Vehicle-mounted multi-camera image position error detection method and device and medium

Abstract

The invention discloses a vehicle-mounted multi-view camera image position error detection method and device and a medium, and the method comprises the steps: controlling a vehicle-mounted multi-view camera to shoot an image card, obtaining actual measurement coordinates, comparing the distance between the actual measurement coordinates of each camera and the theoretical coordinates of a preset feature point, and calculating the self error; and relative operation is carried out on self errors of the reference camera and other cameras, so that system errors among the cameras of the vehicle-mounted multi-view camera can be obtained. According to the vehicle-mounted multi-camera image position error detection method and device and the medium, the actually measured coordinates and the preset theoretical coordinates are compared and calculated, the error of a single camera can be directly quantified, and interference of other cameras is eliminated; meanwhile, the reference camera is introduced, relative operation is carried out on self errors of the reference camera and self errors of other cameras, interference of common deviation can be counteracted, and the problem that self image position errors of each single camera in the vehicle-mounted multi-view camera and system errors among the cameras cannot be accurately detected can be solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a method, device, and medium for detecting image position errors in vehicle-mounted multi-view cameras. Background Technology

[0002] With the continuous development of automotive camera technology, the application of multi-camera systems is becoming increasingly widespread. In scenarios where multiple cameras work collaboratively after assembly, image positional deviations directly affect the effectiveness of subsequent image calibration and practical applications. Therefore, there is an urgent need for a solution that can identify the image position and inter-image deviations of automotive multi-camera systems to address the impact of these issues. Existing automotive camera image position testing technologies mainly focus on individual camera modules, with OC (Open Position) and COD (Cost Per Dimension) as the core testing metrics. This technology focuses on detecting image position-related parameters of a single module, acquiring image position data of a single module through specific testing methods, thereby evaluating the image position performance of a single camera module and meeting the image position testing needs at the single-module level.

[0003] However, existing vehicle camera image position testing technologies only target individual camera modules, using OC and COD as core test indicators, and can only achieve image position performance evaluation at the single module level. This technology cannot cover scenarios involving assembled multi-camera systems; it cannot accurately detect the individual image position errors of each camera after assembly, nor does it have effective means to capture systematic errors between cameras, making it difficult to meet the accuracy testing requirements of multi-camera collaborative applications. Summary of the Invention

[0004] This invention provides a method, device, and medium for detecting image position errors in vehicle-mounted multi-view cameras, in order to solve the problem that the image position errors of each individual camera in a vehicle-mounted multi-view camera, as well as the system errors between cameras, cannot be accurately detected.

[0005] To achieve the above objectives, this application provides a method for detecting image position errors using a vehicle-mounted multi-view camera, comprising: Control the vehicle-mounted multi-view camera to capture images of the map card, and extract the measured coordinates of the center point and circumferential feature points from the images captured by each camera; The measured coordinates of each camera are compared with the theoretical coordinates of the corresponding preset feature points to calculate the distance, thus obtaining the error of each camera. The system error between the cameras in the vehicle-mounted multi-camera system is obtained by performing a relative calculation between the error of the reference camera and the errors of the other cameras.

[0006] This invention uses a map as a unified reference, controlling multiple vehicle-mounted cameras to simultaneously capture images of the map. It extracts the measured coordinates of the center point and circumferential feature points in the images captured by each camera, providing a precise measurement data foundation for error detection. Compared to random shooting without a reference, the fixed positions of the feature points on the map, along with their center and circumferential features, comprehensively reflect the image position information, avoiding measurement data deviations caused by complex shooting scenes and unclear feature points, thus ensuring the targeted nature of individual camera image position detection. Secondly, by comparing the measured coordinates of each camera with the theoretical coordinates of corresponding preset feature points, the imaging deviation of a single camera, i.e., its own error, can be directly quantified. This comparison method focuses on the difference between the measured and theoretical coordinates of a single camera, eliminating interference from other cameras and achieving precise localization of the image position error of each individual camera, solving the problem of the difficulty in independently quantifying the error of a single camera. Furthermore, this invention introduces a reference camera as a benchmark, performing relative calculations on its own error and the errors of other cameras. By utilizing the fixed reference function of the reference camera, the common deviations that may exist in the individual camera's own error are offset, and the relative deviations between cameras, i.e., system errors, are accurately extracted. Since system errors are essentially relative positional deviations between cameras, rather than absolute deviations of individual cameras, this relative calculation can effectively remove the interference of individual camera's own error, achieving accurate detection of system errors between cameras. This simultaneously solves the core problem of the inability to accurately detect both types of errors.

[0007] Compared to existing technologies, this invention compares and calculates measured coordinates with preset theoretical coordinates, which can directly quantify the error of a single camera, eliminate interference from other cameras, and achieve accurate positioning and independent quantification of the error. At the same time, a reference camera is introduced, and its own error is calculated relative to the own errors of other cameras. This can cancel common deviations, eliminate interference from single camera errors, and accurately extract the relative deviation between cameras, i.e., the system error. Therefore, it can solve the problem of the inability to accurately detect the image position error of each individual camera in a vehicle-mounted multi-camera system, as well as the system error between cameras.

[0008] As a preferred embodiment, the chart is a five-point chart, including a center point and four circumferential feature points, and the distance between any two of the four circumferential feature points is the first distance.

[0009] This preferred solution employs a five-point chart, with one center point and four circumferential feature points. The distance between any two of the four circumferential feature points is fixed as the first distance. This simplifies the feature point extraction process and establishes a stable coordinate reference system through symmetrically distributed feature points. The center point serves as the core benchmark for quickly locating the image coordinate origin, while the four circumferential feature points comprehensively cover different orientations of the image, effectively capturing horizontal and vertical positional deviations during camera capture and reducing error interference from single feature point extraction. The fixed first distance provides clear parameter support for the accurate calculation of the theoretical coordinates of subsequent preset feature points, making the comparison between measured and theoretical coordinates more targeted, improving the accuracy of error calculation for each camera, and thus laying a solid foundation for subsequent system error calculation. This approach balances detection accuracy and efficiency, adapting to the actual detection scenario requirements of vehicle-mounted multi-view cameras.

[0010] As a preferred embodiment, the method for obtaining the theoretical coordinates of the preset feature points is as follows: The camera installed at the system reference position in the preset vehicle-mounted multi-view camera is set as the preset reference camera. The center point coordinates of the five-point map card are set as the theoretical origin coordinates of the preset reference camera. Combined with the first distance, the coordinates of the four circumferential feature points are calculated according to the four-directional symmetrical preset distance arrangement rules based on the origin. The theoretical origin coordinates and the coordinates of the four circumferential feature points constitute the reference coordinate set. Obtain the optical axis center of the preset reference camera and other preset cameras in the preset vehicle-mounted multi-view camera; Based on the optical axis center, the horizontal and vertical optical axis errors between the preset reference camera and other preset cameras are measured. Based on the coordinates of each feature point in the reference coordinate set, and combined with the horizontal and vertical optical axis errors, the theoretical coordinate sets of preset feature points for other preset cameras are calculated; wherein, all coordinates in the reference coordinate set and the theoretical coordinate sets of preset feature points for other preset cameras together constitute the theoretical coordinates of the preset feature points.

[0011] This preferred scheme uses the center point of a five-point chart as the theoretical origin of the preset reference camera, combined with a fixed-distance reference coordinate set of circumferential feature points, to ensure the uniformity and standardization of the reference coordinates and reduce subjective errors in coordinate setting. By obtaining the optical axis center of each camera, the horizontal and vertical errors of the optical axes of the preset reference camera and other preset cameras are measured, and these errors are incorporated into the calculation of the theoretical coordinates of other cameras. This ensures that the theoretical coordinates fit the actual installation position of the cameras, avoiding the problem of theoretical coordinates being out of sync with the actual shooting scene due to ignoring camera installation deviations, and ensuring that the comparison between theoretical coordinates and measured coordinates is more targeted. Integrating the theoretical coordinates of all cameras forms a complete theoretical coordinate system of preset feature points, which can accurately support the calculation of its own errors and system errors, improving the accuracy and reliability of the entire detection method and providing strong support for the accurate calibration of vehicle-mounted multi-view cameras.

[0012] As a preferred embodiment, the measured coordinates include the measured coordinates of the center point, the measured coordinates of the upper left point, the measured coordinates of the upper right point, the measured coordinates of the lower left point, and the measured coordinates of the lower right point.

[0013] In this preferred solution, the measured coordinates cover the center point and four circumferential measured coordinates, precisely corresponding to the feature point layout of the five-point chart. This comprehensively captures the positional deviations of the camera-captured images, avoiding omissions of local deviations due to missing measured point coordinates, and ensuring that the error calculation fully reflects the camera's shooting accuracy. The five measured point coordinates correspond to the center and surrounding areas of the chart, simultaneously capturing both center and edge deviations from the camera's images. This reduces the bias caused by measured data from a single area, improving the comprehensiveness and accuracy of the error calculation. The clear division of measured point coordinates provides a clear measurement object for subsequent error calculations, avoiding confusion during the calculation process, standardizing the testing procedure, improving the standardization and operability of the testing method, and adapting to the high-precision testing needs of vehicle-mounted multi-view cameras.

[0014] As a preferred embodiment, the self-error includes horizontal image error, vertical image error, and self-image rotation error, and the self-image rotation error is obtained as follows: For the first captured image corresponding to the first camera in the vehicle-mounted multi-view camera, taking the measured coordinates of the upper left point as the calculation object, based on the first distance, the square of the distance from the measured coordinates of the upper left point to the theoretical coordinates of the center point is calculated to obtain a first value, and the square of the distance from the measured coordinates of the upper left point to the theoretical coordinates of the upper left point is calculated to obtain a second value. Based on the first distance, the first value, and the second value, the angle contribution value is calculated using the inverse cosine theorem; wherein, the angle contribution value is the angle between the line connecting the measured coordinates of the upper left point and the theoretical coordinates of the center point, and the reference direction; the reference direction is the direction from the theoretical coordinates of the center point to the theoretical coordinates of the upper left point; By traversing all the measured circumferential feature point coordinates of the first captured image, the contribution values ​​of the four included angles are obtained; The summation of the four angle contribution values ​​and the average value are used to obtain the image rotation error of the first camera.

[0015] This preferred solution categorizes inherent errors into three types: horizontal, vertical, and rotational. This comprehensive approach covers all potential shooting deviations inherent in vehicle-mounted cameras. The inclusion of rotational error detection fills a gap in traditional testing methods that only focus on positional deviations while ignoring rotational ones. Since vehicle-mounted cameras may rotate slightly after installation, this error directly affects the multi-view image fusion effect. Therefore, detecting rotational error is crucial for improving camera shooting accuracy. The inherent image rotational error is calculated by determining the distance and angle contribution value of a single circumferential measurement point, followed by multi-feature point averaging. This effectively reduces the impact of individual feature point deviations on the rotational error calculation results, improving accuracy. This calculation method requires no additional testing equipment and can be completed based on existing measurement data, simplifying the testing process, reducing costs, and allowing for standardized calculations for industrial-scale application, further enhancing the practicality of the testing method.

[0016] As a preferred embodiment, the distance at which the vehicle-mounted multi-view camera captures images of the map card is controlled to be a preset vertical shooting distance, which is the closest distance that satisfies the depth-of-field range of all cameras in the vehicle-mounted multi-view camera.

[0017] This preferred solution sets the shooting distance to the closest possible distance within the depth of field of all cameras. This ensures that the images captured by each camera are clear and feature points have distinct edges, avoiding image blurring and difficulty in feature point extraction due to excessive distance, or image distortion caused by distance exceeding the depth of field of some cameras. This effectively improves the accuracy of feature point coordinate extraction. Simultaneously, setting the closest distance shortens the shooting distance, reduces inspection time, and adapts to the industrial needs of batch inspection of vehicle-mounted multi-view cameras, improving inspection efficiency. Furthermore, the fixed shooting distance standard reduces distance fluctuations in different inspection scenarios and batches, ensuring consistency of inspection conditions, reducing the impact of distance deviations on error calculation, improving the repeatability and comparability of inspection results, and further guaranteeing the reliability of inspection data.

[0018] As a preferred embodiment, the system errors include horizontal image system errors, vertical image system errors, and image rotation system errors; The image rotation system error is calculated based on the image rotation error of the reference camera and the image rotation errors of other cameras.

[0019] This preferred solution categorizes systematic errors into three types: horizontal, vertical, and rotational. This precisely corresponds to various relative deviations that may exist between multiple cameras, overcoming the limitations of traditional detection methods that suffer from vague systematic error classification and the inability to pinpoint specific deviation types. This allows for targeted subsequent calibration work, significantly improving calibration efficiency. Image rotation systematic errors are calculated based on the rotation errors of each camera's own image, achieving a precise correlation between systematic errors and their individual errors. This avoids subjective estimation of relative rotation deviations between cameras, significantly improving the reliability of image rotation systematic error measurement. The core value of in-vehicle multi-view cameras lies in multi-view image collaborative fusion. Systematic errors directly determine fusion accuracy, thus affecting the safety of in-vehicle vision functions such as autonomous driving and panoramic imaging. This solution, through a clear systematic error detection method, provides clear data support for multi-camera collaborative calibration, effectively reducing relative deviations between cameras, optimizing collaborative work effects, ensuring the overall operational accuracy of the in-vehicle vision system, and adapting to the practical application needs of high-precision in-vehicle scenarios.

[0020] This application also provides an image position error detection device for vehicle-mounted multi-view cameras, including a data acquisition module, an inherent error module, and a system error module; The data acquisition module is used to control the vehicle-mounted multi-view camera to capture images of the map card and extract the measured coordinates of the center point and circumferential feature points from the images captured by each camera. The self-error module is used to compare and calculate the distance between the measured coordinates of each camera and the corresponding preset feature point theoretical coordinates to obtain the self-error of each camera. The system error module is used to perform relative calculations between the self-error of the reference camera in the vehicle-mounted multi-camera system and the self-errors of the other cameras to obtain the system error between the cameras in the vehicle-mounted multi-camera system.

[0021] This application also provides a storage medium storing a computer program, which is called and executed by a computer to implement the image position error detection method for vehicle-mounted multi-view cameras as described above.

[0022] This application also provides a computer program product, including a computer program or instructions, which, when executed by a communication device, implements the image position error detection method for vehicle-mounted multi-view cameras as described above. Attached Figure Description

[0023] Figure 1This is a flowchart illustrating an image position error detection method for a vehicle-mounted multi-view camera provided in an embodiment of this application; Figure 2 This is a schematic diagram of a five-point drawing provided in an embodiment of this application; Figure 3 This is a diagram of the installation environment provided in the embodiments of this application; Figure 4 This is an overall flowchart provided in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of an image position error detection device for a vehicle-mounted multi-view camera provided in an embodiment of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" and "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "several" means two or more.

[0026] The image position error detection method for vehicle-mounted multi-cameras provided in this application aims to solve the technical problem that the existing technology can only test the OC or COD parameters of a single vehicle-mounted camera module, but cannot effectively detect the image position error of a single camera and the system error between cameras after the vehicle-mounted multi-camera system is assembled into a whole. This method can compensate for the defects caused by the inability to identify such errors, such as insufficient consistency in product assembly or structural dimensions, outflow of products with abnormal positions, failure to detect image position problems before the vehicle-mounted camera is calibrated and thus causing calibration failure, and the impact on the performance of vehicle-mounted related applications.

[0027] Example 1: Please see Figure 1 The embodiments of this application provide an image position error detection method for vehicle-mounted multi-view cameras, including S1~S3, and the specific implementation steps are as follows: S1. Control the vehicle-mounted multi-camera to capture images of the map card, and extract the measured coordinates of the center point and circumferential feature points from the images captured by each camera.

[0028] Step S1 in this embodiment includes S1.1 to S1.3, specifically as follows: S1.1 Set up a five-point chart, including one center point and four circumferential feature points. The distance between any two of the four circumferential feature points is the first distance R. The length and width of the chart are obtained as follows: The size design of the diagram is based on the camera with the smallest field of view (FOV) among the preset vehicle multi-view cameras, and the horizontal field of view ∠AH and the vertical field of view ∠BH of the camera are extracted; Meanwhile, a preset vertical shooting distance H is set to be the closest distance that satisfies the depth-of-field range of all cameras in both the preset vehicle-mounted multi-camera system and the subsequent vehicle-mounted multi-camera systems to be tested, thus avoiding the problem of excessively large image size caused by an excessively high H. It should be noted that the "preset vehicle-mounted multi-camera system" is a "reference benchmark" used to derive theoretical coordinates, while the "subsequent vehicle-mounted multi-camera systems to be tested" is the actual object being tested. The camera configurations of the two are identical, so the preset vertical shooting distance H can simultaneously adapt to the depth-of-field requirements of both.

[0029] Based on the tangent calculation logic of a right triangle, the length of the image card is calculated according to the horizontal field of view and the preset vertical shooting distance from the vehicle-mounted multi-view camera to the image card. The calculation formula is: Image card length = 2H·tan(∠AH / 2); the width of the image card is calculated according to the vertical field of view and the preset vertical shooting distance. The calculation formula is: Image card width = 2H·tan(∠BH / 2). It should be noted that the above image card length and width can be adaptively adjusted according to the actual application scenario to meet the usage requirements under different testing conditions.

[0030] For examples of this application, please refer to [link / reference]. Figure 2 , Figure 2 This is a schematic diagram of a five-point pattern card provided in the embodiments of this application. Specifically, the pattern card has four circumferential feature points distributed in the four directions of upper left, upper right, lower left, and lower right, with a center point as the reference. All five feature points are black dots, and the diameter of the dots is set to 50 pixels. This pixel specification is based on the pixel size of the camera with the smallest field of view in the preset vehicle multi-view camera.

[0031] In this embodiment S1.1, a five-point chart is used, with one center point and four circumferential feature points. The distance between any two of the four circumferential feature points is fixed as the first distance. This simplifies the feature point extraction process and forms a stable coordinate reference system through symmetrically distributed feature points. The center point can serve as the core reference to quickly locate the image coordinate origin, while the four circumferential feature points can comprehensively cover different positions of the image, effectively capturing horizontal and vertical positional deviations during camera shooting and reducing error interference caused by single feature point extraction. The fixed first distance provides clear parameter support for the accurate calculation of the theoretical coordinates of subsequent preset feature points, making the distance comparison between measured coordinates and theoretical coordinates more targeted, improving the accuracy of error calculation for each camera, and thus laying a solid foundation for subsequent system error calculation. This approach balances detection accuracy and efficiency, adapting to the actual detection scenario requirements of vehicle-mounted multi-view cameras. Furthermore, the dimensions of the image card are calculated based on the angle parameters of the camera with the smallest field of view and the preset vertical shooting distance. This ensures that all cameras can capture the image card completely, avoiding the problem that some cameras cannot capture all feature points due to the image card being too large or too small. At the same time, there is no need to manually measure the image card size, which improves detection efficiency. It also provides accurate dimensional basis for subsequent setting of preset feature point theoretical coordinates and error calculation, adapting to the detection needs of different specifications of vehicle-mounted multi-view cameras and enhancing the versatility and practicality of the method. Furthermore, setting the shooting distance to the closest possible distance within the depth of field of all cameras ensures that the images captured by each camera are clear and feature point edges are distinct. This avoids image blurring and difficulty in feature point extraction due to excessive distance, or image distortion caused by distances exceeding the depth of field of some cameras, effectively improving the accuracy of feature point coordinate extraction. Simultaneously, setting the closest distance shortens the shooting journey, reduces inspection time, and adapts to the industrial needs of batch inspection of vehicle-mounted multi-view cameras, improving inspection efficiency. In addition, a fixed shooting distance standard reduces distance fluctuations in different inspection scenarios and batches, ensuring consistency of inspection conditions, reducing the impact of distance deviations on error calculation, improving the repeatability and comparability of inspection results, and further guaranteeing the reliability of inspection data.

[0032] S1.2 First, designate the camera installed at the system reference position among the preset vehicle-mounted multi-camera cameras as the preset reference camera (preset camera 1). Use the center point (P1) of the five-point diagram as the theoretical origin of this preset reference camera, and set its coordinates to (0,0). Combining the first distance R, and following the four-directional symmetrical preset distance arrangement rule based on the origin, determine the positions of the four circumferential feature points (P2-P5)—P2 is located to the upper left of the center point, P3 to the upper right, P4 to the lower left, and P5 to the lower right. Then, derive the coordinates of these four circumferential feature points. The reference coordinate set is formed by the coordinates of the theoretical origin and the coordinates of the four circumferential feature points. The "four-directional symmetrical preset distance arrangement rule based on the origin" means that the four circumferential feature points are symmetrically distributed horizontally and vertically around the theoretical origin P1; the horizontal and vertical distances from each feature point to the origin are both equal to half of the first distance R, and the straight-line distance between any two adjacent circumferential feature points remains a fixed first distance R.

[0033] For the other cameras in the preset multi-camera system (denoted as preset camera n), open the 3D structure design file of the camera using CATIA software, and locate the optical axis center of the preset reference camera and preset camera n respectively in the file; Based on the optical axis center, the horizontal direction error Kn and vertical direction error Mn of the optical axis between the preset reference camera (preset camera 1) and other preset cameras (preset camera n) are measured using software measurement tools; Based on the coordinates of each feature point in the reference coordinate set, the reverse values ​​of the horizontal direction error Kn and the vertical direction error Mn of the optical axis are superimposed respectively, and the theoretical coordinates of the five preset feature points of the preset camera n are calculated one by one to form the theoretical coordinate set of the preset feature points of the camera.

[0034] Finally, the coordinates of the reference coordinate set and all the coordinates of the preset feature point theoretical coordinate sets of other preset cameras together constitute a complete preset feature point theoretical coordinate system, and the coordinates in this system are the preset feature point theoretical coordinates.

[0035] After setting the drawing and theoretical coordinates, the initialization settings can be completed.

[0036] It should be noted that the selection method for the aforementioned preset reference camera and other cameras is as follows: Preset reference camera (preset camera 1): According to the multi-view system design scheme, a camera with stable optical performance and whose installation position is the system reference is selected, such as the "main camera" marked in the design drawings, as the origin of the preset coordinate system. Other preset cameras (preset camera n): All cameras in the preset system except the preset reference camera, defined sequentially according to design number / installation position, such as left camera, right camera, etc., whose theoretical coordinates are derived based on the parameters of "preset camera 1".

[0037] The specific reference coordinate set corresponding to the preset reference camera (preset camera 1) is as follows, with each coordinate corresponding to a specified feature point on the map: P1 (center point): (X1,Y1)=(0,0); P2 (top left point): (X2, Y2) = (-R / 2, R / 2); P3 (top right point): (X3, Y3) = (R / 2, R / 2); P4 (bottom left point): (X4, Y4) = (-R / 2, -R / 2); P5 (bottom right point): (X5, Y5) = (R / 2, -R / 2).

[0038] The theoretical coordinate set of the preset feature points corresponding to the other cameras (preset camera n) is as follows, and each coordinate corresponds to a specified feature point on the map: P1n (center point): (X1n, Y1n) = (0-Kn, 0-Mn); P2n (top left point): (X2n, Y2n) = (-R / 2 - Kn, R / 2 - Mn); P3n (top right point): (X3n, Y3n) = (R / 2 - Kn, R / 2 - Mn); P4n (bottom left point): (X4n, Y4n) = (-R / 2 - Kn, -R / 2 - Mn); P5n (bottom right point): (X5n, Y5n) = (R / 2 - Kn, -R / 2 - Mn).

[0039] It should be noted that P1(X1,Y1), P2(X2,Y2), P3(X3,Y3), P4(X4,Y4), and P5(X5,Y5) are the theoretical coordinates of the center point, upper left point, upper right point, lower left point, and lower right point of the preset camera 1 when there is no image deviation, respectively. P1n(X1n,Y1n), P2n(X2n,Y2n), P3n(X3n,Y3n), P4n(X4n,Y4n), and P5n(X5n,Y5n) are the theoretical coordinates of the center point, upper left point, upper right point, lower left point, and lower right point of the preset camera n when there is no image deviation, respectively.

[0040] In this embodiment, S1.2 uses the center point of the five-point chart as the preset theoretical origin of the reference camera. Combined with a fixed distance between circumferential feature points, a reference coordinate set is set to ensure the uniformity and standardization of the reference coordinates, reducing subjective errors in coordinate setting. By obtaining the optical axis center of each camera, the horizontal and vertical errors of the optical axes of the preset reference camera and other preset cameras are measured, and these errors are incorporated into the calculation of the theoretical coordinates of other cameras. This ensures that the theoretical coordinates fit the actual installation position of the cameras, avoiding the problem of theoretical coordinates being out of sync with the actual shooting scene due to ignoring camera installation deviations, and ensuring that the comparison between theoretical and measured coordinates is more targeted. Integrating the theoretical coordinates of all cameras forms a complete preset feature point theoretical coordinate system, which can accurately support the calculation of its own errors and system errors, improving the accuracy and reliability of the entire detection method and providing strong support for the accurate calibration of vehicle-mounted multi-view cameras.

[0041] S1.3. Secure the vehicle-mounted multi-view camera system to the dedicated fixture, ensuring the camera is firmly installed without any loosening or shifting. Then, place the five-point reference card and the multi-view camera system facing each other, maintaining a distance strictly within the preset vertical shooting distance H. Finally, check and confirm that the fixture and the five-point reference card are not displaced, ensuring that the shooting angle and distance remain consistent throughout the test, thus completing the setup and testing environment. The "dedicated fixture" refers to a customized fixing device adapted to this camera system, which can accurately position the camera's installation posture and limit displacement.

[0042] The vehicle-mounted multi-camera system is controlled to capture images of a five-point map card using the vehicle-mounted multi-camera system. The measured coordinates of the center point and circumferential feature points are extracted from the images captured by each camera.

[0043] The measured coordinates include the measured coordinates of the center point, the top left point, the top right point, the bottom left point, and the bottom right point. Specifically: The measured coordinates of the reference camera (camera 1) include the position data corresponding to five feature points: the measured coordinates of the center point P1'(X1', Y1'), the measured coordinates of the upper left point P2'(X2', Y2'), the measured coordinates of the upper right point P3'(X3', Y3'), the measured coordinates of the lower left point P4'(X4', Y4'), and the measured coordinates of the lower right point P5'(X5', Y5').

[0044] For the other cameras (camera n) in the multi-camera system, their measured coordinates also include the position data of five feature points: the measured coordinates of the center point P1n'(X1n', Y1n'), the measured coordinates of the top left point P2n'(X2n', Y2n'), the measured coordinates of the top right point P3n'(X3n', Y3n'), the measured coordinates of the bottom left point P4n'(X4n', Y4n'), and the measured coordinates of the bottom right point P5n'(X5n', Y5n').

[0045] It should be noted that the selection method for the aforementioned reference camera and other cameras to be tested is as follows: Reference camera (camera 1): Among the currently tested vehicle-mounted multi-view cameras, the camera whose design position and functional identification are consistent with the "preset reference camera (preset camera 1)", such as the "main camera" marked on the product casing or the corresponding mounting hole position, ensuring that it matches the role of the preset reference. Other cameras (camera n): The cameras in the current system other than the above-mentioned reference camera, which are assigned to the serial number / position of the "other preset cameras (preset camera n)" according to their design number / installation position, to achieve a one-to-one correspondence with the preset vehicle-mounted multi-view cameras.

[0046] In this embodiment S1.3, the measured coordinates cover the center point and four circumferential measured coordinates, precisely corresponding to the feature point layout of the five-point chart. This comprehensively captures the positional deviation of the camera-captured image, avoiding omissions of local deviations due to missing measured point coordinates, and ensuring that the error calculation fully reflects the camera's shooting accuracy. The five measured point coordinates correspond to the center and surrounding areas of the chart, simultaneously capturing both center and edge deviations from the camera's capture, reducing the bias caused by single-area measured data, and improving the comprehensiveness and accuracy of the error calculation. The clear division of measured point coordinates provides a clear measurement object for subsequent error calculations, avoiding confusion during the calculation process, standardizing the testing procedure, improving the standardization and operability of the testing method, and adapting to the high-precision testing needs of vehicle-mounted multi-view cameras.

[0047] S2. Compare the measured coordinates of each camera with the theoretical coordinates of the corresponding preset feature points to calculate the distance and obtain the error of each camera.

[0048] Step S2 in this embodiment of the application is specifically as follows: The measured coordinates of each camera are compared with the theoretical coordinates of the corresponding preset feature points to calculate the distance, thus obtaining the inherent error of each camera. It should be noted that since the camera configuration and image card shooting logic of the preset vehicle-mounted multi-view cameras (including the preset reference camera and other preset cameras) are consistent with those of the currently tested vehicle-mounted multi-view cameras (including the reference camera and other cameras), there is a one-to-one correspondence between the camera coordinates and the captured image card feature point coordinates of the two.

[0049] The inherent error includes horizontal image error, vertical image error, and self-image rotation error. The self-image rotation error is obtained as follows: For the first captured image corresponding to the first camera in the vehicle-mounted multi-view camera, taking the measured coordinates of the upper left point as the calculation object, based on the first distance R, the square of the distance from the measured coordinates of the upper left point to the theoretical coordinates of the center point is calculated to obtain the first value, and the square of the distance from the measured coordinates of the upper left point to the theoretical coordinates of the upper left point is calculated to obtain the second value. Based on the first distance R, the first numerical value, and the second numerical value, the angle contribution value is calculated using the inverse cosine theorem. The angle contribution value is the angle between the line connecting the measured coordinates of the upper left point and the theoretical coordinates of the center point, and the reference direction. The reference direction is the direction from the theoretical coordinates of the center point to the theoretical coordinates of the upper left point. By traversing all the measured circumferential feature point coordinates of the first captured image, the contribution values ​​of the four included angles are obtained; The average of the summation of the contribution values ​​of the four included angles is used to obtain the rotation error of the first camera's own image.

[0050] The following provides a detailed explanation of the formulas for calculating the inherent errors of each camera in a vehicle-mounted multi-view camera system: ① Formula for the inherent error of the reference camera (camera 1): Formula 1 (Horizontal Image Error): CX1 = X1' - X1 = X1', (Since X1 = 0, the horizontal coordinate of the theoretical center point of the preset reference camera is 0) Formula 2 (Vertical Image Error): CY1=Y1'-Y1=Y1', (Since Y1=0, the ordinate of the theoretical center point of the preset reference camera is 0) Formula 3 (self-image rotation error): ② Formula for the inherent error of other cameras (camera n): Formula 4 (Horizontal Image Error): CXn = X1n' - X1n Formula 5 (Vertical Image Error): CYn = Y1n' - Y1n Formula 6 (self-image rotation error): In this embodiment S2, the inherent error is categorized into three types: horizontal, vertical, and rotational. This comprehensively covers the potential shooting deviations of the vehicle-mounted camera itself. The addition of rotational error detection fills the gap in traditional detection methods that only focus on positional deviations while ignoring rotational deviations. Since vehicle-mounted cameras may rotate slightly after installation, this error directly affects the multi-view image fusion effect. Therefore, detecting rotational error is crucial for improving the camera's shooting accuracy. The inherent image rotational error is calculated by determining the distance and angle contribution value of a single circumferential measurement point, followed by multi-feature point averaging. This effectively reduces the impact of individual feature point deviations on the rotational error calculation results, improving calculation accuracy. This calculation method requires no additional detection equipment and can be completed based on existing measurement data, simplifying the detection process, reducing detection costs, and allowing for standardized calculation processes for industrial-scale application, further enhancing the practicality of the detection method.

[0051] S3. Perform relative calculations between the self-error of the reference camera in the vehicle-mounted multi-view camera and the self-errors of the other cameras to obtain the systematic error between the cameras in the vehicle-mounted multi-view camera.

[0052] Step S3 in this embodiment of the application is specifically as follows: The system error between the cameras in the vehicle multi-camera system is obtained by performing relative calculations between the reference camera's own error and the errors of the other cameras. This includes horizontal image system error, vertical image system error, and image rotation system error. The image rotation system error is calculated based on the reference camera's own image rotation error and the errors of the other cameras.

[0053] The final output includes the inherent error of the reference camera in the vehicle-mounted multi-view camera system, as well as the systematic error between each camera.

[0054] The specific formulas for calculating each systematic error are as follows: ① Formula 7 (Horizontal image system error between camera 1 and camera n): Hx = CX1 - CXn ② Formula 8 (Vertical image system error between camera 1 and camera n): Vx = CY1 - CYn ③ Formula 9 (Image rotation system error between camera 1 and camera n): ∠Q' = ∠Q1 - ∠Qn For examples of this application, please refer to [link / reference]. Figure 3-4 , Figure 3This is an installation environment diagram provided in the embodiments of this application, which shows the test setup scene for vehicle-mounted multi-view camera image position detection. It clearly presents the relative positional relationship of the vehicle-mounted multi-view camera system, the five-point map card and the supporting fixture. That is, the multi-view camera system is fixed on the fixture and placed opposite the five-point map card. The distance between the two is set to the closest distance that meets the depth of field range of all cameras. At the same time, the size of the map card is adapted to the camera with the smallest FOV in the system, ensuring that each camera can clearly capture the five feature points on the map card.

[0055] Figure 4 This is an overall flowchart provided in the embodiments of this application, which shows the complete execution logic of vehicle-mounted multi-view camera image position detection. That is, first, the initialization settings and test environment setup are completed, then the reference camera (camera 1) and other cameras (camera n) respectively take pictures of the five-point map card, and the measured coordinates are obtained by calculating the center coordinates of the map card center and the center coordinates of the surrounding feature points. Then, based on the measured coordinates and theoretical coordinates, the error of each camera is calculated, and the systematic error between the cameras is derived. Finally, all error data are output.

[0056] In this embodiment, S3 categorizes system errors into three types: horizontal, vertical, and rotational. This accurately corresponds to various relative deviations that may exist between multiple cameras, overcoming the limitations of traditional detection methods that suffer from vague system error classification and the inability to pinpoint specific deviation types. This allows for targeted subsequent calibration work, significantly improving calibration efficiency. Rotational system errors are calculated based on the rotational errors of each camera's own image, achieving a precise correlation between system errors and their own inherent errors. This avoids subjective estimation of relative rotational deviations between cameras, significantly improving the reliability of rotational system error calculation. The core value of in-vehicle multi-view cameras lies in multi-view image collaborative fusion. System errors directly determine fusion accuracy, thus affecting the safety of in-vehicle vision functions such as autonomous driving and panoramic imaging. This embodiment, through a clearly defined system error detection method, provides clear data support for multi-camera collaborative calibration, effectively reducing relative deviations between cameras, optimizing collaborative work effects, ensuring the overall operational accuracy of the in-vehicle vision system, and adapting to the practical application needs of high-precision in-vehicle scenarios.

[0057] Overall, this embodiment has the following beneficial effects: This invention uses a map as a unified reference, controlling multiple vehicle-mounted cameras to simultaneously capture images of the map. It extracts the measured coordinates of the center point and circumferential feature points in the images captured by each camera, providing a precise measurement data foundation for error detection. Compared to random shooting without a reference, the fixed positions of the feature points on the map, along with their center and circumferential features, comprehensively reflect the image position information, avoiding measurement data deviations caused by complex shooting scenes and unclear feature points, thus ensuring the targeted nature of individual camera image position detection. Secondly, by comparing the measured coordinates of each camera with the theoretical coordinates of corresponding preset feature points, the imaging deviation of a single camera, i.e., its own error, can be directly quantified. This comparison method focuses on the difference between the measured and theoretical coordinates of a single camera, eliminating interference from other cameras and achieving precise localization of the image position error of each individual camera, solving the problem of the difficulty in independently quantifying the error of a single camera. Furthermore, this invention introduces a reference camera as a benchmark, and performs relative calculations on its own error and the own errors of other cameras. By utilizing the fixed reference function of the reference camera, the common deviations that may exist in the individual camera's own error are offset, and the relative deviations between cameras, i.e., system errors, are accurately extracted. Since the system error is essentially the relative positional deviation between cameras, rather than the absolute deviation of a single camera, this relative calculation can effectively remove the interference of the individual camera's own error, and achieve accurate detection of the system error between cameras, thereby simultaneously solving the core problem that both types of errors cannot be accurately detected. In summary, this invention can accurately identify the image errors of individual cameras and the system image errors between cameras after the assembly of a multi-camera system, effectively improving the consistency of product assembly and structural dimensions, and preventing products with abnormal positions from entering the market. At the same time, it can check image position problems in advance before the calibration of the vehicle camera, avoiding the risk of calibration failure due to image deviation. It can also accurately calculate the system image error data after the multi-camera system is assembled, providing a reliable error compensation basis for subsequent algorithm applications, thereby significantly improving the overall performance of vehicle-related applications.

[0058] Example 2: Please see Figure 5 The embodiments of this application provide an image position error detection device for vehicle-mounted multi-view cameras, including a data acquisition module 10, an inherent error module 20, and a system error module 30; Among them, the data acquisition module 10 is used to control the vehicle-mounted multi-view camera to take pictures of the map card and extract the measured coordinates of the center point and circumferential feature points from the pictures taken by each camera. The self-error module 20 is used to compare and calculate the distance between the measured coordinates of each camera and the theoretical coordinates of the corresponding preset feature points to obtain the self-error of each camera. The system error module 30 is used to perform relative calculations between the self-error of the reference camera in the vehicle-mounted multi-view camera and the self-errors of other cameras to obtain the system error between the cameras in the vehicle-mounted multi-view camera.

[0059] It should be noted that the technical concept of this second embodiment is completely consistent with that of the first embodiment. The two maintain a high degree of synergy at the technical logic level. The specific technical details can be referred to the relevant description of the first embodiment, which will not be repeated here.

[0060] Example 3: This application provides a computer-readable storage medium, which includes a stored computer program, wherein the computer program controls the device where the computer-readable storage medium is located to execute the image position error detection method of a vehicle-mounted multi-view camera when it is running. The image position error detection method for vehicle-mounted multi-view cameras, when implemented as a software functional unit and used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0061] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this application are performed entirely or partially. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, or other programmable device. The computer program or instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; it can also be an optical medium, such as a digital video optical disc; or it can be a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both types of storage media.

[0062] The above are preferred embodiments of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for detecting image position error in a vehicle-mounted multi-view camera, characterized in that, include: Control the vehicle-mounted multi-view camera to capture images of the map card, and extract the measured coordinates of the center point and circumferential feature points from the images captured by each camera; The measured coordinates of each camera are compared with the theoretical coordinates of the corresponding preset feature points to calculate the distance, thus obtaining the error of each camera. The system error between the cameras in the vehicle-mounted multi-camera system is obtained by performing a relative calculation between the error of the reference camera and the errors of the other cameras.

2. The image position error detection method for vehicle-mounted multi-view cameras as described in claim 1, characterized in that, The chart is a five-point chart, including a center point and four circumferential feature points, and the distance between any two of the four circumferential feature points is the first distance.

3. The image position error detection method for vehicle-mounted multi-view cameras as described in claim 2, characterized in that, The method for obtaining the theoretical coordinates of the preset feature points is as follows: The camera installed at the system reference position in the preset vehicle-mounted multi-view camera is set as the preset reference camera. The center point coordinates of the five-point map card are set as the theoretical origin coordinates of the preset reference camera. Combined with the first distance, the coordinates of the four circumferential feature points are calculated according to the four-directional symmetrical preset distance arrangement rules based on the origin. The theoretical origin coordinates and the coordinates of the four circumferential feature points constitute the reference coordinate set. Obtain the optical axis center of the preset reference camera and other preset cameras in the preset vehicle-mounted multi-view camera; Based on the optical axis center, the horizontal and vertical optical axis errors between the preset reference camera and other preset cameras are measured. Based on the coordinates of each feature point in the reference coordinate set, and combined with the horizontal and vertical optical axis errors, the theoretical coordinate sets of preset feature points for other preset cameras are calculated; wherein, all coordinates in the reference coordinate set and the theoretical coordinate sets of preset feature points for other preset cameras together constitute the theoretical coordinates of the preset feature points.

4. The image position error detection method for vehicle-mounted multi-view cameras as described in claim 2, characterized in that, The measured coordinates include the measured coordinates of the center point, the upper left point, the upper right point, the lower left point, and the lower right point.

5. The image position error detection method for vehicle-mounted multi-view cameras as described in claim 4, characterized in that, The self-error includes horizontal image error, vertical image error, and self-image rotation error. The self-image rotation error is obtained as follows: For the first captured image corresponding to the first camera in the vehicle-mounted multi-view camera, taking the measured coordinates of the upper left point as the calculation object, based on the first distance, the square of the distance from the measured coordinates of the upper left point to the theoretical coordinates of the center point is calculated to obtain a first value, and the square of the distance from the measured coordinates of the upper left point to the theoretical coordinates of the upper left point is calculated to obtain a second value. Based on the first distance, the first value, and the second value, the angle contribution value is calculated using the inverse cosine theorem; wherein, the angle contribution value is the angle between the line connecting the measured coordinates of the upper left point and the theoretical coordinates of the center point, and the reference direction; the reference direction is the direction from the theoretical coordinates of the center point to the theoretical coordinates of the upper left point; By traversing all the measured circumferential feature point coordinates of the first captured image, the contribution values ​​of the four included angles are obtained; The summation of the four angle contribution values ​​and the average value are used to obtain the image rotation error of the first camera.

6. The image position error detection method for vehicle-mounted multi-view cameras as described in claim 1, characterized in that, The distance at which the vehicle-mounted multi-view camera captures images of the map card is a preset vertical shooting distance, which is the closest distance that satisfies the depth-of-field range of all cameras in the vehicle-mounted multi-view camera.

7. The image position error detection method for vehicle-mounted multi-view cameras as described in claim 1, characterized in that, The systematic errors include horizontal image systematic errors, vertical image systematic errors, and image rotation systematic errors; The image rotation system error is calculated based on the image rotation error of the reference camera and the image rotation errors of other cameras.

8. An image position error detection device for vehicle-mounted multi-view cameras, characterized in that, It includes a data acquisition module, an intrinsic error module, and a system error module; The data acquisition module is used to control the vehicle-mounted multi-view camera to capture images of the map card and extract the measured coordinates of the center point and circumferential feature points from the images captured by each camera. The self-error module is used to compare and calculate the distance between the measured coordinates of each camera and the corresponding preset feature point theoretical coordinates to obtain the self-error of each camera. The system error module is used to perform relative calculations between the self-error of the reference camera in the vehicle-mounted multi-camera system and the self-errors of the other cameras to obtain the system error between the cameras in the vehicle-mounted multi-camera system.

9. A storage medium, characterized in that, The storage medium stores a computer program, which is called and executed by a computer to implement an image position error detection method for a vehicle-mounted multi-view camera as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by the communication device, they implement the image position error detection method for vehicle-mounted multi-view cameras as described in any one of claims 1 to 7.

Images