fusing an inertial measurement unit (IMU) and an optical flow sensor, constructing a camera azimuth angle ESO model, and estimating an azimuth angle state and external disturbance in real time; then, whether the obtained current azimuth angle of the camera meets critical conditions or not is judged, and an azimuth angle self-adaptive control strategy based on the train speed is designed by combining a Lyapunov theory and a self-adaptive backstepping method; and finally, approximately calculating the derivative of the virtual control quantity by adopting a dynamic surface control (DSC) method, optimizing the response time of camera azimuth angle control, and stably obtaining a positioning light source image. The rotation amount of the camera is adaptively adjusted according to the line curve radius, it is ensured that the camera stably obtains a positioning light source image, and the train positioning precision is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of train positioning, and particularly relates to a method for adaptively obtaining a visible light imaging communication train positioning light source. Background Art

[0002] In recent years, visible light communication (VLC) technology has been widely applied in the fields of indoor positioning, underwater optical communication, medical treatment, Internet of Things, intelligent transportation, etc. The uniformly distributed LED lighting sources in subway tunnels provide a natural communication carrier for realizing train positioning by using VLC, and can be used as the transmitting end of VLC to transmit optical signals containing the position information of the LED light sources. The visible light non-imaging communication technology with a photodetector as the receiving end is vulnerable to tunnel ambient light interference and has a low positioning accuracy; while the visible light imaging communication technology with a complementary metal-oxide semiconductor (CMOS) camera as the receiving end has a stronger anti-interference ability to background light, higher positioning accuracy, and is more suitable for train positioning in tunnel scenarios. When realizing train positioning by using visible light imaging communication, the camera installed at the top of the train head receives the optical signals sent by the LED light sources on the tunnel wall in real time, and the on-vehicle computer identifies the identification (ID) information of the LED light sources by extracting the region of interest and feature information of the LED light source images, and calculates the actual position of the train according to the coordinate positions of the LED light sources. During the train turning process, due to the small curve radius of the subway line, it is difficult for the camera to stably obtain the LED light source images in front of the train running, which may cause the positioning error to diverge rapidly and even interrupt the positioning.

[0003] Currently, there have been a large number of studies on realizing positioning by using VLC, such as:

[0004] (1) Using the modulated LED light source as the VLC signal transmitter to realize positioning, which has achieved good positioning effects in indoor positioning and low-speed vehicle positioning, but does not consider the influence of external interference light sources on the positioning performance.

[0005] (2) The visible light imaging communication positioning method with a camera as the receiving end significantly reduces the light source interference by using the spatial separation advantage of the camera, and the positioning accuracy reaches the centimeter level, but when the positioning light source is insufficient due to the limited camera field of view, the positioning accuracy will decrease significantly.

[0006] (3) The single light source positioning algorithm can calculate the actual position of the receiving end through a single light source when the LED light source image obtained by the camera cannot meet the positioning requirements, but its calculation complexity is high and the positioning accuracy is low.

[0007] (4) Predict the subsequent light source based on the position of the current light source in the image. This method can achieve high positioning accuracy when the spatial layout of the light source remains stable. However, when the spatial layout of the light source changes, the positional relationship between adjacent light sources in the image changes accordingly, and the prediction error will increase significantly.

[0008] (5) For the problem of how to stably obtain target information in the fields of driverless, mobile robots, and navigation, the camera angle can be dynamically adjusted by using an Extended State Observer (ESO) to ensure that the target is always within the field of view, but the system response time is slow.

[0009] Although the existing technologies have verified the feasibility and effectiveness of positioning using VLC by different methods and provided a reference for how to adjust the camera rotation amount to stably obtain target information, less attention has been paid to how to stably obtain the positioning light source for fast-moving objects on different curve radius lines in tunnel scenarios, making it difficult to meet the train positioning requirements for vehicle-to-vehicle communication. Summary of the Invention

[0010] Aiming at the problems existing in the above-mentioned background technology, the purpose of the present invention is to provide a method for adaptively obtaining a positioning light source for visible light imaging communication trains, which adjusts the camera azimuth angle through an Extended State Observer (ESO) to adaptively obtain the positioning light source.

[0011] To achieve the above purpose, the present invention adopts the following technical solutions:

[0012] A method for adaptively obtaining a positioning light source for visible light imaging communication trains includes the following steps:

[0013] S1. Analyze the positional relationship between the LED positioning light source and the camera, and obtain the critical conditions of the camera azimuth angle by determining the minimum and maximum values of the camera azimuth angle;

[0014] S2. Construct a camera azimuth angle ESO model:

[0015]

[0016] Among them, x 2 =β is the drift angle generated by external disturbances, indicating the additional deviation of the camera caused by disturbances; is the azimuth angular velocity, is the azimuth angular acceleration; Δ β is the external disturbance in the drift angle, x 5 =d(t)=f(x 3 , x 4 )+τ, is the unmodeled dynamics, u is the control input, b 1is the model parameter, τ is the external perturbation, and h(t) is the derivative of the composite perturbation;

[0017] Convert the camera azimuth ESO model into a matrix equation form, construct the ESO equation, and estimate the azimuth state in real time according to the ESO equation;

[0018] S3. Use the inertial measurement unit (IMU) to collect the camera azimuth in real time, use the optical flow sensor to monitor the relative displacement of the LED light source image in the camera field of view, calculate the offset angle of the LED light source in the camera field of view, and correct the cumulative error of the camera azimuth collected by the IMU in real time through the offset angle to obtain the current camera azimuth;

[0019] S4. Determine whether the current camera azimuth satisfies the critical condition obtained in step S1:

[0020] If the critical condition is satisfied, further determine whether it is the desired azimuth; if so, directly capture and locate the light source image and determine the positioning light source; if not, calculate the observation error between the current camera azimuth and the azimuth estimated by the ESO equation, and estimate the disturbance information;

[0021] If the critical condition is not satisfied, calculate the critical error between the current camera azimuth and the critical value, and jointly estimate the disturbance information by combining the observation error between the current camera azimuth and the azimuth estimated by the ESO equation;

[0022] S5. According to the estimated disturbance information, construct a Lyapunov function based on the adaptive backstepping method, and minimize the control error through the Lyapunov function;

[0023] Design the train speed gain function:

[0024]

[0025] where K 1 is the reference gain, μ is the adjustment coefficient of the relationship between speed and gain, and v is the real-time speed of the train;

[0026] Adaptive adjust the control gain according to the train speed, gradually compensate the disturbance by using the virtual control quantity, realize the adaptive adjustment of the camera azimuth, and ensure that the adjusted azimuth always satisfies the critical condition; use the dynamic surface control (DSC) method to approximately calculate the derivative of the virtual control quantity, optimize the response time of the camera azimuth control, and stably obtain the positioning light source image.

[0027] Further, in step S1, the method for obtaining the critical condition of the camera azimuth angle is as follows: LED light sources are evenly distributed on the left tunnel wall along the train running direction. The camera azimuth angle is the angle by which the camera deflects relative to both sides of the tunnel wall. The first LED light source in front of the train is used as the LED positioning light source. When the camera deflects from the right tunnel wall to the left, when the upper boundary of the camera's field of view can capture the first LED light source in front of the train and subsequent light sources, the camera azimuth angle reaches the minimum value; when the camera deflects towards the LED light source side and the lower boundary of the camera's field of view can only obtain the first LED light source in front of the train, the camera azimuth angle reaches the maximum value.

[0028] The minimum value ψ of the camera azimuth angle min The calculation formula is:

[0029]

[0030] The maximum value ψ of the camera azimuth angle max The calculation formula is:

[0031]

[0032] Where, F is the camera field of view angle, L is the distance between adjacent LED light sources, and W is the vertical distance between the camera and the LED light source.

[0033] Further, in step S2, the method for constructing the camera azimuth angle ESO model is as follows:

[0034] S21. Taking the camera azimuth angle and the drift angle as control parameters, establish a non-linear state space model of the camera azimuth angle:

[0035]

[0036] S22. Combine the external disturbance and the unmodeled dynamics in the non-linear state space model of the camera azimuth angle into a composite disturbance, and define the state variable x 5 = d(t) = f(x 3 , x 4 ) + τ, thus obtaining the camera azimuth angle ESO model.

[0037] Further, in step S2, the method for constructing the ESO equation is as follows:

[0038] Convert the camera azimuth angle ESO model into the following matrix equation form:

[0039]

[0040] Where, x is the camera azimuth angle state, is the state matrix, c 1 and c 2is the attenuation coefficient of the drift angle; is the control input matrix, y is the measurement output, C = [1 0 0 0] is the output matrix, and φ x is the composite disturbance of the external disturbance and the unmodeled dynamics;

[0041] Construct the ESO equation according to the form of the matrix equation:

[0042]

[0043] where, is the estimated value of the camera azimuth angle, is the ESO gain matrix, and α i > 0 (i = 1, 2, 3, 4) are the ESO gain coefficients, and 0 <ε <1 is the convergence parameter of the ESO.

[0044] Furthermore, in step S3, the method for the IMU to collect the camera azimuth angle in real time is:

[0045] The camera angular velocity measured by the IMU is ω = (ω x , ω y , ω z ), and the angular velocity information is converted into a quaternion q:

[0046]

[0047] where, Δt is the sampling time interval;

[0048] Using the quaternion conversion formula, q is converted into the camera azimuth angle ψ 0 :

[0049]

[0050] where, q ω , q x , q y and q z represent the four components of the quaternion q respectively, q ω is the scalar component, usually related to the rotation angle; q x , q y and q z are all vector components, representing the directions of the rotation axes (x, y, z).

[0051] Furthermore, in step S3, the offset angle Δψ of the LED light source in the camera field of view is:

[0052] Δψ = kl;

[0053] where, l is the relative displacement of the LED light source image in the camera field of view, l = (l x, l y ), l = (l x , l y ), l x and l y are the relative displacements of the LED light source image in the x - direction and y - direction in the camera's field of view respectively, and k is the proportionality coefficient of the optical flow sensor;

[0054] Then the current azimuth angle ψ of the camera is:

[0055] ψ = ψ 0 + Δψ.

[0056] Compared with the disadvantages and deficiencies of the prior art, the present invention has the following beneficial effects:

[0057] (1) The present invention uses the data of the IMU and the optical flow sensor to judge the state of the camera's azimuth angle, compensates for the disturbance error by designing an azimuth angle control strategy, and combines DCS to shorten the response time of the azimuth angle control.

[0058] (2) The present invention adaptively adjusts the rotation amount of the camera according to the radius of the line curve, can ensure that the camera stably acquires the positioning light source image, and thus improves the train positioning accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a flowchart of a method for adaptively acquiring a train positioning light source by visible light imaging communication provided by an embodiment of the present invention;

[0060] Figure 2 is a schematic structural diagram for determining the critical conditions of the camera's azimuth angle provided by an embodiment of the present invention. (a) is for determining the minimum value of the camera's azimuth angle, and (b) is for determining the maximum value of the camera's azimuth angle;

[0061] Figure 3 is the experimental result of the adaptive adjustment of the camera's azimuth angle provided by an embodiment of the present invention; in the figure, (a) represents the adjustment result when the curve radius is 500m; (b) represents the adjustment result when the curve radius is 400m; (c) represents the adjustment result when the curve radius is 300m; (d) represents the adjustment result when the curve radius is 250m;

[0062] Figure 4 is the success rate of acquiring the LED light source under different acquisition methods provided by an embodiment of the present invention;

[0063] Figure 5 is the success rate of acquiring the LED light source at different speeds provided by an embodiment of the present invention;

[0064] Figure 6 is the train positioning result at different speeds provided by an embodiment of the present invention;

[0065] Figure 7 It is the train positioning result provided by the embodiment of the present invention using different light source acquisition methods. Specific embodiments

[0066] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below in conjunction with specific 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.

[0067] 1. During the rapid operation of the train, the train measures its own position, speed and other information in real time, providing support for functions such as train-to-train interval control and overspeed protection in the CBTC system for train-to-train communication. By establishing a visible light imaging communication link between the train and the LED light source, the on-board computer receives and decodes the optical signal sent by the LED light source and independently calculates the train position information. When the train passes through lines with different curve radii, the present invention adaptively adjusts the camera rotation amount to stably acquire the positioning light source and improve the train positioning accuracy. The flowchart of the visible light imaging communication train positioning light source adaptive acquisition method is as Figure 1 shown and will be described in detail below.

[0068] 1. Determine the critical conditions of the camera azimuth angle

[0069] When the train is running on a straight section, the success rate of the camera in acquiring the image of the front LED light source is relatively high; while at a turning point, affected by the curve radius of the subway line, the camera's field of view is limited and the difficulty of acquiring the LED light source image increases. Therefore, the camera rotation amount is adaptively adjusted according to the curve radius of the line to ensure that the acquired LED light source meets the positioning requirements. The camera rotation amount refers to the rotation angle of the camera in three-dimensional space, usually represented by the azimuth angle, pitch angle and roll angle. Since the vertical height between the LED light source and the rail plane remains constant and the train always runs along the rail, the influence of the pitch angle and roll angle on the camera's acquisition of the LED light source image can be ignored. Therefore, adjusting the camera rotation amount mainly focuses on the azimuth angle. When adjusting the camera azimuth angle, two key factors need to be balanced: one is to ensure that there are always enough LED light sources in the camera's field of view to prevent positioning interruption; the other is to take into account the quality of the LED light source image to prevent image distortion and affect the train positioning accuracy. Since the adjustment range of the camera azimuth angle is limited, too large a camera azimuth angle will result in insufficient LED light sources in the field of view, while too small a camera azimuth angle will cause geometric distortion of the image. Therefore, determining the critical conditions of the camera azimuth angle is a prerequisite for reasonably adjusting the camera azimuth angle.

[0070] LED light sources are evenly distributed on the left tunnel wall along the train running direction. The camera azimuth angle is the angle at which the camera deflects relative to both sides of the tunnel wall, and the first LED light source in front of the train is used as the LED positioning light source.

[0071] (1) Determination of the minimum value of the camera azimuth angle

[0072] Refer to Figure 2 (a), when the camera deflects from the right tunnel wall to the left and the upper boundary of the camera's field of view can capture the first LED light source in front of the train and subsequent light sources, the camera azimuth angle reaches the minimum value ψ min :

[0073]

[0074] (2) Determination of the maximum value of the camera azimuth angle

[0075] Refer to Figure 2 (b), when the camera deflects towards the LED light source side and the lower boundary of the camera's field of view can only obtain the first LED light source in front of the train, the camera azimuth angle reaches the maximum value ψ max :

[0076]

[0077] where F is the camera field of view angle, L is the distance between adjacent LED light sources, and W is the vertical distance between the camera and the LED light source;

[0078] If the camera azimuth angle continues to increase, no effective positioning light source can be obtained.

[0079] 2. Acquisition of camera azimuth angle information

[0080] Use an inertial measurement unit (IMU) to collect the camera azimuth angle in real time:

[0081] The angular velocity of the camera measured by the IMU is ω = (ω x , ω y , ω z ), and convert the angular velocity information into a quaternion q:

[0082]

[0083] where Δt is the sampling time interval;

[0084] Use the quaternion conversion formula to convert q into the camera azimuth angle ψ 0 :

[0085]

[0086] where q ω , q x , q y and q z represent the four components of the quaternion q respectively, q ω is the scalar component, usually related to the rotation angle; q x, q y and q z are both vector components, representing the directions of the rotation axes (x, y, z).

[0087] Use an optical flow sensor to monitor the relative displacement of the LED light source image in the camera's field of view, and calculate the offset angle Δψ of the LED light source in the camera's field of view:

[0088] Δψ = kl(5)

[0089] where l is the relative displacement of the LED light source image in the camera's field of view, l = (l x , l y ), l = (l x , l y ), l x and l y are respectively the relative displacements of the LED light source image in the x - direction and y - direction in the camera's field of view, and k is the proportionality coefficient of the optical flow sensor.

[0090] Correct the cumulative error of the camera azimuth angle collected in real - time by the IMU through the offset angle to obtain the current camera azimuth angle ψ:

[0091] ψ = ψ 0 +Δψ(6).

[0092] 3. Camera Azimuth Angle Disturbance Estimation

[0093] The adjustment of the camera azimuth angle has non - linear characteristics and is vulnerable to external disturbances and unmodeled dynamics such as azimuth angle angular velocity and angular acceleration. The ESO is an intelligent observer that can estimate the observation error by estimating the azimuth angle state, thereby estimating the disturbances caused by factors such as external disturbances and unmodeled dynamics.

[0094] Taking the camera azimuth angle and drift angle as control parameters, establish a non - linear state - space model of the camera azimuth angle:

[0095]

[0096] where, x 2 =β is the drift angle generated by external disturbances, representing the additional deviation of the camera caused by disturbances; is the azimuth angle angular velocity, is the azimuth angle angular acceleration; Δ β is the external disturbance in the drift angle, is the unmodeled dynamics, u is the control input, b 1 is the model parameter, and τ is the external disturbance.

[0097] To reduce the estimation error, combine the external disturbances and unmodeled dynamics in the non - linear state - space model of the camera azimuth angle into a composite disturbance, and define the state variable x5 = d(t) = f(x 3 , x 4 ) + τ, thus obtaining the camera azimuth ESO model:

[0098]

[0099] where h(t) is the derivative of the composite disturbance.

[0100] For the convenience of subsequent analysis, the camera azimuth ESO model is transformed into the following matrix equation form:

[0101]

[0102] where x is the camera azimuth state, is the state matrix, c 1 and c 2 are the decay coefficients of the drift angle; is the control input matrix, y is the measurement output, C = [1 0 0 0] is the output matrix, and φ x is the composite disturbance of the external disturbance and the unmodeled dynamics.

[0103] Construct the ESO equation according to the matrix equation form:

[0104]

[0105] where, is the estimated value of the camera azimuth state, is the ESO gain matrix, and α i > 0 (i = 1, 2, 3, 4) are the ESO gain coefficients, and 0 < ε < 1 is the convergence parameter of the ESO.

[0106] Estimate the azimuth state in real time according to the ESO equation.

[0107] When the current azimuth measured by the IMU and the optical flow sensor does not meet the critical condition, estimate the azimuth according to the current camera azimuth and the ESO equation, calculate the observation error e o , and calculate the critical error e c in combination with the critical condition. At this time, the total camera azimuth error e is:

[0108] e = e o + e c (11)

[0109] where:

[0110]

[0111] If the current camera azimuth meets the critical condition, then the total error e = e o .

[0112] By adjusting the parameters of the observer gain matrix H, e is amplified and fed back into the ESO to update the ESO equation:

[0113]

[0114] where, represents the state variables estimated by the ESO.

[0115] The updated camera azimuth state and disturbance information are:

[0116]

[0117] 4. Design of Camera Azimuth Control Strategy

[0118] To compensate for the disturbances caused by external interference and unmodeled dynamics, an adaptive control strategy for the camera azimuth is designed to reduce the azimuth adjustment deviation during the rapid operation of the train. The adaptive backstepping method is a recursive design method using Lyapunov theory, which can adjust the parameters online to achieve disturbance compensation; the Dynamic Surface Control (DSC) method simplifies the high-order differential calculation by introducing a first-order filter, reduces the computational complexity, and shortens the control response time.

[0119] (1) To compensate for the camera drift angle β, the azimuth control error e is defined using the adaptive backstepping method 1 as:

[0120] e 1 = ψ - ψ d + β = x 1 - ψ d + x 2 (16)

[0121] where, ψ d is the desired azimuth;

[0122] To minimize e 1 , a Lyapunov function V 1 is constructed as:

[0123]

[0124] Differentiating the Lyapunov function V 1 yields:

[0125]

[0126] In a straight section, it is relatively easy for the camera to acquire the positioning light source, and there is no need to significantly adjust the azimuth angle. Moreover, at this time, the train is running at a relatively high speed, and a too large adjustment of the azimuth angle may cause the light source image to be blurred. While at a turning point, it is more difficult to acquire the positioning light source, and the train speed is relatively slow, so the adjustment range can be increased. Therefore, a speed gain function K(v) is designed to adaptively adjust the control gain according to the train speed.

[0127] Train speed gain function K(v):

[0128]

[0129] where K is the reference gain, μ is the adjustment coefficient of the relationship between speed and gain, and v is the real-time speed of the train.

[0130] In order to make make e 1 converge, a virtual control quantity q 1 is introduced:

[0131]

[0132] where K 1 (v) is the control gain, is the disturbance estimation of the drift angle:

[0133]

[0134] where α β and σ β are designed positive constants.

[0135] The dynamic surface control (DSC) method is adopted, and a filter is introduced to approximately calculate the derivative of the virtual control quantity:

[0136]

[0137] where γ 1 is the filter time constant, and q 1 ′ is the filtered output of q 1 .

[0138] As can be seen from Equation (7), the control parameter x 3 can directly drive x 1 and x 2 . Let the filtering error E 1 = q 1 - q 1 ′, which is used to compensate for the disturbance caused by the unmodeled dynamic angular velocity . Define the new azimuth control error e 2 as:

[0139] e 2 = x3 -q 1 ' = x 3 -q 1 +E 1 (23)

[0140] Substituting equations (20) and (23) into (18), we get:

[0141]

[0142] To stabilize e 2 , choose the Lyapunov function V 2 as:

[0143]

[0144] Taking the derivative of equation (25), we get:

[0145]

[0146] For V 2 choose the virtual control input q 2 as:

[0147]

[0148] where K 2 (v) is the control gain. By adjusting K 2 (v), ensure that makes e 1 and e 1 converge.

[0149]

[0150] where γ 2 is the filter time constant, and q 2 ' is the filtered output of q 2 .

[0151] Similarly, the control parameter x 4 can directly drive x 3 . Let the filtering error E 2 = q 2 -q 2 ' to compensate for the disturbance caused by the unmodeled dynamic angular acceleration . Define the new azimuth control error e 3 as:

[0152] e 3 = x 4 -q 2 ' = x 4 -q 2 +E 2 (29)

[0153] Substituting Eqs. (27) and (29) into (26), we get:

[0154]

[0155] To stabilize e 3 , select the Lyapunov function:

[0156]

[0157] Taking the derivative of Eq. (31), we have:

[0158]

[0159] By minimizing e 1 , e 2 and e 3 , compensating for the disturbances caused by the drift angle, angular velocity, and angular acceleration, ensuring that the camera azimuth control error is consistent and bounded, and finally obtaining the azimuth control strategy as:

[0160]

[0161] where K 3 (v) is the control gain.

[0162] 5. Obtain the positioning light source image

[0163] Judge whether the current azimuth angle of the camera meets the critical condition obtained in step S1:

[0164] If the critical condition is met, further judge whether it is the desired azimuth angle; if so, directly capture the positioning light source image and determine the positioning light source; if not, calculate the observation error between the current azimuth angle of the camera and the estimated azimuth angle of the ESO equation, and estimate the disturbance information;

[0165] If the critical condition is not met, calculate the critical error between the current azimuth angle of the camera and the critical value, and jointly estimate the disturbance information with the observation error;

[0166] According to the estimated disturbance information, compensate for the disturbance through the camera azimuth control strategy, and adaptively adjust the camera azimuth angle. Specifically: adaptively adjust the control gain according to the train speed, gradually compensate for the disturbance using the virtual control quantity, realize the adaptive adjustment of the camera azimuth angle, and ensure that the adjusted azimuth angle always meets the critical condition; adopt the dynamic surface control (DSC) method to approximately calculate the derivative of the virtual control quantity, optimize the response time of the camera azimuth control, and stably obtain the positioning light source image.

[0167] Due to the specific spatial layout of LED light sources in the tunnel and the high-precision requirements of the CBTC system for train positioning in vehicle-to-vehicle communication, a dual-LED light source positioning algorithm with high reliability and low computational complexity is used to calculate the train position information and obtain the actual position of the train.

[0168] II. Simulation Experiments and Result Analysis

[0169] Combined with the line data and equipment information of a certain subway Line 1, a train positioning experimental scenario is constructed to simulate the train running in the subway tunnel, verify the performance of adaptively adjusting the camera azimuth angle using ESO under different curve radii, analyze the influence of the light source acquisition method and the train running speed on the success rate of positioning light source acquisition, and comparatively analyze the positioning results of the train under different running speeds and when using different methods to adjust the camera azimuth angle.

[0170] 1. Simulation Experiment Design

[0171] The curve radius of the subway line is closely related to the train running speed. When the train speed is lower than 80 km / h, the minimum curve radius should generally not be less than 300 m and should not be less than 250 m in difficult cases; if the train speed exceeds 80 km / h, the minimum curve radius should generally not be less than 500 m and should not be less than 400 m in difficult cases. Curve radii of 500 m, 400 m, 300 m, and 250 m are selected. Taking the distance between two LED light sources as a positioning unit, an experimental platform of 20 m × 2.0 m × 1.5 m is built. The vertical distance and horizontal distance between the CMOS camera and the LED light source are 1.5 m and 2.0 m respectively.

[0172] Let the initial position of the camera be the coordinate origin, the side of the LED light source be the X-axis, the running direction of the experimental trolley be the Y-axis, and the direction perpendicular to the ground be the Z-axis. Along the running direction of the experimental trolley at intervals of 0.5 m, 20 test points are set between (-2.0, 0, -1.5) and (-2.0, 10, -1.5), and 10 tests are carried out respectively. When the experimental trolley runs at 0 km / h and 20 km / h, the actual images of the LED light sources captured by the camera are recorded in sequence; according to the image shift amount at different train running speeds, the motion-blurred images of the LED light sources when the train runs at 40, 60, 80, and 100 km / h are simulated using MATLAB software. The simulation parameters are shown in Table 1:

[0173] Table 1 Experimental Parameters

[0174]

[0175] 2. Analysis of Experimental Results

[0176] (1) Analysis of the Performance of Adaptive Adjustment of the Camera Azimuth Angle

[0177] To verify the performance of the adaptive adjustment of the camera azimuth angle, when the curve radius changes, the proposed method is used to adjust the camera azimuth angle. During the period of t = 0 to 100 s, the train runs on a straight line section; at t = 100 s, the curve radius of the line changes. When the curve radii are 500 m, 400 m, 300 m, and 250 m, the maximum deviations between the adjusted camera azimuth angle ψ and the expected value ψ d are 0.26°, 0.28°, 0.31°, and 0.34° respectively; as the curve radius of the line becomes smaller, the error of the azimuth angle adjustment shows an increasing trend, but during the whole adjustment process, there is no obvious overshoot or lag phenomenon, as Figure 3 shown.

[0178] (2) Influence of camera azimuth angle adjustment on obtaining the positioning light source

[0179] The methods of using a fixed azimuth angle and an adaptive adjusted azimuth angle are adopted to obtain the positioning light source under different curve radii, and the influence of the camera azimuth angle on obtaining the positioning light source is analyzed. The experimental results are as Figure 4 shown, indicating that the success rates of both methods on the straight line section (i.e., the curve radius is ∞) are close to 99.06%. When the curve radii of the line are 500 m, 400 m, 300 m, and 250 m, the success rates of the fixed azimuth angle are 79.98%, 71.89%, 62.40%, and 57.01% respectively, and the success rates of the adaptive adjusted azimuth angle are 99.87%, 99.01%, 95.89%, and 93.95% respectively.

[0180] (3) Influence of train running speed on obtaining the positioning light source

[0181] When the train runs at different speeds, the proposed method is used to adaptively adjust the camera azimuth angle to obtain the positioning light source, and the influence of the train running speed on obtaining the positioning light source is analyzed. When the train passes through the line with a curve radius of 250 m at speeds of 20 km / h, 40 km / h, 60 km / h, 80 km / h, and 100 km / h, the success rates of obtaining the positioning light source are 92.29%, 91.84%, 91.63%, 91.48%, and 91.39% respectively, as Figure 5 shown. The faster the train runs, the lower the success rate of the camera in obtaining the LED light source, but when the train runs at different speeds, the success rates of obtaining the positioning light source are all higher than 91%.

[0182] (4) Analysis of train positioning results at different running speeds

[0183] To analyze the influence of train running speed on the positioning result, the proposed method was used to verify the train positioning accuracy at different running speeds. When the train speeds were 20 km / h, 40 km / h, 60 km / h, 80 km / h, and 100 km / h respectively, the maximum positioning errors were 20.87 cm, 22.35 cm, 24.97 cm, 27.84 cm, and 30.04 cm respectively, and the average positioning errors were 10.59 cm, 11.45 cm, 12.41 cm, 13.69 cm, and 15.18 cm respectively, as Figure 6 shown. The experimental results show that the proposed method can achieve good positioning results at different train running speeds and meet the requirements of the IEEE 1474.1-2025 standard for train positioning accuracy.

[0184] (5) Comparison of train positioning results of different methods

[0185] To verify the positioning performance of the proposed method under different line curve radii, when the train was running at 100 km / h, the train positioning results without using ESO, using the traditional ESO, and the proposed method were compared and analyzed. The maximum positioning errors without using ESO and using the traditional ESO under different line curve radii were 25.78, 29.56, 32.28, 35.15, 39.66 cm and 24.99, 27.56, 30.28, 33.15, 38.26 cm respectively, while the maximum positioning error of the method of the present invention was 23.36, 25.69, 27.86, 29.69, and 30.04 cm. Compared with the methods without using ESO and the traditional ESO, the positioning accuracy of the positioning method of the present invention was improved by 24.25% and 21.48% respectively when the curve radius was the smallest, as Figure 7 shown.

[0186] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for adaptively acquiring a light source for train positioning in visible light imaging communication, characterized in that: The steps include: S1, analyzing the positional relationship between the LED positioning light source and the camera, and obtaining the critical condition of the camera azimuth angle by determining the minimum and maximum values ​​of the camera azimuth angle; S2. Construct the camera azimuth ESO model: Among them, x2 = β is the drift angle caused by external disturbance, which represents the additional deviation of the camera caused by the disturbance; is the azimuth angular velocity, is the azimuth acceleration; Δ β is the external disturbance in the drift angle, x5=d(t)=f(x3,x4)+τ, is the unmodeled dynamics, u is the control input, b1 is the model parameter, τ is the external disturbance, and h(t) is the derivative of the composite disturbance; The camera azimuth ESO model is converted into a matrix equation form, an ESO equation is constructed, and the azimuth state is estimated in real time according to the ESO equation; S3, using the inertial measurement unit to collect the camera azimuth in real time, using the optical flow sensor to monitor the relative displacement of the LED light source image in the camera field of view, calculating the offset angle of the LED light source in the camera field of view, and correcting the cumulative error of the camera azimuth collected by the IMU in real time by the offset angle to obtain the current azimuth of the camera; S4, determining whether the current azimuth angle of the camera satisfies the critical condition obtained in step S1: If the critical condition is met, it is further determined whether it is the expected azimuth angle; if so, the positioning light source image is directly captured and the positioning light source is determined; if not, the observation error between the current azimuth angle of the camera and the azimuth angle estimated by the ESO equation is calculated to estimate the disturbance information; If the critical condition is not met, the critical error between the current azimuth angle of the camera and the critical value is calculated, and the disturbance information is estimated by combining the observation error between the current azimuth angle of the camera and the azimuth angle estimated by the ESO equation; S5. Constructing a Lyapunov function based on the adaptive backstepping method according to the estimated disturbance information, and minimizing the control error through the Lyapunov function; Design the train speed gain function: Among them, K1 is the reference gain, μ is the adjustment coefficient of the relationship between speed and gain, and v is the real-time speed of the train; The control gain is adaptively adjusted according to the train speed, and the virtual control amount is used to gradually compensate for the disturbance to achieve adaptive adjustment of the camera azimuth angle, ensuring that the adjusted azimuth angle always meets the critical conditions. The dynamic surface control method is used to approximate the derivative of the virtual control amount, optimize the response time of the camera azimuth angle control, and stably obtain the positioning light source image.

2. The method for adaptively acquiring a train positioning light source for visible light imaging communication according to claim 1, characterized in that: In step S1, the critical condition of the camera azimuth angle is obtained by evenly distributing LED light sources on the left tunnel wall along the running direction of the train, the camera azimuth angle is the angle at which the camera is deflected relative to both sides of the tunnel wall, and the first LED light source in front of the train is used as the LED positioning light source. When the camera deflects from the right tunnel wall to the left, the upper boundary of the camera field of view angle can capture the first LED light source and subsequent light sources in front of the train, and the camera azimuth angle reaches the minimum value; When the camera deflects toward the LED light source, the camera azimuth angle reaches its maximum value when the lower boundary of the camera field of view can only obtain the first LED light source in front of the train; The minimum value of the camera azimuth angle ψ min The calculation formula is: The maximum value of the camera azimuth ψ max The calculation formula is: Wherein, F is the field of view of the camera, L is the distance between adjacent LED light sources, and W is the vertical distance between the camera and the LED light source.

3. The method for adaptively acquiring a train positioning light source for visible light imaging communication according to claim 1, characterized in that: In step S2, the camera azimuth ESO model is constructed by: S21. Using the camera azimuth angle and drift angle as control parameters, a nonlinear state space model of the camera azimuth angle is established: S22. Combine the external disturbance and unmodeled dynamics in the camera azimuth nonlinear state space model into a composite disturbance, and define the state variable x5=d(t)=f(x3,x4)+τ to obtain the camera azimuth ESO model.

4. The method for adaptively acquiring a train positioning light source for visible light imaging communication according to claim 3, characterized in that: In step S2, the ESO equation is constructed as follows: The camera azimuth ESO model is converted into the following matrix equation form: Among them, x is the camera azimuth state, is the state matrix, c1 and c2 are the attenuation coefficients of the drift angle; is the control input matrix, y is the measured output, C = [1 0 0 0] is the output matrix, φ x is a composite perturbation of external perturbations and unmodeled dynamics; The ESO equation is constructed according to the matrix equation form: in, is the estimated value of the camera azimuth state, is the ESO gain matrix, α i >0(i=1,2,3,4) is the ESO gain coefficient, and 0<ε<1 is the ESO convergence parameter.

5. The method for adaptively acquiring a train positioning light source for visible light imaging communication according to claim 1, characterized in that: In step S3, the method for the inertial measurement unit to collect the camera azimuth angle in real time is: The camera angular velocity measured by the inertial measurement unit is ω=(ω x ,ω y ,ω z ), convert the angular velocity information into quaternion q: in, Δt is the sampling time interval; Using the quaternion conversion formula, q is converted into the camera azimuth angle ψ0: Among them, q ω ,q x ,q y and q z Represents the four components of the quaternion q, q ω is a scalar component, usually related to the rotation angle; q x ,q y and q z They are all vector components, representing the direction of the rotation axis (x, y, z).

6. The method for adaptively acquiring a train positioning light source for visible light imaging communication according to claim 5, characterized in that: In step S3, the offset angle Δψ of the LED light source in the camera field of view is: Δψ=kl; Where l is the relative displacement of the LED light source image in the camera field of view, l = (l x , l y ), l x and l y are the relative displacements of the LED light source image in the x-direction and y-direction in the camera field of view, respectively, and k is the scale factor of the optical flow sensor; Then the current azimuth angle ψ of the camera is: ψ=ψ0+Δψ。

Images