Vision-based positioning phase interferometer static calibration method

Abstract

The present application relates to a visual positioning-based phase interferometer static calibration method, which is assisted by a calibration device including a camera, a feature plate, a positioning feature point, a signal source and a control computer; the control computer is connected with the camera and the signal source; the feature plate adopts a black-and-white checkboard, which is used for camera positioning and is placed near an antenna array of the phase interferometer; the area of the checkboard is determined by a calibration distance of the camera, and the distance d of each grid of the checkboard is known; the positioning feature point is placed on the surface of the antenna array and is used for calculating a baseline of the antenna array on a carrier platform; the camera adopts a digital camera; and the signal source uses a general microwave signal source.

Description

Technical Field

[0001] This invention relates to a static calibration method for phase interferometers based on visual positioning. Background Technology

[0002] In electronic reconnaissance, phase interferometer direction finding systems offer advantages such as wide bandwidth, high accuracy, and high sensitivity, while also being moderately sized, making them widely used on various airborne, shipborne, and vehicle-mounted platforms. The basic principle of phase interferometer direction finding is to determine the azimuth of a radiation source by analyzing the phase difference between signals received by multiple elements of a reconnaissance antenna array from the same radiation source. In engineering implementation, because the signal transmission paths from each antenna to the receiver and then to the processor in a phase interferometer direction finding system are not perfectly identical, inherent phase differences inevitably exist between the antennas and receiving channels. Therefore, in practical applications, calibration is necessary to eliminate these inherent phase differences to ensure the accuracy and high precision of the final direction finding results. Especially after the direction finding system has been developed and installed on an application platform for a period of time, changes in the equipment hardware (aging, component replacement, etc.) necessitate periodic or conditional calibration to maintain direction finding accuracy.

[0003] In the installed state, there are currently two main methods for calibrating a phase interferometer direction finding system:

[0004] One type is dynamic calibration, also known as internal source calibration. This involves using a signal source built into the phase interferometer direction finding equipment to generate a calibration signal, correcting phase inconsistencies between the receiver and processor for each channel. This is a built-in function of the equipment, easy to operate, and requires no additional testing equipment. However, since the calibration signal from the built-in signal source is generally injected from the receiver, this calibration method cannot correct phase inconsistencies in the channel from the antenna to the receiver. Therefore, dynamic calibration cannot achieve its correction purpose after the antenna, feeder, or receiver equipment ages or is replaced.

[0005] Another type is static calibration, also known as external radiation calibration. This involves using an independent signal transmitter to emit calibration signals, illuminating the reconnaissance equipment from outside, thus reconstructing the complete direction-finding process of the equipment. This allows for calibration of each channel from the antenna to the processor. Clearly, static calibration offers more comprehensive and accurate correction. Furthermore, according to the maintenance requirements of interferometer direction-finding equipment, static calibration must be performed after the equipment has been used for a period of time (generally several months), or after replacing antennas, feeders, receivers, or other components.

[0006] For existing field static calibration implementation solutions:

[0007] The main equipment for static calibration in the field includes a total station, a signal source, and a control computer. The traditional calibration process is as follows: Figure 1As shown. Currently, in field installation environments, static calibration of phase interferometer direction-finding systems is hampered by factors such as the attitude of the installation platform, ground flatness, and manual operation, resulting in low efficiency and limited accuracy in determining the normals of the phase interferometer direction-finding array. This invention, based on computer vision principles, uses visual positioning technology to automatically determine the normal positions of the interferometer array, thus replacing manual measurement methods. This achieves automatic, high-precision calibration of the phase interferometer direction-finding array normals in an installation environment, improving the overall efficiency and accuracy of static calibration of phase interferometer direction-finding systems under field conditions.

[0008] The first step is to determine the normal: use a total station to determine the normal OS of the phase interferometer antenna array.

[0009] The second step is to radiate a signal: at a sufficiently far distance, along the normal OS direction, place a signal source to radiate a calibration signal.

[0010] The third step is to generate a calibration table: measure the phase difference between the signals received by each element of the antenna array, which is taken as the inherent phase difference of the phase interferometer at that frequency. Then, change the frequency and measure the phase difference between each element channel at each frequency point to form a phase calibration table.

[0011] The fourth step is to inject the phase calibration table into the direction finding equipment to correct the direction finding results in real time and ensure the accuracy of the direction finding.

[0012] Currently, under the computer control of calibration, steps two through four can be completed efficiently and automatically by the equipment. The main bottleneck affecting the speed and accuracy of static calibration in the field lies in the first step, namely, measuring the normal of the phase interferometer antenna array.

[0013] Taking the most commonly used total station measurement method as an example, assume that fixed points A and B are two fixed points when the phase interferometer antenna array is installed. This is determined when the direction finding system is installed on the platform. The line connecting A and B represents the installation baseline of the interferometer array.

[0014] First, use the plumb line method to find the projection A'B' of points A and B on the ground, and find the center O' of A'B';

[0015] Then, use the total station to align with point O' on the ground to perform balancing, that is, the center of the total station should be aimed at point O', and at the same time adjust its own posture to a horizontal position.

[0016] Next, align the total station with the extension line of A'B and mark the orientation at this point as 0°.

[0017] Finally, by precisely rotating the total station by 90°, the normal OS was found.

[0018] Therefore, it can be seen that the process of determining the normal OS of a phase interferometer antenna array using a total station is essentially a manual measurement process. This has the following drawbacks:

[0019] First, it is inefficient: Under field conditions, operating a total station for projection, balancing, and aiming often requires repeated corrections, which is time-consuming and labor-intensive. In particular, for current practical direction finding systems, a phase interferometer array often has multiple baselines for resolving direction finding ambiguities. Therefore, it is necessary to mark the normals of all the baselines. The process of finding the normals is very cumbersome, resulting in a single calibration often taking tens of hours, or even hundreds of hours, which is inefficient.

[0020] Secondly, the accuracy is limited: Since the operation process is mainly done manually, the steps of projection, balancing, and aiming are linked one after another. An improper operation can easily introduce a large number of errors, which will accumulate in subsequent steps and eventually lead to a large deviation in the calibrated normal angle, exceeding the direction finding error range of the equipment and failing to meet the calibration accuracy requirements.

[0021] Third, the site is limited: Since the observation and calibration are carried out by projecting onto the ground, a relatively flat and wide site is required firstly, and secondly, the posture of the installation platform must be basically level with the ground; otherwise, the platform needs to be leveled before installation. Summary of the Invention

[0022] In general, the technical problem to be solved by this invention is to provide a visual positioning-based static calibration method for phase interferometers. This method uses a regular camera instead of a total station and, based on computer vision and mapping geometry techniques, reconstructs the three-dimensional coordinates of the phase interferometer array on a platform in the external environment. On this basis, visual positioning technology is used to automatically determine the normal position of the interferometer array, thereby replacing the manual measurement method using a total station. This achieves automatic and high-precision calibration of the phase interferometer normal, thus improving the overall efficiency and accuracy of static calibration under external conditions.

[0023] To solve the above problems, the technical solution adopted by the present invention is as follows:

[0024] The calibration equipment used in this invention includes a camera, a feature plate, positioning feature points, a signal source, and a control computer; the control computer is connected to the camera and the signal source; the feature plate uses a black and white checkerboard pattern for camera positioning and is placed near the phase interferometer antenna array, the area of ​​the checkerboard pattern is determined by the camera's calibration distance, and the spacing d between each checkerboard cell is known; the positioning feature points are placed on the surface of the antenna array and are used to calculate the baseline of the antenna array on the vehicle platform; the camera is a digital camera; and the signal source is a general-purpose microwave signal source.

[0025] The control computer, a processing center equipped with visual positioning processing and automatic calibration software, interacts with the camera to acquire photos taken by the camera in real time to reconstruct the three-dimensional coordinates, while dynamically calculating the current camera position and alignment error, and indicating the alignment correction direction; the control computer completes the cross-linking with the phase interferometer direction finding equipment, controls the microwave signal source to work, and automatically calculates and generates a phase calibration table for the entire frequency band.

[0026] The camera, signal source, and control computer are integrated on a calibration cart with small wheels;

[0027] The calibration method is as follows:

[0028] S1. Construct a static calibration environment. First, construct the calibration environment under outdoor conditions, place the 3D reconstruction feature plate and locate the feature points, and connect the camera, signal source, and control computer; then connect the control computer to the direction finding equipment of the phase interferometer under test.

[0029] S2, Visual Reconstruction of 3D Coordinates: First, a camera is used to take several photos and / or videos of the phase interferometer array from different angles. The photos and / or videos contain feature plates and localization feature points. Then, based on computer vision algorithms, the 3D coordinates of the phase interferometer array on the vehicle platform in the field environment are automatically reconstructed.

[0030] S3, calculate the perpendicular bisector of the interferometer array baseline in the reconstructed three-dimensional space, including the normal to be calibrated;

[0031] S4 uses a visual positioning algorithm to dynamically indicate the angle between the current camera shooting position and the vertical plane, and uses a calibration trolley to guide and correct the shooting position to 0°, thereby completing the alignment of the normal OS of the phase interferometer antenna array.

[0032] S5, under the automatic control of the control computer, the signal source on the calibration trolley at the normal position radiates the calibration signal, compares it with the phase difference obtained by each antenna of the direction finding device of the phase interferometer under test, and automatically generates a phase calibration table, thereby completing the static calibration.

[0033] In step S2, step S2.1 is executed, where the camera captures an image at the set position P1, completely capturing the feature plate and feature points pre-placed at the phase interferometer direction-finding antenna array. Since the feature plate and positioning feature points both use a checkerboard pattern with alternating light and dark areas, a checkerboard corner detection algorithm is used to detect all feature corner points on the feature plate and positioning feature points, calculate the two-dimensional coordinates of the corner points in the image, and arrange them in order from left to right and from top to bottom to ensure that the index of the same corner point is consistent in different images. Corner points with consistent indices are called paired points.

[0034] S2.2, Perform the feature plate 3D coordinate initialization step;

[0035] First, let the first intersection point at the top left corner of the feature plate be the origin of the world coordinate system O(0, 0, 0), and the feature plate be defined on the XOY plane. Since the corner spacing d is known, the coordinates of the corner point in the m-th row and n-th column in the world coordinate system are [(n-1)·d, (m-1)·d, 0]. This determines the three-dimensional spatial coordinates of all corner points on the 3D reconstruction feature plate in the world coordinate system.

[0036] S2.3, Calculate the camera pose. First, when the camera takes a picture from point P1, it is assumed that the camera's coordinate system (P1-x1y1z1) undergoes rigid body motion relative to the world coordinate system (O-XYZ). Then, according to the principle of rigid body motion in three-dimensional space, the motion between the two coordinate systems is defined as a rotation plus a translation. The three-dimensional coordinates of the corner point N of the 3D reconstructed feature plate in the world coordinate system are known, denoted as N = [X, Y, Z]. T In the P1 camera coordinate system (P1-x1y1z1), it is denoted as N1=[x n1 y n1 , z n1 ] T ;

[0037] Then the coordinates of N1 are: N1 = RN + t;

[0038] Where R is a 3×3 rotation matrix and t is a 3×1 translation matrix, R and t represent the camera's position and pose, i.e., the camera pose; secondly, defining the augmented matrix [R|t] as a 3×4 matrix, we can express equation ①:

[0039]

[0040] The camera pose [R|t] is unknown. The camera pose [R|t] at point P1 is determined by capturing corner images on the 3D reconstruction board. N is imaged as n1 in P1. Let the two-dimensional coordinates of n1 be n1 = [u n1 v n1 ] T ;

[0041] Secondly, assuming the camera uses a pinhole camera model, if f is the focal length of the camera, then:

[0042]

[0043] That is, z n1 n1 = KN1;

[0044] Where K is the camera's intrinsic parameter matrix, which is the preset coefficients for all images;

[0045] From equation ②, we can obtain:

[0046] Then, substituting equation ③ into equation ① and expanding it, we obtain two constraints:

[0047]

[0048] Among them, the camera pose augmentation matrix [R|t] at point P1 has 12 dimensions. The camera intrinsic parameters K and the three-dimensional coordinates (X, Y, Z) of the corner points on the 3D reconstruction feature plate are known. Therefore, by providing the constraint of Equation ④, the matrix [R|t] can be solved with 6 feature corner points. When there are more than 6 points, the SVD method is used to find the least square solution of the equation.

[0049] Next, after solving for the rotation matrix R and translation matrix t, the position and orientation of the camera at point P1 are determined. Then, the same method is used to determine the position and orientation of the camera at all subsequent shooting points (pn, n = 2, 3, 4…).

[0050] S2.4, Calculation of 3D coordinates of positioning feature points: After determining the camera poses at two shooting points P1 and P2 in S2.3, the 3D coordinates of all positioning feature points in space are determined by triangulation. First, positioning feature point A is selected, where a1 and a2 are the pixel coordinates of this positioning feature point in the images captured by the camera at positions P1 and P2. Then, all points on ray P1A are projected onto the same pixel point a1. When the spatial position of A is unknown, ray P2d2 intersects ray P1a1 at point A, thus inferring the spatial position of A and realizing the 3D reconstruction of the positioning feature points.

[0051] S2.5: When the error in calculating the least squares solution of the image exceeds a set threshold, bundle adjustment is performed. The calculated coordinates of point A are projected onto camera P. i The two-dimensional coordinates obtained are a i However, the actual imaging coordinates of point A in the image are a′. i , when a i With a′ i When they do not coincide, a i With a′ i The distance between two points is called the reprojection error. Clustering adjustment minimizes the reprojection error of all images by adjusting the camera pose and the coordinates of point A. Ideal results are obtained when the number of images M reaches 10.

[0052] In S3-S4, visual positioning guides alignment;

[0053] In S3, the perpendicular bisector of the interferometer antenna array is measured. After obtaining the three-dimensional coordinates of all positioning feature points, a straight line l is fitted in space using these points, and a plane perpendicular to the straight line is obtained. This plane is the perpendicular bisector of the phase interferometer antenna array. Let the perpendicular bisector intersect the straight line l at point C.

[0054] In S4, the alignment is dynamically guided by the calibration position deviation. First, in the camera pose, t is considered to be the coordinates of the camera center P in space at the current shooting position. Given the coordinates of point P and point C, the direction vector t of the line connecting the camera center P and point C is obtained by subtracting the coordinates of point P and point C. pc The current calibration position is the current shooting position, and the angle between it and the perpendicular bisector is the vector t. pc Let θ be the angle between the perpendicular bisector and the perpendicular plane.

[0055] Then, when there is an offset angle θ, the current camera position is dynamically adjusted by the magnitude and sign of the offset angle, and the shooting position is equal to the calibration position, until the offset angle θ < ±δ alignment accuracy, thus completing the phase interferometer antenna array normal measurement and alignment.

[0056] The advantages of this invention are as follows:

[0057] First, calibration time is significantly reduced. Using a fully automated 3D reconstruction method, the normal direction of the phase interferometer array antenna on the field platform can be quickly determined by taking a dozen photos (or a short video). Simultaneously, the location and movement of the instrument can be determined through visual positioning, allowing for dynamic adjustment of the radiation alignment with the normal. This avoids the time-consuming and labor-intensive process of manual projection, balancing, and aiming, which is a common practice in traditional total station methods. The calibration time is reduced from tens of hours to less than half an hour, greatly improving calibration efficiency.

[0058] Secondly, the calibration accuracy is guaranteed. First, the normal measurement and deviation indication are completed automatically, eliminating human error. Second, the normal alignment accuracy is determined by the accuracy of 3D reconstruction and visual positioning. Strictly speaking, this is an optical measurement method. Based on the correct calibration of the camera's intrinsic parameters, the normal alignment accuracy is less than 0.1° (experimental value), which meets the calibration requirements.

[0059] Third, the requirements for personnel operation are not high. The calibration process only requires the operator to take clear photos or videos of the feature plate and feature points from multiple angles, and then complete the normal alignment under the guidance of the system; the operator no longer needs to master complex measurement operations such as total station balancing and aiming; the ability requirements for calibration personnel are reduced, which is conducive to the use and maintenance personnel of phase interferometer direction finding equipment mastering the calibration method and reducing maintenance costs.

[0060] Fourth, it has low requirements for the condition of the site and platform. Even when the site is uneven or the platform on which the antenna array is installed is tilted or rolling, it does not affect the visual reconstruction of the camera images. It can still correctly reconstruct the three-dimensional coordinates of the antenna array without the need for laborious work such as traction or leveling the platform. Therefore, the invention has a wider range of application scenarios.

[0061] Fifth, the cost of calibration equipment is not high. The calibration system can be upgraded by replacing the total station in the traditional static calibration system with a camera and deploying visual positioning software on the control computer. Attached Figure Description

[0062] Figure 1 This is a diagram of the existing phase interferometer field static calibration scheme.

[0063] Figure 2 This is a diagram of the field static calibration scheme of the present invention.

[0064] Figure 3 This is a flowchart illustrating the implementation steps of the present invention.

[0065] Figure 4 This is a schematic diagram of the three-dimensional coordinates of the visual reconstruction antenna array of the present invention.

[0066] Figure 5 This is a schematic diagram of the corner coordinates of the three-dimensional reconstruction feature plate of the present invention.

[0067] Figure 6 This is a schematic diagram of the clustering adjustment of the present invention.

[0068] Figure 7 This is a schematic diagram of the visual positioning guidance alignment of the present invention. Detailed Implementation

[0069] based on Figure 1-7 The calibration equipment used in this invention includes a camera 3, a feature plate 4, a signal source 1, a control computer 2, a camera interferometer antenna array 5, and positioning feature points 6; the overall implementation scheme is as follows: Figure 2 As shown. Feature plate 4 is a 3D reconstruction feature plate.

[0070] exist Figure 2 There are two types of feature plates. One type is a black and white checkerboard pattern, which is placed near the phase interferometer antenna array and is less than 1 square meter in size (determined by the calibration distance). The spacing d between each grid is known and is used for visual reconstruction of the three-dimensional coordinates of space. The other type is a positioning feature point, which is placed at the antenna array positioning point and is used for visual positioning of the baseline of the antenna array on the vehicle platform.

[0071] A standard digital camera is used, with performance sufficient to capture clear images of the feature plate and feature points at the calibration distance. A conventional, general-purpose microwave signal source is used, covering the operating frequency band and is programmable. The control computer serves as the processing center for the visual positioning and automatic calibration software. Its functions include: 1) interacting with the camera, acquiring real-time photos (videos) to reconstruct 3D coordinates, dynamically calculating the current camera position and alignment error, and indicating the alignment correction direction; and 2) connecting with the phase interferometer direction-finding equipment, controlling the microwave signal source, and automatically calculating and generating a full-band phase calibration table.

[0072] The specific steps for implementing this invention are shown in Figure 3.

[0073] The first step is to construct a static calibration environment. Under outdoor conditions, the calibration environment is constructed by placing the 3D reconstruction feature plate and positioning feature points, and connecting the camera, signal source, and control computer (these three can be integrated into a small calibration cart with wheels); then the control computer is connected to the direction finding equipment of the phase interferometer under test.

[0074] The second step involves visually reconstructing the 3D coordinates by taking multiple photos or videos of the phase interferometer array from different angles using a camera (including feature plates and positioning feature points); then, based on computer vision algorithms, the 3D coordinates of the phase interferometer array on the outdoor environment download platform are automatically reconstructed.

[0075] The third step is to calculate the perpendicular bisector of the interferometer array baseline (including the normal to be calibrated) in the reconstructed three-dimensional space.

[0076] The fourth step involves using a visual positioning algorithm to dynamically indicate the angle between the current camera shooting position (calibration trolley position) and the vertical plane, thereby guiding the rapid correction of the shooting position (calibration trolley) to 0°, thus completing the alignment of the normal OS of the phase interferometer antenna array.

[0077] The fifth step involves the automatic control of a computer, where a signal source on a calibration trolley at the normal position radiates a calibration signal. This signal is then compared with the phase difference acquired by each antenna of the direction-finding device of the interferometer under test, automatically generating a phase calibration table to complete the static calibration.

[0078] The following is a detailed analysis of two key technologies in the implementation process:

[0079] like Figure 4 As shown, both images taken by the camera at positions P1 and P2 completely captured the feature plate and feature points pre-placed at the direction-finding antenna array of the phase interferometer. Since the feature plate and feature points are composed of alternating light and dark grids, resembling a chessboard, a chessboard corner detection algorithm is used to detect all feature corner points on the feature plate and the located feature points, calculate the two-dimensional coordinates of the corner points in the image, and arrange them in order from left to right and top to bottom to ensure that the index of the same corner point is consistent in different images. Corner points with consistent indices are called paired points.

[0080] (1) Initialization of three-dimensional coordinates of feature plate

[0081] Since the feature plate is a plane, we can set the first intersection point at the top left corner as the origin of the world coordinate system O(0, 0, 0), and the feature plate lies in the XOY plane, as shown below. Figure 5As shown. The corner spacing d is known, and the coordinates of the corner point in the m-th row and n-th column in the world coordinate system are [(n-1)·d, (m-1)·d, 0]. Therefore, the three-dimensional spatial coordinates of all corner points on the 3D reconstruction feature plate in the world coordinate system can be determined.

[0082] (2) Camera pose calculation

[0083] When the camera takes a picture from point P1, it can be assumed that the camera's coordinate system (P1-x1y1z1) undergoes rigid body motion relative to the world coordinate system (O-XYZ). According to the principle of rigid body motion in three-dimensional space, the motion between two coordinate systems can consist of a rotation and a translation.

[0084] For example, the 3D coordinates of a corner point N on a 3D reconstructed feature plate are known in the world coordinate system, denoted as N = [X, Y, Z]. T In the P1 camera coordinate system (P1-x1y1z1), it is denoted as N1=[x n1 y n1 , z n1 ] T .

[0085] The coordinates of N1 are:

[0086] N1 = RN + t

[0087] Where R is a 3×3 rotation matrix and t is a 3×1 translation matrix, R and t represent the camera position and pose (or simply camera pose). Defining the augmented matrix [R|t] as a 3×4 matrix, we can then derive the following formula:

[0088]

[0089] At this point, the camera pose [R|t] is unknown. We can determine the camera pose [R|t] at point P1 using the corner images captured on the 3D reconstruction board. N is imaged as n1 in P1. Let the two-dimensional coordinates of n1 in image 1 be n1 = [u...]. n1 v n1 ] T .

[0090] Based on the pinhole camera model, if f is the focal length of the camera, then:

[0091]

[0092] Right now

[0093] z n1 n1=KN1

[0094] Where K is the camera's intrinsic parameter matrix. The camera's intrinsic parameters are fixed after it leaves the factory and will not change during use. Some camera manufacturers will directly label the camera's intrinsic parameters, or they can use mature calibration algorithms to obtain the camera's intrinsic parameters (such as the Zhang Zhengyou calibration method). Therefore, K can be considered known.

[0095] From equation ②, we can obtain:

[0096]

[0097] Substituting equation ① and expanding it yields two constraints:

[0098]

[0099] The camera pose augmentation matrix [R|t] at point P1 has 12 dimensions. The camera intrinsic parameters K and the 3D coordinates (X, Y, Z) of the corner points on the 3D reconstruction feature plate are known, thus providing two of the aforementioned constraints. Therefore, matrix [R|t] can be solved using 6 feature corner points. For more than 6 points, methods such as SVD are used to find the least-squares solution to the equation.

[0100] After solving for the rotation matrix R and the translation matrix t, the position and orientation of the camera at point P1 can be determined. Similarly, the position and orientation of the camera at point P2 and any other shooting point can be determined.

[0101] (3) Calculation of three-dimensional coordinates of positioning feature points

[0102] Once the camera poses at shooting points P1 and P2 are determined, the 3D coordinates of any other localized feature point in space not on the 3D reconstruction feature plate can be determined using triangulation. Figure 4 Taking a localized feature point A as an example, a1 and a2 are the pixel coordinates of this feature point in the images captured by the camera at positions P1 and P2. From the first image, all points on ray P1A will project onto the same pixel a1. If the location of A is unknown, we can see from the second image that ray P2a2 intersects with ray P1a1 at point A, thus allowing us to infer the spatial location of A and achieve 3D reconstruction of the localized feature point.

[0103] This process is sensitive to noise, and the error is relatively large when calculating the least squares solution from two images. Therefore, after calculating the 3D coordinates of the localized feature points using P1 and P2, the 3D coordinates of the localized feature points are usually optimized using subsequent M (M>10) images. This method is called bundle adjustment. Figure 6 As shown.

[0104] The calculated coordinates of point A are projected onto camera P. i The two-dimensional coordinates obtained are a i However, the actual imaging coordinates of point A in the image are a′.i a i With a′ i Typically, the points do not coincide; the distance between two points is called the reprojection error. Bundle adjustment minimizes the reprojection error of all cameras by adjusting the camera pose and the coordinates of point A. Therefore, locating feature points and camera pose is a continuous optimization process. The more images captured, the smaller the error. A stable result can be obtained when the number of images M reaches 10.

[0105] Regarding visual positioning guidance alignment

[0106] (1) Measurement of the vertical plane of the interferometer antenna array

[0107] After obtaining the three-dimensional coordinates of all the positioning feature points, a straight line l is fitted in space using these points, and a plane perpendicular to the straight line is found. This plane is the perpendicular bisector of the phase interferometer antenna array. Let the perpendicular bisector intersect the straight line l at point C.

[0108] (2) Dynamic guidance alignment based on calibration position deviation

[0109] In camera pose, t can be considered as the coordinates of the camera center P in space at the current shooting position. Given the coordinates of points P and C, the direction vector t of the line connecting the camera center P and point C can be obtained by subtracting the two coordinates. pc The current calibration position is the current shooting position, and the angle between it and the perpendicular bisector is the vector t. pc Let θ be the angle between the perpendicular bisector and the perpendicular plane.

[0110] With the offset angle θ, the current camera position (shooting position = calibration position) can be dynamically adjusted by the magnitude and sign of the offset angle until the offset angle θ < ±δ (alignment accuracy), thus completing the determination and alignment of the phase interferometer antenna array normal.

[0111] The present invention has been described in detail for the purpose of making the disclosure clearer, and the prior art will not be listed in detail.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. It is obvious to those skilled in the art that multiple technical solutions of the present invention can be combined. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. All technical contents not described in detail in the present invention are well-known technologies.

Claims

1. A static calibration method for a phase interferometer based on visual positioning, characterized in that: Includes the following steps, S1. Construct a static calibration environment. First, construct the calibration environment under outdoor conditions, place the 3D reconstruction feature plate and locate the feature points, and connect the camera, signal source, and control computer; then connect the control computer to the direction finding equipment of the phase interferometer under test. S2, Visual reconstruction of 3D coordinates: First, a camera is used to take several photos and / or videos of the phase interferometer array from different angles. The photos and / or videos contain feature plates and localization feature points. Then, based on computer vision algorithms, the 3D coordinates of the phase interferometer array on the vehicle platform in the field environment are automatically reconstructed. S3, calculate the perpendicular bisector of the interferometer array baseline in the reconstructed three-dimensional space, including the normal to be calibrated; S4 uses a visual positioning algorithm to dynamically indicate the angle between the current camera shooting position and the vertical plane, and uses a calibration trolley to guide and correct the shooting position to 0°, thereby completing the alignment of the normal OS of the phase interferometer antenna array. S5, under the automatic control of the control computer, the signal source on the calibration trolley at the normal position radiates the calibration signal, compares it with the phase difference obtained by each antenna of the direction finding device of the phase interferometer under test, and automatically generates a phase calibration table, thereby completing the static calibration.

2. The visual positioning-based static calibration method for phase interferometers according to claim 1, characterized in that: The calibration equipment includes a camera, feature plate, signal source, and control computer; The control computer is electrically connected to the camera and signal source; positioning feature points are set on the phase interferometer antenna array; The feature plate uses a black and white checkerboard pattern and is placed near the phase interferometer antenna array. The area of ​​the checkerboard pattern is determined by the camera calibration distance. The spacing d of each checkerboard pattern is known and is used for visual reconstruction of three-dimensional space and camera positioning. The positioning feature points are placed on the surface of the antenna array to calculate the baseline of the antenna array on the visual positioning vehicle platform; The camera used is a digital camera; the signal source uses a general-purpose microwave signal source. The control computer, a processing center equipped with visual positioning processing and automatic calibration software, interacts with the camera to acquire photos taken by the camera in real time to reconstruct three-dimensional coordinates. At the same time, it dynamically calculates the current camera position and the angle error between the camera and the antenna baseline, indicating the alignment correction direction. The control computer completes the cross-linking with the phase interferometer direction finding equipment, controls the microwave signal source to work, and automatically calculates and generates a phase calibration table for the entire frequency band. The camera, signal source, and control computer are integrated on a calibration cart with small wheels.

3. The visual positioning-based static calibration method for phase interferometers according to claim 1, characterized in that: in, In S2, S2.1 is executed, and the camera is in the set... The image was captured at the location, completely capturing the feature plate and feature points pre-placed at the direction-finding antenna array of the phase interferometer. Since the feature plate and the positioning feature points both use a checkerboard pattern with alternating light and dark areas, a checkerboard corner detection algorithm was used to detect all feature corner points on the feature plate and the positioning feature points, calculate the two-dimensional coordinates of the corner points in the image, and arrange them in order from left to right and from top to bottom to ensure that the index of the same corner point is consistent in different images; among them, corner points with consistent indices are called paired points. S2.2, Perform the feature plate 3D coordinate initialization step; First, let the first intersection point at the top left corner of the feature plate be the origin of the world coordinate system. Furthermore, the feature plate is defined on the XOY plane, where the corner spacing is... Given that, the coordinates of the corner point in the m-th row and n-th column in the world coordinate system are: This allows us to determine the three-dimensional spatial coordinates of all corner points on the three-dimensional reconstruction feature plate in the world coordinate system. S2.3, Calculate camera pose. First, when the camera moves from... When taking a picture, the camera's coordinate system is assumed to be... Relative to the world coordinate system Rigid body motion occurred; then, according to the principle of rigid body motion in three-dimensional space, the motion between the two coordinate systems is defined as a rotation plus a translation, and the corner points of the 3D reconstructed feature plate are obtained. N The three-dimensional coordinates in the world coordinate system are known, denoted as . ;exist Camera coordinate system Below, recorded as ; but The coordinates are: ; Where R is a 3×3 rotation matrix and t is a 3×1 translation matrix, R and t represent the camera's position and pose, i.e., the camera pose; secondly, the augmented matrix [R|t] is defined as a... For a matrix, then equation ① is given: Formula ①; The camera pose [R|t] is unknown and is determined by capturing corner images on the 3D reconstruction board. The pose of the point camera [R|t], N is in Medium imaging is ,set up The two-dimensional coordinates are ; Secondly, the camera uses a pinhole camera model, if Let the focal length of the camera be: Formula ②; Right now = ; in, K is the camera's intrinsic parameter matrix, which contains the preset coefficients for all images; From equation ②, we can obtain: Formula ③; Then, substituting equation ③ into equation ① and expanding it, we obtain two constraints: Formula ④; Among them, the one to be sought The camera pose augmentation matrix [R|t] has 12 dimensions, and the camera intrinsic parameters are... K and the three-dimensional coordinates of the corner points on the three-dimensional reconstruction feature plate ( Given that Equation ④ provides the constraint, the matrix [R|t] can be solved using 6 feature corner points. When there are more than 6 points, the SVD method is used to find the least squares solution of the equation. Then, after solving for the rotation matrix R and the translation matrix t, the position of the camera is determined. The position and orientation of the camera at each shooting point are determined; then, the same method is used to determine the position and orientation of the camera at all subsequent shooting points (pn, n=2,3,4…); S2.4, Calculation of the three-dimensional coordinates of the location feature points, which was determined in S2.

3. and After posing the cameras at the two shooting points, the three-dimensional coordinates of all spatial positioning feature points are determined using triangulation. First, positioning feature point A is selected. and It's the camera. and In the image captured at the location, the pixel coordinates of the localized feature point; then, the ray... All points on the surface are projected onto the same pixel. When the spatial location of A is unknown, then the ray With rays Intersecting at point A, the spatial location of A is inferred, and the three-dimensional reconstruction of the localized feature points is achieved; S2.5, when the error in finding the least squares solution for the image exceeds a set threshold, bundle adjustment is performed, utilizing the method in S2.

4. and After calculating the three-dimensional coordinates of the localized feature points, the three-dimensional coordinates of the localized feature points are optimized using the subsequent M images, where M>5; S2.6, Project the calculated coordinates of point A onto the camera. The two-dimensional coordinates obtained are However, the actual imaging coordinates of point A in the image are... ,when and When they do not overlap, and The distance between two points is called the reprojection error. Clustering adjustment minimizes the reprojection error of all cameras by adjusting the camera pose and the coordinates of point A. Ideal results are obtained when the number of images M reaches 10.

4. The visual positioning-based static calibration method for phase interferometers according to claim 1, characterized in that: In S3-S4, visual positioning guides alignment; In S3, the perpendicular bisector of the interferometer antenna array is measured, and the three-dimensional coordinates of all positioning feature points are obtained. Then, a straight line is fitted in space using these points. Find the plane perpendicular to the line, which is the perpendicular bisector of the phase interferometer antenna array. They intersect at point C.

Images