Mobile phone information fusion-based method and system for measuring down tilt angle of plate antenna

Abstract

The application provides a kind of plate-shaped antenna downtilt angle measurement method and system based on mobile phone information fusion, comprising: basic measurement module: through the antenna image of mobile phone shooting, eliminate the parameters related to distance in camera imaging principle, solve the downtilt angle of antenna in combination with IMU data and antenna specific model information;Virtual pole module: through the image of the checkerboard of mobile phone shooting and the direction of gravity, solve the gravity vector on the image, obtain the downtilt angle after error correction;Image aspect ratio solving module: when the antenna model is unknown, the aspect ratio of antenna on the image is obtained from the front of the antenna, the aspect ratio of antenna on the image is used to replace the size ratio of real antenna, and the aspect ratio of antenna is continuously iterated to solve the downtilt angle of antenna.The downtilt angle measured by the algorithm is within two degrees through the test on the actual shooting data set, compared with the traditional cumbersome and dangerous measurement method, the accuracy is also guaranteed, and the requirements in actual network optimization and other aspects are met.

Description

Technical Field

[0001] This invention relates to the field of wireless measurement technology, specifically to a method and system for measuring the downtilt angle of a plate antenna based on mobile phone information fusion, and more specifically to a method and system for measuring the downtilt angle of a plate antenna based on mobile phone multi-sensor information fusion. Background Technology

[0002] In the field of communications, the precise downtilt angle of an antenna is crucial data. The downtilt angle determines the coverage range of the antenna signal, and as fundamental data for daily network optimization, its accuracy directly impacts the quality and efficiency of subsequent network optimization. Convenient and accurate measurement of the antenna downtilt angle will greatly facilitate practical network optimization. However, existing methods for measuring antenna downtilt angle have some drawbacks: the most common method, using physical equipment such as inclinometers and total stations, requires climbing antenna poles or carrying large equipment, which is extremely dangerous and inconvenient. Given the mature imaging principles of mobile phones and the prevalence of high-zoom phones capable of capturing sufficiently clear images from a distance, inferring the antenna downtilt angle by taking a photo of the antenna with a mobile phone and combining the pixel values ​​with the phone's IMU data becomes a feasible approach.

[0003] Patent document CN110186430A (application number: CN201910497585.5) discloses a method for measuring the mechanical downtilt angle of a plate antenna. This method integrates a sensor for measuring the mechanical downtilt angle of the plate antenna inside the plate antenna, thereby achieving automatic measurement of the mechanical downtilt angle. The mechanical downtilt angle is defined as the angle β between the back plate wall of the plate antenna and the vertical direction at the mounting location after the plate antenna is installed and fixed. By integrating the measurement unit capable of measuring the mechanical downtilt angle of the plate antenna as a functional unit within the plate antenna, the automatic measurement function of the plate antenna's own mechanical downtilt angle is achieved. However, this invention cannot calculate the downtilt angle based on a mobile phone photo, resulting in insufficient portability. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for measuring the downtilt angle of a plate antenna based on mobile phone information fusion.

[0005] A downtilt angle measurement system for a plate antenna based on mobile phone information fusion according to the present invention includes:

[0006] Basic Measurement Module: By capturing antenna images with a mobile phone, distance-related parameters in the camera imaging principle are eliminated. The antenna's bottom and side vectors are marked on the image. The antenna's downtilt angle is calculated by combining IMU data and specific antenna model information.

[0007] Virtual support pole module: By taking an image of a chessboard grid aligned with the direction of gravity using a mobile phone, the gravity vector on the image is calculated, and the downward tilt angle after error correction is obtained by replacing the support pole vector with the gravity direction corresponding to the gravity vector in the image.

[0008] Image aspect ratio calculation module: When the antenna model is unknown, the aspect ratio of the antenna is obtained by taking a picture of the front of the antenna. The aspect ratio of the antenna in the image is used to replace the actual antenna size ratio, and the antenna downtilt angle is calculated by iterating the aspect ratio of the antenna.

[0009] Preferably, in the basic measurement module:

[0010] By approximating the camera imaging principle based on the pinhole camera model when shooting at a preset distance, distance-related parameters in the camera imaging principle are eliminated. The approximation process means that in the image pixel coordinate system, the physical dimensions dx and dy of each pixel in the u-axis and v-axis directions are the same. Using the known physical dimensions of the antenna's bottom and side, the antenna's bottom and side vectors are mapped to the two-dimensional image coordinate system through the above approximated camera imaging principle. The antenna's downtilt angle is solved by fitting the bottom and side vectors of the known image.

[0011] The specific implementation steps are as follows:

[0012] On the image captured by the mobile phone, the points corresponding to points a, b, and c are explicitly marked respectively, and vectors av, bv, and cv are obtained.

[0013] Random initialization parameters include: tilt angle θ, distance and camera intrinsic parameters K, yaw angle α, pitch angle β, roll angle γ;

[0014] The original coordinates are transformed according to the above approximate camera imaging model to obtain the coordinates 'a' projected onto the two-dimensional pixel coordinate system from the three-dimensional coordinates. RGB b RGB c RGB ;

[0015] Minimize the vectors (av, bv, cv) and (a) formed after the graph markings are optimized using the optimizer. RGB b RGB c RGB The mean squared error between the two sides, where the loss function is defined as:

[0016] loss = (a RGB -av) 2 +(b RGB -bv) 2 +(c RGB -cv) 2

[0017] Search and update parameters, iterating repeatedly until the error is less than the preset error coefficient ∈.

[0018] Preferably, in the virtual pole-holding module:

[0019] By taking an image of a checkerboard pattern aligned with the direction of gravity using a mobile phone, determine the gravity vector on the image:

[0020] Hang the chessboard grid vertically, ensuring the plumb line coincides with the grid lines within the chessboard grid.

[0021] Record a video of a chessboard pattern and change the phone's orientation;

[0022] Record IMU data, and denote the gravity vector as r;

[0023] Solve for the camera intrinsic and extrinsic parameter matrices (R, T) using existing data;

[0024] The gravity vector v in each image is solved using intrinsic and extrinsic parameters, coinciding with the y-axis of the image's physical coordinate system;

[0025] Establish the mapping f(r) = v;

[0026] A virtual pole-holding database is established through the above steps, and the image gravity direction corresponding to the gravity vector with the highest similarity at the time of shooting is used to replace the measured pole-holding vector;

[0027] The specific implementation steps are as follows:

[0028] When calibrating the chessboard, the camera intrinsic parameters are first calculated by extracting corner points from some chessboard images. Then, the camera extrinsic parameters are calculated using corner points obtained from other images. Finally, the projection vector of the gravity vector [0, 0, 1] onto the image is calculated using the camera intrinsic and extrinsic parameters, and the results are stored in the database.

[0029] When collecting IMU data, data is recorded before and after the photo is taken at preset times. The average value of the collected IMU data is then taken. The vectors in the virtual pole library are sorted according to their cosine similarity to the IMU data. Data with a similarity to the IMU data exceeding a preset value are filtered out, with a maximum of a preset value of data being filtered. The vectors on the image corresponding to the IMU data are then filtered out, removing vectors with a similarity to most other vectors below the preset value. The average value of the filtered vectors is then taken as the final virtual pole vector on the image.

[0030] Preferably, in the image aspect ratio calculation module:

[0031] The farther the image is taken from the front of the antenna, the closer the antenna aspect ratio obtained from the image will be to the actual antenna aspect ratio; when the actual downtilt angle of the antenna is β, the relationship between the aspect ratio in the image and the actual aspect ratio satisfies:

[0032]

[0033] Where L′ and H′ represent the projected lengths of the antenna's bottom and side sides on the image plane, and L and H represent the physical lengths of the antenna's bottom and side sides, respectively. The phone's elevation angle during shooting is θ.

[0034] Therefore, the aspect ratio obtained from the image is affected by both the phone's elevation angle and the antenna's downtilt angle.

[0035] An iterative downtilt angle optimization algorithm based on formula (5) is designed. The dimensions are corrected by continuously solving the downtilt angle and antenna aspect ratio. The iterative process is as follows:

[0036] The aspect ratio information from close-up shooting is used to solve for the tilt angle θ, and the phone's elevation angle β is calculated from the phone's IMU data; the aspect ratio information is obtained using formula (5). in It is the aspect ratio result of the previous iteration;

[0037] Repeat the above steps until the preset number of iterations is reached to obtain the downtilt angle θ.

[0038] According to the present invention, a method for measuring the downtilt angle of a plate antenna based on mobile phone information fusion is provided, which employs the aforementioned plate antenna downtilt angle measurement system based on mobile phone information fusion and performs the following steps:

[0039] Step S1: Take an image of the antenna from a preset distance using a mobile phone;

[0040] Step S2: Process the image to eliminate distance-related parameters in the camera imaging principle;

[0041] Step S3: If the antenna information is known, map the antenna bottom and side vectors onto the two-dimensional image coordinate system, and solve the antenna downtilt angle by fitting the bottom and side vectors on the image.

[0042] If the antenna information is unknown, the antenna downtilt angle is solved by iterating through the antenna's aspect ratio.

[0043] Step S4: Take an image of the chessboard grid aligned with the direction of gravity using a mobile phone, solve for the gravity vector on the image, and obtain the downward tilt angle after error correction.

[0044] Preferably, in step S2:

[0045] By approximating the imaging principle of a camera based on a pinhole camera model when shooting at a preset distance, distance-related parameters in the imaging principle are eliminated. The approximation process means that in the image pixel coordinate system, the physical dimensions dx and dy of each pixel in the u-axis and v-axis directions are the same.

[0046] Preferably, in step S3:

[0047] Using the known physical dimensions of the antenna's base and sides, the antenna's base and side vectors are mapped onto a two-dimensional image coordinate system using the aforementioned approximated camera imaging principle. The antenna's downtilt angle is then solved by fitting the base and side vectors of the known image.

[0048] The specific implementation steps are as follows:

[0049] On the image captured by the mobile phone, the points corresponding to points a, b, and c are explicitly marked respectively, and vectors av, bv, and cv are obtained.

[0050] Random initialization parameters include: tilt angle θ, distance and camera intrinsic parameters K, yaw angle α, pitch angle β, roll angle γ;

[0051] The original coordinates are transformed according to the above approximate camera imaging model to obtain the coordinates 'a' projected onto the two-dimensional pixel coordinate system from the three-dimensional coordinates. RGB b RGB c RGB ;

[0052] Minimize the vectors (av, bv, cv) and (a) formed after the graph markings are optimized using the optimizer. RGB b RGB c RGB The mean squared error between the two sides, where the loss function is defined as:

[0053] loss = (a RGB -av) 2 +(b RGB -bv) 2 +(c RGB -cv) 2

[0054] Search and update parameters, iterating repeatedly until the error is less than the preset error coefficient ∈.

[0055] Preferably, in step S3:

[0056] When the antenna model is unknown, the aspect ratio of the antenna is obtained by taking a picture of the front of the antenna. The aspect ratio of the antenna in the image is used to replace the actual antenna size ratio, and the antenna downtilt angle is solved by iterating the aspect ratio of the antenna.

[0057] Preferably, in step S3:

[0058] The farther the image is taken from the front of the antenna, the closer the antenna aspect ratio obtained from the image will be to the actual antenna aspect ratio; when the actual downtilt angle of the antenna is β, the relationship between the aspect ratio in the image and the actual aspect ratio satisfies:

[0059]

[0060] Where L′ and H′ represent the projected lengths of the antenna's bottom and side sides on the image plane, and L and H represent the physical lengths of the antenna's bottom and side sides, respectively. The phone's elevation angle during shooting is θ.

[0061] Therefore, the aspect ratio obtained from the image is affected by both the phone's elevation angle and the antenna's downtilt angle.

[0062] An iterative downtilt angle optimization algorithm based on formula (5) is designed. The dimensions are corrected by continuously solving the downtilt angle and antenna aspect ratio. The iterative process is as follows:

[0063] The aspect ratio information from close-up shooting is used to solve for the tilt angle θ, and the phone's elevation angle β is calculated from the phone's IMU data; the aspect ratio information is obtained using formula (5). in It is the aspect ratio result of the previous iteration;

[0064] Repeat the above steps until the preset number of iterations is reached to obtain the downtilt angle θ.

[0065] Preferably, in step S4:

[0066] By taking an image of a checkerboard pattern aligned with the direction of gravity using a mobile phone, determine the gravity vector on the image:

[0067] Hang the chessboard grid vertically, ensuring the plumb line coincides with the grid lines within the chessboard grid.

[0068] Record a video of a chessboard pattern and change the phone's orientation;

[0069] Record IMU data, and denote the gravity vector as r;

[0070] Solve for the camera intrinsic and extrinsic parameter matrices (R, T) using existing data;

[0071] The gravity vector v in each image is solved using intrinsic and extrinsic parameters, coinciding with the y-axis of the image's physical coordinate system;

[0072] Establish the mapping f(r) = v;

[0073] A virtual pole-holding database is established through the above steps, and the image gravity direction corresponding to the gravity vector with the highest similarity at the time of shooting is used to replace the measured pole-holding vector;

[0074] The specific implementation steps are as follows:

[0075] When calibrating the chessboard, the camera intrinsic parameters are first calculated by extracting corner points from some chessboard images. Then, the camera extrinsic parameters are calculated using corner points obtained from other images. Finally, the projection vector of the gravity vector [0, 0, 1] onto the image is calculated using the camera intrinsic and extrinsic parameters, and the results are stored in the database.

[0076] When collecting IMU data, data is recorded before and after the photo is taken at preset times. The average value of the collected IMU data is then taken. The vectors in the virtual pole library are sorted according to their cosine similarity to the IMU data. Data with a similarity to the IMU data exceeding a preset value are filtered out, with a maximum of a preset value of data being filtered. The vectors on the image corresponding to the IMU data are then filtered out, removing vectors with a similarity to most other vectors below the preset value. The average value of the filtered vectors is then taken as the final virtual pole vector on the image.

[0077] Compared with the prior art, the present invention has the following beneficial effects:

[0078] 1. The algorithm for solving the downtilt angle of a plate antenna based on mobile phone photography and mobile phone IMU data proposed in this invention has extremely high portability, ease of use and low cost compared with the measurement methods of traditional slope gauges and total stations;

[0079] 2. The virtual pole-holding technology proposed in this invention can effectively solve the pole-holding tilt problem caused by sensor noise and the fact that the gravity sensor is not parallel to the mobile phone plane due to welding and other issues in the tilt angle algorithm, thus effectively improving the universality and stability of the algorithm.

[0080] 3. The algorithm proposed in this invention, which calculates the downtilt angle by the aspect ratio of the antenna in the image, solves the problem of difficulty in calculating the tilt angle caused by the inability to distinguish the antenna model when shooting at a distance in some scenarios, and improves the adaptability and robustness of the algorithm in extreme scenarios.

[0081] 4. This invention provides rigorous simulation experiments and solvability proofs for the core algorithm, ensuring the stability and reliability of the algorithm's operation; Attached Figure Description

[0082] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0083] Figure 1 This is a flowchart of the present invention;

[0084] Figure 2 This is a flowchart of the algorithm for fusing sensor data measurement with mobile phone photography in this invention;

[0085] Figure 3 This is a schematic diagram showing the vectors of the antenna bottom edge, side edge, and mast in this invention;

[0086] Figure 4 This is a diagram illustrating the principle of using the antenna aspect ratio in an image to calculate the downtilt angle in this invention. Detailed Implementation

[0087] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0088] Example 1:

[0089] According to the present invention, a plate antenna downtilt angle measurement system based on mobile phone information fusion is provided, such as... Figures 1-4 As shown, it includes:

[0090] Basic Measurement Module: By capturing antenna images with a mobile phone, distance-related parameters in the camera imaging principle are eliminated. The antenna's bottom and side vectors are marked on the image. The antenna's downtilt angle is calculated by combining IMU data and specific antenna model information.

[0091] Virtual support pole module: By taking an image of a chessboard grid aligned with the direction of gravity using a mobile phone, the gravity vector on the image is calculated, and the downward tilt angle after error correction is obtained by replacing the support pole vector with the gravity direction corresponding to the gravity vector in the image.

[0092] Image aspect ratio calculation module: When the antenna model is unknown, the aspect ratio of the antenna is obtained by taking a picture of the front of the antenna. The aspect ratio of the antenna in the image is used to replace the actual antenna size ratio, and the antenna downtilt angle is calculated by iterating the aspect ratio of the antenna.

[0093] Specifically, in the basic measurement module:

[0094] By approximating the camera imaging principle based on the pinhole camera model when shooting at a preset distance, distance-related parameters in the camera imaging principle are eliminated. The approximation process means that in the image pixel coordinate system, the physical dimensions dx and dy of each pixel in the u-axis and v-axis directions are the same. Using the known physical dimensions of the antenna's bottom and side, the antenna's bottom and side vectors are mapped to the two-dimensional image coordinate system through the above approximated camera imaging principle. The antenna's downtilt angle is solved by fitting the bottom and side vectors of the known image.

[0095] The specific implementation steps are as follows:

[0096] On the image captured by the mobile phone, the points corresponding to points a, b, and c are explicitly marked respectively, and vectors av, bv, and cv are obtained.

[0097] Random initialization parameters include: tilt angle θ, distance and camera intrinsic parameters K, yaw angle α, pitch angle β, roll angle γ;

[0098] The original coordinates are transformed according to the above approximate camera imaging model to obtain the coordinates 'a' projected onto the two-dimensional pixel coordinate system from the three-dimensional coordinates. RGB b RGB c RGB ;

[0099] Minimize the vectors (av, bv, cv) and (a) formed after the graph markings are optimized using the optimizer. RGB b RGB c RGB The mean squared error between the two sides, where the loss function is defined as:

[0100] loss = (a RGB -av) 2 +(b RGB -bv) 2 +(c RGB -cv) 2

[0101] Search and update parameters, iterating repeatedly until the error is less than the preset error coefficient ∈.

[0102] Specifically, in the virtual pole-holding module:

[0103] By taking an image of a checkerboard pattern aligned with the direction of gravity using a mobile phone, determine the gravity vector on the image:

[0104] Hang the chessboard grid vertically, ensuring the plumb line coincides with the grid lines within the chessboard grid.

[0105] Record a video of a chessboard pattern and change the phone's orientation;

[0106] Record IMU data, and denote the gravity vector as r;

[0107] Solve for the camera intrinsic and extrinsic parameter matrices (R, T) using existing data;

[0108] The gravity vector v in each image is solved using intrinsic and extrinsic parameters, coinciding with the y-axis of the image's physical coordinate system;

[0109] Establish the mapping i(r) = v;

[0110] A virtual pole-holding database is established through the above steps, and the image gravity direction corresponding to the gravity vector with the highest similarity at the time of shooting is used to replace the measured pole-holding vector;

[0111] The specific implementation steps are as follows:

[0112] When calibrating the chessboard, the camera intrinsic parameters are first calculated by extracting corner points from some chessboard images. Then, the camera extrinsic parameters are calculated using corner points obtained from other images. Finally, the projection vector of the gravity vector [0, 0, 1] onto the image is calculated using the camera intrinsic and extrinsic parameters, and the results are stored in the database.

[0113] When collecting IMU data, data is recorded before and after the photo is taken at preset times. The average value of the collected IMU data is then taken. The vectors in the virtual pole library are sorted according to their cosine similarity to the IMU data. Data with a similarity to the IMU data exceeding a preset value are filtered out, with a maximum of a preset value of data being filtered. The vectors on the image corresponding to the IMU data are then filtered out, removing vectors with a similarity to most other vectors below the preset value. The average value of the filtered vectors is then taken as the final virtual pole vector on the image.

[0114] Specifically, in the image aspect ratio calculation module:

[0115] The farther the image is taken from the front of the antenna, the closer the antenna aspect ratio obtained from the image will be to the actual antenna aspect ratio; when the actual downtilt angle of the antenna is β, the relationship between the aspect ratio in the image and the actual aspect ratio satisfies:

[0116]

[0117] Where L′ and H′ represent the projected lengths of the antenna's bottom and side sides on the image plane, and L and H represent the physical lengths of the antenna's bottom and side sides, respectively. The phone's elevation angle during shooting is θ.

[0118] Therefore, the aspect ratio obtained from the image is affected by both the phone's elevation angle and the antenna's downtilt angle.

[0119] An iterative downtilt angle optimization algorithm based on formula (5) is designed. The dimensions are corrected by continuously solving the downtilt angle and antenna aspect ratio. The iterative process is as follows:

[0120] The aspect ratio information from close-up shooting is used to solve for the tilt angle θ, and the phone's elevation angle β is calculated from the phone's IMU data; the aspect ratio information is obtained using formula (5). in It is the aspect ratio result of the previous iteration;

[0121] Repeat the above steps until the preset number of iterations is reached to obtain the downtilt angle θ.

[0122] According to the present invention, a method for measuring the downtilt angle of a plate antenna based on mobile phone information fusion is provided, which employs the aforementioned plate antenna downtilt angle measurement system based on mobile phone information fusion and performs the following steps:

[0123] Step S1: Take an image of the antenna from a preset distance using a mobile phone;

[0124] Step S2: Process the image to eliminate distance-related parameters in the camera imaging principle;

[0125] Step S3: If the antenna information is known, map the antenna bottom and side vectors onto the two-dimensional image coordinate system, and solve the antenna downtilt angle by fitting the bottom and side vectors on the image.

[0126] If the antenna information is unknown, the antenna downtilt angle is solved by iterating through the antenna's aspect ratio.

[0127] Step S4: Take an image of the chessboard grid aligned with the direction of gravity using a mobile phone, solve for the gravity vector on the image, and obtain the downward tilt angle after error correction.

[0128] Specifically, in step S2:

[0129] By approximating the imaging principle of a camera based on a pinhole camera model when shooting at a preset distance, distance-related parameters in the imaging principle are eliminated. The approximation process means that in the image pixel coordinate system, the physical dimensions dx and dy of each pixel in the u-axis and v-axis directions are the same.

[0130] Specifically, in step S3:

[0131] Using the known physical dimensions of the antenna's base and sides, the antenna's base and side vectors are mapped onto a two-dimensional image coordinate system using the aforementioned approximated camera imaging principle. The antenna's downtilt angle is then solved by fitting the base and side vectors of the known image.

[0132] The specific implementation steps are as follows:

[0133] On the image captured by the mobile phone, the points corresponding to points a, b, and c are explicitly marked respectively, and vectors av, bv, and cv are obtained.

[0134] Random initialization parameters include: tilt angle θ, distance and camera intrinsic parameters K, yaw angle α, pitch angle β, roll angle γ;

[0135] The original coordinates are transformed according to the above approximate camera imaging model to obtain the coordinates 'a' projected onto the two-dimensional pixel coordinate system from the three-dimensional coordinates. RGB b RGB c RGB ;

[0136] Minimize the vectors (av, bv, cv) and (a) formed after the graph markings are optimized using the optimizer. RGB b RGB c RGB The mean squared error between the two sides, where the loss function is defined as:

[0137] loss = (a RGB -av) 2 +(b RGB -bv) 2 +(c RGB -cv) 2

[0138] Search and update parameters, iterating repeatedly until the error is less than the preset error coefficient ∈.

[0139] Specifically, in step S3:

[0140] When the antenna model is unknown, the aspect ratio of the antenna is obtained by taking a picture of the front of the antenna. The aspect ratio of the antenna in the image is used to replace the actual antenna size ratio, and the antenna downtilt angle is solved by iterating the aspect ratio of the antenna.

[0141] Specifically, in step S3:

[0142] The farther the image is taken from the front of the antenna, the closer the antenna aspect ratio obtained from the image will be to the actual antenna aspect ratio; when the actual downtilt angle of the antenna is β, the relationship between the aspect ratio in the image and the actual aspect ratio satisfies:

[0143]

[0144] Where L′ and H′ represent the projected lengths of the antenna's bottom and side sides on the image plane, and L and H represent the physical lengths of the antenna's bottom and side sides, respectively. The phone's elevation angle during shooting is θ.

[0145] Therefore, the aspect ratio obtained from the image is affected by both the phone's elevation angle and the antenna's downtilt angle.

[0146] An iterative downtilt angle optimization algorithm based on formula (5) is designed. The dimensions are corrected by continuously solving the downtilt angle and antenna aspect ratio. The iterative process is as follows:

[0147] The aspect ratio information from close-up shooting is used to solve for the tilt angle θ, and the phone's elevation angle β is calculated from the phone's IMU data; the aspect ratio information is obtained using formula (5). in It is the aspect ratio result of the previous iteration;

[0148] Repeat the above steps until the preset number of iterations is reached to obtain the downtilt angle θ.

[0149] Specifically, in step S4:

[0150] By taking an image of a checkerboard pattern aligned with the direction of gravity using a mobile phone, determine the gravity vector on the image:

[0151] Hang the chessboard grid vertically, ensuring the plumb line coincides with the grid lines within the chessboard grid.

[0152] Record a video of a chessboard pattern and change the phone's orientation;

[0153] Record IMU data, and denote the gravity vector as r;

[0154] Solve for the camera intrinsic and extrinsic parameter matrices (R, T) using existing data;

[0155] The gravity vector v in each image is solved using intrinsic and extrinsic parameters, coinciding with the y-axis of the image's physical coordinate system;

[0156] Establish the mapping f(r) = v;

[0157] A virtual pole-holding database is established through the above steps, and the image gravity direction corresponding to the gravity vector with the highest similarity at the time of shooting is used to replace the measured pole-holding vector;

[0158] The specific implementation steps are as follows:

[0159] When calibrating the chessboard, the camera intrinsic parameters are first calculated by extracting corner points from some chessboard images. Then, the camera extrinsic parameters are calculated using corner points obtained from other images. Finally, the projection vector of the gravity vector [0, 0, 1] onto the image is calculated using the camera intrinsic and extrinsic parameters, and the results are stored in the database.

[0160] When collecting IMU data, data is recorded before and after the photo is taken at preset times. The average value of the collected IMU data is then taken. The vectors in the virtual pole library are sorted according to their cosine similarity to the IMU data. Data with a similarity to the IMU data exceeding a preset value are filtered out, with a maximum of a preset value of data being filtered. The vectors on the image corresponding to the IMU data are then filtered out, removing vectors with a similarity to most other vectors below the preset value. The average value of the filtered vectors is then taken as the final virtual pole vector on the image.

[0161] Example 2:

[0162] Example 2 is a preferred embodiment of Example 1, and is used to illustrate the present invention in more detail.

[0163] 1. A method for measuring the downtilt angle of a plate antenna based on a combination of mobile phone camera photography and IMU data. The method involves taking a picture of the antenna with a mobile phone, marking the bottom and side vectors of the antenna on the image, and combining IMU data and specific antenna model information to solve the downtilt angle of the plate antenna by minimizing the reprojection error.

[0164] include:

[0165] By approximating the camera imaging principle based on a pinhole camera model for long-distance shooting, distance-related parameters in the camera imaging principle are eliminated. Then, using the known physical dimensions of the bottom and side of the plate antenna, the antenna's bottom and side vectors are mapped to a two-dimensional image coordinate system through the improved camera imaging principle. The downtilt angle of the plate antenna is then solved by fitting it to the known bottom and side vectors on the image.

[0166] Approximation refers to assuming that the physical dimensions of each pixel in the u-axis and v-axis directions, dx and dy, are the same in the image pixel coordinate system, so that the number of equations and the number of unknowns are the same, thereby obtaining the correct solution.

[0167] 2. A gravity vector calibration technique based on mobile phone photography and IMU data, which calculates the gravity vector in an image by taking a picture of a checkerboard pattern aligned with the direction of gravity. The core operation includes the following steps:

[0168] (1) Hang the chessboard grid vertically so that the plumb line coincides with the grid line inside the chessboard grid;

[0169] (2) Record a video of the chessboard pattern and continuously change the phone's orientation;

[0170] (3) Simultaneously record IMU data (gravity vector denoted as r);

[0171] (4) Solve for the camera intrinsic and extrinsic parameters (R, T) using the existing data;

[0172] (5) Solve for the gravity vector v in each image using intrinsic and extrinsic parameters (coinciding with the y-axis of the physical coordinate system of the image);

[0173] (6) Establish the mapping f(r) = v;

[0174] 3. A method for measuring the downtilt angle of a plate antenna based on a combination of mobile phone camera photos and IMU data, which does not require specific antenna information. This method uses the length and width of the plate antenna in the image to replace the actual size ratio of the plate antenna for measuring the downtilt angle. The method optimizes the solution of the antenna downtilt angle by continuously iterating and optimizing the aspect ratio of the antenna.

[0175] Example 3:

[0176] Example 3 is a preferred example of Example 1, and is used to illustrate the present invention in more detail.

[0177] This invention relates to antenna downtilt angle measurement technology based on mobile phone photography and IMU data fusion, specifically applicable to outdoor panel antenna downtilt angle measurement, providing a convenient downtilt angle measurement method for practical network optimization, etc.

[0178] The technical problem this invention aims to solve is how to accurately measure the downtilt angle of a plate antenna using data from a mobile phone camera and a mobile phone IMU.

[0179] To address this technical problem, this invention provides an algorithm for calculating the downtilt angle of a plate antenna based on mobile phone photography and mobile phone IMU data. To improve the algorithm's accuracy and robustness in different scenarios, a virtual pole technique and a technique for calculating the downtilt angle using the antenna's aspect ratio in the image are proposed. At the same time, rigorous simulation experiments and proof of the algorithm's solvability are provided.

[0180] The method includes the following hardware: a smartphone with a camera and an IMU sensor.

[0181] The method includes the following steps:

[0182] This invention aims to obtain the antenna downtilt angle by capturing images of a plate antenna at a distance (>30 meters) and then using the "basic measurement algorithm" described below. When using the "basic measurement algorithm," the direction of the gravity vector is required. The algorithm uses the direction of the antenna mast to represent this direction, but in actual measurements, the antenna mast may be misaligned. Therefore, a "virtual mast technique" is proposed to correct this error. In other words, combining the "basic measurement algorithm" and the "virtual mast technique" allows for a more accurate measurement of the downtilt angle.

[0183] Basic Measurement Algorithm: During camera capture, the distance information from the camera's optical center to the antenna panel in 3D space is lost. According to the basic principles of camera imaging, it is impossible to directly construct the specific pose of an object in 3D space from a single 2D image. To address this problem, the camera imaging principle is improved, eliminating distance-related parameters. Then, using the known physical dimensions of the base and sides of the plate antenna, the base and side vectors are mapped onto the 2D image coordinate system using the improved camera imaging principle. The downtilt angle of the plate antenna is then calculated by fitting the base and side vectors from the known image.

[0184] Virtual mast technology: Virtual mast technology is a gravity calibration technique proposed to improve algorithm accuracy. In the downtilt angle calculation process, the antenna mast vector is a crucial parameter determining the accuracy of the downtilt angle. However, in actual operation, the antenna mast is not necessarily perfectly perpendicular to the ground. When directly projecting the mast vector from IMU data, issues such as sensor noise and inconsistencies between the sensor plane and the phone's plane also exist, which can affect the accuracy of the downtilt angle calculation. To address this problem, we propose virtual mast technology to determine the direction of gravity in the image. The basic process of establishing virtual mast technology is as follows:

[0185] 1. Solve for the camera intrinsic parameter matrix;

[0186] 2. Establish the mapping relationship between the gravity vector and the gravity vector on the image by photographing a vertically suspended chessboard.

[0187] The virtual pole-mounting database is established through the two steps described above. In practical applications, the image gravity direction corresponding to the gravity vector with the highest similarity to that captured during the shooting is used to replace the pole-mounting vector measured in this instance. After extensive practical testing, it was found that the virtual pole-mounting technology can significantly improve the accuracy of the downtilt angle calculation.

[0188] Image Aspect Ratio Determination: In the downtilt angle calculation algorithm, the actual physical dimensions of the antenna are required as constraints to solve for the downtilt angle. The specific dimensions of the antenna can be obtained by consulting the corresponding antenna model. However, in reality, the shape of plate antennas is not easily distinguishable, making it impossible to identify the antenna model from its appearance. Therefore, a method was designed to obtain the antenna's aspect ratio by taking a picture of the antenna directly facing the plate, replacing the actual physical dimensions. However, at large elevation angles, obtaining the aspect ratio from the image will produce significant errors. Therefore, through derivation, an iterative aspect ratio correction algorithm is proposed, improving the algorithm's robustness and allowing it to be applied to more scenarios.

[0189] The following section will elaborate on this content, detailing the algorithm for measuring the downtilt angle of a plate antenna based on mobile phone photography and mobile phone IMU data. The "virtual pole technology" and "image aspect ratio calculation" will also be described below.

[0190] Downtilt Angle Measurement Algorithm: Measurement Principle and Steps

[0191] When shooting at a long distance, we can approximate the actual physical dimensions dx and dy corresponding to a unit pixel in the image coordinate system to be the same. Secondly, the actual physical distance Z from the camera's optical center to the antenna plate cannot be recovered from a single image. c Therefore, the camera imaging principle is simplified by eliminating distance-related parameters: the translation vector T and the distance Z from the camera's optical center to the antenna plate. c The following coordinate transformation model is obtained:

[0192]

[0193] Where (u, v) represents the coordinates of a point (X, V) in the three-dimensional world coordinate system. w Y w Z w The coordinates projected onto the camera image pixel coordinate system. This is a parameter related to camera intrinsics and distance. dx and dy represent the actual physical size of each pixel in the two axes of the image coordinate system. When shooting from a distance, dx can be approximated as dy. The algorithm will use formula (1) for spatial coordinate transformation. The specific algorithm principle is as follows:

[0194] 1. Establish three vectors at the corner points of the antenna, corresponding to the bottom edge, side edge, and mast of the antenna, respectively (see attached diagram). Figure 3 (As shown). It is also assumed that the antenna base, mast, and coordinate axes coincide at the original vector position. Assuming the antenna downtilt angle is θ, the above three vectors can be represented in the physical coordinate system as:

[0195] a0 = [L, 0, 0] T b0 = [0, Hsinθ, Hcosθ] T c0 = [0, 0, 1] T

[0196] 2. After undergoing a rigid body transformation, the above vector can be represented in the camera coordinate system as follows:

[0197] a = Ra0, b = Rb0, c = Rc0

[0198] Where R is the rotation matrix of the actual vector relative to the camera, containing three unknown parameters (yaw angle, pitch angle, and roll angle).

[0199] 3. For long-distance observations exceeding 30 meters, the coordinate transformation model improved by formula (1) is used to approximately obtain the coordinates of points a, b, and c mapped to three points on the image:

[0200] a RGB =Ma,b RGB =Mb,c RGB =M c

[0201] in

[0202]

[0203] At this point, we can obtain the constraint conditions for solving the downtilt angle θ:

[0204]

[0205] The above solution principle contains a total of five degrees of freedom: the rotation matrix R contains three degrees of freedom (represented by yaw angle α, pitch angle β and roll angle γ respectively), and the distance and camera intrinsic parameters K and downtilt angle θ. The three constraints provided in formula (2) leave only two degrees of freedom for the antenna pose parameters, namely: the distance Z from the antenna corner point to the optical center. c The rotation angle of the antenna normal vector (based on constraints 1 and 2, only one degree of freedom remains), and the actual length of the antenna brings two constraint equations, which are solvable in terms of degrees of freedom and number of equations.

[0206] The above describes the theoretical solution principle of the downtilt angle algorithm. However, in the algorithm implementation, the downtilt angle θ is an unknown parameter and therefore cannot be solved directly. Therefore, this method uses a fitting approach to optimize the solution of the downtilt angle θ. The specific implementation steps are as follows (see attached diagram). Figure 2 As shown):

[0207] 1. Explicitly mark the positions corresponding to points a, b, and c on the image captured by the mobile phone, and obtain vectors av, bv, and cv.

[0208] 2. Randomly initialize parameters: θ, K, α, β, γ;

[0209] 3. Transform the original coordinates according to the simplified camera imaging model described above to obtain the coordinates 'a' projected onto the two-dimensional pixel coordinate system from the three-dimensional coordinates. RGB b RGB c RGB ;

[0210] 4. Minimize the vectors (av, bv, cv) and (a) formed after marking on the graph using the optimizer. RGB b RGB c RGB The mean squared error between ) and ), where the loss function is defined as:

[0211] loss = (a RGB -av) 2 +(b RGB -bv) 2 +(c RGB -cv) 2

[0212] 5. Use the Powell optimization algorithm (a algorithm for finding local minima of a function) to search and update the parameters, iterating repeatedly until the error is less than the given error coefficient ∈.

[0213] Virtual pole mounting technology:

[0214] In actual downtilt angle measurements, the antenna mast is not necessarily perfectly perpendicular to the ground, and a tilted mast can significantly affect the accuracy of the downtilt angle calculated by the algorithm. Therefore, this section introduces virtual mast technology to replace the tilted mast in real-world scenarios. The core operation steps are as follows:

[0215] 1. Hang the chessboard grid vertically, ensuring the plumb line coincides with the grid lines within the chessboard grid;

[0216] 2. Record a video of the chessboard pattern while continuously changing the phone's orientation;

[0217] 3. Simultaneously record IMU data (gravity vector denoted as r);

[0218] 4. Solve for the camera intrinsic and extrinsic parameter matrices (R, T) using the existing data;

[0219] 5. Solve for the gravity vector v in each image using intrinsic and extrinsic parameters (coinciding with the y-axis of the image's physical coordinate system);

[0220] 6. Establish the mapping f(r) = v;

[0221] During checkerboard calibration, the camera intrinsic parameters are first calculated by extracting corner points from a portion of the checkerboard image. Then, using corner points obtained from other images, the PnP algorithm (a method for solving 3D to 2D point pair motion) is employed to calculate the camera extrinsic parameters. Using the camera intrinsic and extrinsic parameters, the projection vector of the gravity vector [0, 0, 1] (which coincides with the z-axis due to the vertically suspended checkerboard) onto the image is calculated, and the results are then stored in a database.

[0222] To further improve the robustness and accuracy of the algorithm's measurements and reduce the influence of the external environment, measures such as recording data one second before and after taking a picture can be used when acquiring IMU data. This reduces the impact of large instantaneous IMU data deviations caused by factors such as hand tremors during shooting. The acquired IMU data is then processed using a mean-shift algorithm and averaged to reduce the deviation of the virtual boom caused by IMU bias, thus obtaining the accurate IMU value.

[0223] After basic processing of the IMU data, the vectors in the virtual pole library are sorted according to their cosine similarity to the IMU data. Then, data with a similarity exceeding 99.5% is selected, up to a maximum of 200. Next, the vectors on the corresponding images from these IMU data are filtered out, removing vectors with a similarity below 99% to most other vectors, reducing errors caused by virtual pole calibration mistakes. Finally, the average of the selected vectors is taken as the final virtual pole vector on the image.

[0224] Image aspect ratio determination:

[0225] Antenna dimensions can be obtained by consulting the corresponding antenna model number, and accurate antenna dimensions can significantly improve the accuracy of downtilt angle calculation. However, in real-world scenarios, many antennas look very similar, making it impossible to directly determine the antenna model from its appearance. This poses a new challenge to the practicality of the algorithm proposed in this patent. To solve this problem, the aspect ratio of the antenna in the image taken from the front of the antenna is obtained, and then the aspect ratio is used to replace the actual antenna size for calculation.

[0226] Through theoretical derivation, it can be determined that the projected lengths of the antenna's bottom and sides can lead to two constraint equations:

[0227] l=||F(q0+v1L)-F(q0)||, h=||F(q0+v2H)-F(q0)|| (3)

[0228] in It is a mapping function from a spatial point to the camera image coordinate system (pinhole camera model). For long-distance shooting, there must exist a D such that D >> max(||X||). p ||,||Y p When ||), F(p) can be linearized into F D (p), that is, replacing Z in formula (3) with a fixed denominator D. p At this point, we can obtain:

[0229]

[0230]

[0231] We can eliminate one unknown D by solving the ratio of the two equations above:

[0232]

[0233] From formula (4), it can be derived that the ratio of the lengths of the two sides of the antenna on a given image is... Ratio to the actual antenna size Equation (4) is the new constraint equation, used to determine the remaining degree of freedom. Therefore, from theoretical analysis, we can conclude that substituting the antenna size ratio information is completely solvable.

[0234] As shown in formula (4), the physical size ratio of the antenna is required to solve for the downtilt angle. Therefore, we approximate the physical size ratio of the antenna by taking a picture from the front of the antenna. When taking a picture from the front of the antenna, the distance between the shooting point and the antenna mast determines the error between the aspect ratio of the antenna in the image and the actual aspect ratio of the antenna. Therefore, in the following text, this patent will theoretically analyze the degree of difference between the aspect ratio of the antenna in the image and the actual aspect ratio of the antenna under different shooting elevation angles (distance between the shooting point and the antenna mast) and give a specific optimization scheme.

[0235] As attached Figure 4 The diagram illustrates a simplified process of antenna imaging on the camera plane. For ease of analysis, it is assumed that the antenna tilt angle is 0 degrees during this imaging process. In the diagram, d represents the distance from the shooting point to the antenna mast, θ represents the phone's elevation angle during shooting, h represents the mast height, L and H represent the physical lengths of the antenna's base and side, respectively, and L′ and H′ represent the projected lengths of the antenna's base and side on the image plane. For the pinhole camera model, based on triangle similarity, we can obtain:

[0236]

[0237]

[0238] Solving for:

[0239]

[0240] in Right now When d >> h, we get: The farther the image is taken from the front of the antenna, the closer the antenna size ratio obtained from the image will be to the actual antenna size ratio. When the actual downtilt angle of the antenna is β, it is not difficult to find that the relationship between the aspect ratio of the image and the actual aspect ratio satisfies:

[0241]

[0242] That is, the aspect ratio obtained from the image and the aspect ratio of the actual antenna are both affected by the phone's elevation angle and the antenna's downtilt angle.

[0243] In practical applications, geographical factors may prevent us from taking photos from a distance far enough from the antenna mast. Therefore, we designed an iterative downtilt angle optimization algorithm based on formula (5). By continuously solving for the downtilt angle and antenna aspect ratio, we correct the dimensions, thereby achieving the goal of correcting the antenna downtilt angle error. The iterative process is as follows:

[0244] 1. Incorporating aspect ratio information with large near-distance errors, the downtilt angle θ is calculated using the downtilt angle solution algorithm mentioned in this patent, while the phone's elevation angle β is calculated from the phone's IMU data;

[0245] 2. Use formula (5) to obtain the new aspect ratio information. in It is the aspect ratio result of the previous iteration;

[0246] 3. Repeat steps 1 and 2 above until the given number of iterations is reached to obtain the final downtilt angle θ.

[0247] In this application description, it should be understood that the term "downtilt angle" refers to the angle between the surface of the plate antenna and the direction of gravity. For ease of description, in this patent description, "downtilt angle" refers to the concept defined above.

[0248] In the description of this application, the explanation of concepts such as the camera imaging principle based on the pinhole camera model and the PnP algorithm, which are familiar to those skilled in the art, will be omitted. Those skilled in the art can fully construct and use this algorithm based on its essential content.

[0249] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0250] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A system for measuring the downtilt angle of a plate antenna based on mobile phone information fusion, characterized in that, include: Basic Measurement Module: By capturing antenna images with a mobile phone, distance-related parameters in the camera imaging principle are eliminated. The antenna's bottom and side vectors are marked on the image. The antenna's downtilt angle is calculated by combining IMU data and specific antenna model information. Virtual support pole module: By taking an image of a chessboard grid aligned with the direction of gravity using a mobile phone, the gravity vector on the image is calculated, and the downward tilt angle after error correction is obtained by replacing the support pole vector with the gravity direction corresponding to the gravity vector in the image. Image aspect ratio calculation module: When the antenna model is unknown, the aspect ratio of the antenna is obtained by taking a picture of the front of the antenna. The aspect ratio of the antenna in the image is used to replace the actual antenna size ratio, and the antenna downtilt angle is calculated by iterating the aspect ratio of the antenna. In the virtual pole-mounting module: By taking an image of a checkerboard pattern aligned with the direction of gravity using a mobile phone, determine the gravity vector on the image: Hang the chessboard grid vertically, ensuring the plumb line coincides with the grid lines within the chessboard grid. Record a video of a chessboard pattern and change the phone's orientation; Record IMU data, with the gravity vector denoted as... ; Solve for the camera intrinsic and extrinsic parameters using existing data. ; The gravity vector v in each image is solved using intrinsic and extrinsic parameters, coinciding with the y-axis of the image's physical coordinate system; Establish mapping ; A virtual pole-holding database is established through the above steps, and the image gravity direction corresponding to the gravity vector with the highest similarity at the time of shooting is used to replace the measured pole-holding vector; The specific implementation steps are as follows: During checkerboard calibration, the camera's intrinsic parameters are first calculated by extracting corner points from a portion of the checkerboard image. Then, using corner points obtained from other images, the camera's extrinsic parameters are calculated. Finally, the gravity vector is determined using these intrinsic and extrinsic parameters. The projection vectors onto the image are then stored in the database. Record data before and after taking a picture at preset times when collecting IMU data, and take the average of the collected IMU data; The vectors in the virtual pole library are sorted according to their cosine similarity to the IMU data. Data with a similarity to the IMU data exceeding a preset value are selected, with a maximum of a preset value of data being selected. The vectors on the images corresponding to the IMU data are filtered to remove vectors with a similarity to most other vectors below the preset value. The average of the selected vectors is taken as the final virtual pole vector on the image.

2. The downtilt angle measurement system for a plate antenna based on mobile phone information fusion according to claim 1, characterized in that, In the basic measurement module: By approximating the camera imaging principle based on the pinhole camera model when shooting at a preset distance, distance-related parameters in the camera imaging principle are eliminated. The approximation process means that in the image pixel coordinate system, the physical dimensions di and di of each pixel in the 𝑢-axis and 𝑣-axis directions are the same. Using the known physical dimensions of the antenna's bottom and side, the antenna's bottom and side vectors are mapped to the two-dimensional image coordinate system through the above approximated camera imaging principle. The antenna's downtilt angle is solved by fitting it with the known bottom and side vectors in the image. The specific implementation steps are as follows: Explicitly mark the images captured by the mobile phone. The vector is obtained from the point corresponding to the point. ; Random initialization parameters include: downtilt angle , ， ， Roll angle ; The original coordinates are transformed according to the above approximate camera imaging model to obtain coordinates that are three-dimensional coordinates projected onto a two-dimensional pixel coordinate system. ; Minimize the vector formed after marking the graph using the optimizer. )and( The mean squared error between ) and the loss function Defined as: Search and update parameters, iterating repeatedly until the error is less than the preset error coefficient. .

3. The plate antenna downtilt angle measurement system based on mobile phone information fusion according to claim 1, characterized in that, In the image aspect ratio calculation module: The farther the image is taken from the front of the antenna, the closer the antenna size ratio obtained from the image will be to the actual antenna size ratio; when the actual downtilt angle of the antenna is... At this time, the relationship between the aspect ratio of the image and the actual aspect ratio satisfies: (5) in, and This represents the projected lengths of the antenna's base and side edges onto the image plane. 𝐿 and 𝐻 represent the physical lengths of the antenna's base and side edges, respectively. The phone's elevation angle during shooting is [value missing]. ; Therefore, the aspect ratio obtained from the image is affected by both the phone's elevation angle and the antenna's downtilt angle. An iterative downtilt angle optimization algorithm based on formula (5) is designed. The dimensions are corrected by continuously solving the downtilt angle and antenna aspect ratio. The iterative process is as follows: Calculate the tilt angle by incorporating the aspect ratio information from close-up shots. Simultaneously, the phone's tilt angle is calculated from the phone's IMU data. Use formula (5) to obtain aspect ratio information. ,in It is the aspect ratio result of the previous iteration; Repeat the above steps until the preset number of iterations is reached to obtain the downslope angle. .

4. A method for measuring the downtilt angle of a plate antenna based on mobile phone information fusion, characterized in that, The downtilt angle measurement system for a plate antenna based on mobile phone information fusion as described in claim 1 is used to perform the following: Step S1: Take an image of the antenna from a preset distance using a mobile phone; Step S2: Process the image to eliminate distance-related parameters in the camera imaging principle; Step S3: If the antenna information is known, map the antenna bottom and side vectors onto the two-dimensional image coordinate system, and solve the antenna downtilt angle by fitting the bottom and side vectors on the image. If the antenna information is unknown, the antenna downtilt angle is solved by iterating through the antenna's aspect ratio. Step S4: Take an image of the chessboard grid aligned with the direction of gravity using a mobile phone, solve for the gravity vector on the image, and obtain the downward tilt angle after error correction; In step S4: By taking an image of a checkerboard pattern aligned with the direction of gravity using a mobile phone, determine the gravity vector on the image: Hang the chessboard grid vertically, ensuring the plumb line coincides with the grid lines within the chessboard grid. Record a video of a chessboard pattern and change the phone's orientation; Record IMU data, with the gravity vector denoted as... ; Solve for the camera intrinsic and extrinsic parameters using existing data. ; The gravity vector v in each image is solved using intrinsic and extrinsic parameters, coinciding with the y-axis of the image's physical coordinate system; Establish mapping ; A virtual pole-holding database is established through the above steps, and the image gravity direction corresponding to the gravity vector with the highest similarity at the time of shooting is used to replace the measured pole-holding vector; The specific implementation steps are as follows: During checkerboard calibration, the camera's intrinsic parameters are first calculated by extracting corner points from a portion of the checkerboard image. Then, using corner points obtained from other images, the camera's extrinsic parameters are calculated. Finally, the gravity vector is determined using these intrinsic and extrinsic parameters. The projection vectors onto the image are then stored in the database. When collecting IMU data, data is recorded before and after the photo is taken at preset times. The average value of the collected IMU data is then taken. The vectors in the virtual pole library are sorted according to their cosine similarity to the IMU data. Data with a similarity to the IMU data exceeding a preset value are filtered out, with a maximum of a preset value of data being filtered. The vectors on the image corresponding to the IMU data are then filtered out, removing vectors with a similarity to most other vectors below the preset value. The average value of the filtered vectors is then taken as the final virtual pole vector on the image.

5. The method for measuring the downtilt angle of a plate antenna based on mobile phone information fusion according to claim 4, characterized in that, In step S2: By approximating the imaging principle of a camera based on a pinhole camera model when shooting at a preset distance, distance-related parameters in the imaging principle are eliminated. Approximation means that in the image pixel coordinate system, the physical dimensions di and di of each pixel in the y-axis and y-axis directions are the same.

6. The method for measuring the downtilt angle of a plate antenna based on mobile phone information fusion according to claim 5, characterized in that, In step S3: Using the known physical dimensions of the antenna's base and sides, the antenna's base and side vectors are mapped onto a two-dimensional image coordinate system using the aforementioned approximated camera imaging principle. The antenna's downtilt angle is then solved by fitting the base and side vectors of the known image. The specific implementation steps are as follows: Explicitly mark the images captured by the mobile phone. The vector is obtained from the point corresponding to the point. ; Random initialization parameters include: downtilt angle , ， ， Roll angle ; The original coordinates are transformed according to the above approximate camera imaging model to obtain coordinates that are three-dimensional coordinates projected onto a two-dimensional pixel coordinate system. ; Minimize the vector formed after marking the graph using the optimizer. )and( The mean squared error between ) and the loss function Defined as: Search and update parameters, iterating repeatedly until the error is less than the preset error coefficient. .

7. The method for measuring the downtilt angle of a plate antenna based on mobile phone information fusion according to claim 4, characterized in that, In step S3: When the antenna model is unknown, the aspect ratio of the antenna is obtained by taking a picture of the front of the antenna. The aspect ratio of the antenna in the image is used to replace the actual antenna size ratio, and the antenna downtilt angle is solved by iterating the aspect ratio of the antenna.

8. The method for measuring the downtilt angle of a plate antenna based on mobile phone information fusion according to claim 7, characterized in that, In step S3: The farther the image is taken from the front of the antenna, the closer the antenna size ratio obtained from the image will be to the actual antenna size ratio; when the actual downtilt angle of the antenna is... At this time, the relationship between the aspect ratio of the image and the actual aspect ratio satisfies: (5) in, and This represents the projected lengths of the antenna's base and side edges onto the image plane. 𝐿 and 𝐻 represent the physical lengths of the antenna's base and side edges, respectively. The phone's elevation angle during shooting is [value missing]. ; Therefore, the aspect ratio obtained from the image is affected by both the phone's elevation angle and the antenna's downtilt angle. An iterative downtilt angle optimization algorithm based on formula (5) is designed. The dimensions are corrected by continuously solving the downtilt angle and antenna aspect ratio. The iterative process is as follows: Calculate the tilt angle by incorporating the aspect ratio information from close-up shots. Simultaneously, the phone's tilt angle is calculated from the phone's IMU data. Use formula (5) to obtain aspect ratio information. ,in It is the aspect ratio result of the previous iteration; Repeat the above steps until the preset number of iterations is reached to obtain the downslope angle. .

Images