Calibration method for definition and photographing pose based on self-made checkerboard

Abstract

The invention discloses a self-made checkerboard-based definition and photographing pose calibration method and system. According to the method, a comprehensive calibration board integrating a high-frequency checkerboard and precise reference dots is adopted as a unified physical reference. The core is to provide a definition quantitative evaluation algorithm, and an objective definition evaluation value is generated by calculating the gray difference before and after the image is slightly translated. By using the evaluation value, the focusing of the main camera and the Sammer camera and the debugging of the Sammer angle can be automatically completed, and the Sammer angle is optimized by taking the optimal global definition as the target. Furthermore, on the basis of the same calibration plate, image resolution calculation, lens distortion correction and perspective distortion correction can be completed in sequence. According to the method, a traditional calibration process depending on subjective experience is converted into objective and quantitative systematic operation, the problems that the calibration process of a multi-camera system is tedious, the consistency is poor and the efficiency is low are solved, and the calibration precision and efficiency and the consistency between equipment are remarkably improved.

Description

A method for calibrating image sharpness and photographic pose based on a self-made chessboard grid. Technical Field

[0001] This invention belongs to the field of machine vision and photoelectric measurement technology, specifically a method for calibrating the sharpness and photographic pose based on a self-made chessboard grid. Background Technology

[0002] In fields such as high-end automated optical inspection (AOI), 3D topography measurement, and precision industrial inspection, multi-camera collaborative imaging systems have become the mainstream technical solution for comprehensively acquiring surface information of objects, eliminating blind spots, or performing 3D reconstruction. A typical and crucial structure is the "1+N" model: a vertically mounted main camera is responsible for acquiring the front view, while multiple tilting cameras mounted around it, strictly adhering to Scheimpflug's principle, are used to acquire side view information, effectively eliminating shadows, accurately detecting sidewalls, or completing high-precision 3D reconstruction.

[0003] However, such advanced multi-camera systems face significant challenges in deployment and calibration. The core issue lies in the lack of an efficient, accurate, and consistent systematic calibration method and dedicated tools. Current technical deficiencies primarily include: First, the calibration process is cumbersome and complex, lacking a unified physical benchmark and integrated workflow. Existing technologies typically require independent focusing, lens distortion correction, resolution calibration, and precise adjustment of the Sham angle for each camera. This process often necessitates the alternating use of different calibration tools or patterns (e.g., one calibration board for focusing, another for distortion correction), resulting in fragmented workflows and cumbersome operations. Frequent tool switching and calibration board repositioning not only significantly reduce efficiency but also introduce unnecessary cumulative errors, severely impacting the final system calibration accuracy.

[0004] Secondly, the key processes of focusing and Sham angle adjustment are highly dependent on manual experience, resulting in low automation and poor consistency. In particular, the ideal goal of Sham angle adjustment is to achieve sharp imaging across the entire tilted focal plane. However, traditional methods rely entirely on the operator's subjective judgment of whether the center and corner areas of the image are "simultaneously sharp" by visually observing the image. This subjective judgment method has inherent drawbacks: it is difficult to adjust, inefficient, and its results are heavily dependent on the skill level of the operators and the on-site conditions. This leads to inconsistent imaging performance across different equipment and batches after adjustment, severely restricting the quality stability of mass-produced equipment.

[0005] Third, perspective distortion correction of tilt camera images is difficult and hard to integrate effectively with the preceding calibration process. Images captured by tilt cameras suffer from severe perspective distortion and must be corrected to a similar top-down perspective to the main camera before effective multi-view fusion or joint analysis can be performed. Existing methods typically require complex, multi-step independent calibration to calculate the perspective transformation matrix. This process is disconnected from preceding steps such as focusing and distortion correction, lacking a systematic solution that runs throughout the entire process.

[0006] In summary, existing technologies for calibrating "1+N" type multi-camera imaging systems suffer from significant drawbacks, including lengthy and discrete processes, strong subjective dependence, and difficulty in guaranteeing accuracy and consistency. The root cause lies in the lack of a dedicated tool that can provide a unified, high-precision physical benchmark for all cameras, as well as a core methodology that can objectify subjective judgments and integrate complex processes. This has become a technical bottleneck restricting the large-scale application and further performance improvement of such high-end vision systems.

[0007] Therefore, there is an urgent need in this field for an innovative technical solution that can solve all the above problems in one stop, thereby achieving efficient, high-precision, and highly consistent calibration of multi-camera systems. Summary of the Invention

[0008] The purpose of this invention is to provide a method for calibrating the sharpness and photographic pose based on a self-made checkerboard pattern. This invention transforms the traditional calibration process that relies on subjective experience into an objective, quantitative, and systematic operation, solving the problems of cumbersome, inconsistent, and inefficient calibration processes for multi-camera systems, and significantly improving calibration accuracy, efficiency, and consistency between devices.

[0009] The technical solution adopted in this invention is as follows: a method for calibrating the sharpness and photographic pose based on a self-made checkerboard pattern, wherein the self-made checkerboard pattern is a comprehensive calibration board integrating a high-frequency checkerboard pattern and precision reference dots on a substrate; the method includes the following steps: a sharpness quantification evaluation step: based on the captured image of the self-made checkerboard pattern, a quantified sharpness evaluation value is generated by calculating the grayscale difference before and after a predetermined pixel translation of the image region; a resolution calculation and camera calibration step: based on the pixel coordinates of the precision reference dots identified in the image and their known physical distances, the image resolution of the camera is calculated, and the sharpness evaluation value is used for automatic focusing; a pose-related correction step: based on the high-frequency checkerboard pattern, lens distortion correction and / or perspective distortion correction are performed.

[0010] Preferably, in the clarity quantification evaluation step, the clarity evaluation value is calculated according to the following formula: ;in, This represents the final sharpness rating; a higher value indicates a sharper image. Indicates the first The average gray value of the absolute difference between the image obtained after translation in each direction and the original image. These correspond to the four translation directions: up, down, left, and right. The calculation formula is: ;in, This represents the original image region to be evaluated. Indicates the image region Along the first Translation in each direction The new image obtained after [number] pixels. It is a positive integer (usually 1 or 2). Represents image region Width (in pixels). Represents image region Height (in pixels). This represents the coordinate index of a pixel in the image.

[0011] Preferably, the image resolution in the resolution calculation and camera calibration steps is calculated using the following formula: ; where Scale represents the image resolution, that is, the actual physical size represented by each pixel, in millimeters per pixel. It represents the actual physical distance between two precision reference points, in millimeters. This indicates the center pixel coordinates of the first precision reference point in the image. This indicates the center pixel coordinates of the second precision reference point in the image.

[0012] Preferably, the lens distortion correction in the pose-related correction step adopts the following model formula: ; ;in, This represents the normalized image coordinates before distortion occurred (the origin of the coordinates is located at the principal point of the image, and has been normalized by dividing by the focal length). This represents the normalized image coordinates after distortion correction. Represents the normalized image coordinates. Distance to the principal point . This represents the radial distortion coefficient of the lens, used to correct distortion distributed along the radial direction. This represents the tangential distortion coefficient of the lens, used to correct distortion caused by the lens not being parallel to the imaging plane.

[0013] Preferably, the perspective distortion correction in the pose-related correction step adopts the following perspective transformation model: ;in, This represents the homogeneous coordinates of a point in the corrected image. This represents the homogeneous coordinates of the corresponding point in the source image to be corrected. to Constructing a 3x3 perspective transformation matrix elements, This represents the scale equivalence relation in homogeneous coordinates. The perspective transformation matrix... The solution is obtained by solving for four or more corresponding point pairs, including the point coordinates in the source image and their target coordinates in the corrected image.

[0014] Preferably, the autofocus process includes: controlling the camera to move along its optical axis and calculating in real time the sharpness evaluation value of a specified area in the captured image. ,Will The position of the camera when the maximum value is obtained is determined as its optimal focus position.

[0015] Preferably, the method is applied to an imaging system comprising a vertical main camera and at least one tilting SAM camera; the calibration of the SAM camera includes: after the main camera has completed focusing and the calibration plate position is fixed, fine-tuning the angle and focal length of the SAM camera, and calculating in real time the sharpness evaluation values ​​of multiple preset regions of interest in its image. , in all regions The goal is to achieve maximum values ​​and close to the target values, thus completing the focusing and adjustment of the Sham camera.

[0016] Preferably, after completing lens distortion correction or perspective distortion correction, the obtained distortion coefficients or perspective transformation matrix are saved as correction parameters and applied to all subsequent working images captured by the camera for real-time or offline correction.

[0017] Preferably, the substrate of the self-made checkerboard pattern is made of a dimensionally stable ceramic material, and the high-frequency checkerboard pattern and precision reference dots are machined on its surface with high precision.

[0018] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described.

[0019] In summary, due to the adoption of the above technical solutions, the beneficial effects of this invention are as follows: the calibration method provided by this invention achieves a fundamental paradigm shift compared to traditional techniques that rely on discrete tools and subjective experience. Its beneficial effects are not simply additive, but rather, through the two core innovations of "unified physical benchmarks" and "quantified evaluation standards," it triggers a series of synergistic effects from quantitative to qualitative changes, ultimately giving rise to certain technical effects that exceed the expectations of those skilled in the art. First, the most direct effect is reflected in the significant improvement in efficiency and the great simplification of the operation process. In traditional methods, the cumbersome tool switching and repositioning processes are replaced by a highly integrated comprehensive calibration board, which directly eliminates the accumulation of errors and time consumption between processes, enabling the complex multi-camera system calibration process to be streamlined. However, the core contribution of this invention goes far beyond this. Its breakthrough lies in the introduction of a clear quantified evaluation algorithm (G... This method transforms the calibration process from an "art" into a "science." This brings about a second layer of effect: a revolution in automation and consistency driven by objective quantification. Specifically, the focusing process achieves precise automatic optimization due to a clear numerical target. More importantly, this method unexpectedly and successfully solves the long-standing problem of Sham angle adjustment that has plagued the industry for years. The "global sharpness" target, traditionally relying on engineers' feel and visual judgment, is transformed into an executable mathematical instruction—"to make the G values ​​of multiple preset regions in the image..." "The _cl value is maximized and approaches simultaneously." This transformation not only makes SAM debugging learnable and repeatable, but also fundamentally guarantees the high consistency of imaging performance of mass-produced equipment, achieving a qualitative leap from "craftsmanship skills" to "standardized processes." Furthermore, the deep integration of the aforementioned "unified benchmark" and "quantitative standard" has given rise to deeper and more disruptive unexpected technical effects—a systematic leap in overall system accuracy and an intelligent innovation in operation and maintenance models.

[0020] Firstly, it eliminates the systematic accuracy gain caused by the "bucket effect." In traditional discrete calibration, errors in each step are propagated and amplified. This invention constructs a positively promoting accuracy guarantee chain: based on G... The precise focusing of _cl provides the conditions for high-precision feature point extraction; and high-precision feature points are the foundation for accurately solving the lens distortion coefficient and perspective transformation matrix. This interconnected precision enables the final overall system calibration accuracy to exceed the theoretical limit that can be achieved by optimizing each link independently in the traditional mode, resulting in a synergistic precision amplification effect of 1+1>2.

[0021] Secondly, it endows equipment with "traceable health status" and predictive maintenance capabilities. Traditional methods store the "optimal state" in the engineer's instantaneous judgment, which cannot be recorded. However, this invention generates a complete digital file (including G...) after each calibration. The system uses parameters such as peak value, resolution, distortion parameters, and perspective matrix to quantify and archive the "health status" of the equipment. When equipment performance drifts, comparing current calibration data with factory baseline data allows for precise diagnosis of whether the problem stems from focusing deviation, changes in mechanical stress, or damage to optical components. This upgrades equipment maintenance from an "experience-driven" passive response to a "data-driven" predictive maintenance approach. This is not merely an improvement in accuracy, but a revolution in operational philosophy, bringing users long-term stable value and low maintenance costs.

[0022] In summary, the beneficial effects of this invention form a progressive system: it begins with efficiency optimization, succeeds with breakthroughs in the automation and consistency of key technologies, and ultimately culminates in the ultimate breakthrough in overall system performance and the emergence of full lifecycle management capabilities. These profound effects fully demonstrate that this invention is not simply a tool improvement, but a comprehensive solution capable of leading technological transformation in the industry. Attached Figure Description

[0023] Figure 1 is a flowchart of the present invention; Figure 2 is a flowchart of the sharpness quantification evaluation step of the present invention; Figure 3 is a flowchart of the resolution calculation and camera calibration steps of the present invention; Figure 4 is a flowchart of the pose correlation correction step of the present invention; Figure 5 is a schematic diagram of the correction of an image captured by the SAM camera to an orthophoto angle similar to that of the main camera; Figure 6 is a schematic diagram of the image sharpness correction curve of the present invention; Figure 7 is a schematic diagram of the sharpness correction effect of the present invention; Figure 8 is a schematic diagram of the perspective correction effect comparison of the present invention. Detailed Description of Embodiments To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.

[0024] Referring to Figures 1 to 8, this invention relates to a method for calibrating sharpness and photographic pose based on a self-made checkerboard pattern. The self-made checkerboard pattern is a comprehensive calibration board integrating a high-frequency checkerboard pattern and precision reference dots on a substrate. The method includes the following steps: S1. Sharpness quantification evaluation step: Referring to Figure 2, based on the captured image of the self-made checkerboard pattern, a quantified sharpness evaluation value is generated by calculating the grayscale difference before and after a predetermined pixel shift in the image region. In the sharpness quantification evaluation step, the sharpness evaluation value is calculated according to the following formula: ;in, This represents the final sharpness rating; a higher value indicates a sharper image. Indicates the first The average gray value of the absolute difference between the image obtained after translation in each direction and the original image. These correspond to the four translation directions: up, down, left, and right. The calculation formula is: ;in, This represents the original image region to be evaluated. Indicates the image region Along the first Translation in each direction The new image obtained after [number] pixels. It is a positive integer (usually 1 or 2). Represents image region Width (in pixels). Represents image region Height (in pixels). This represents the coordinate index of a pixel in the image.

[0025] This step operates based on the intrinsic relationship between image gradient characteristics and sharpness. Its core mechanism lies in converting the high-frequency information richness (i.e., sharpness) of an image into a calculable gray-level difference statistic by introducing a small translation difference method. Specifically, performing a finite-pixel translation operation on the original image region essentially introduces a controllable, small spatial offset, thereby amplifying the gray-level discontinuities caused by edges and textures at the pixel level. Calculating the absolute difference between the image before and after the translation and statistically analyzing its average gray level essentially quantifies the expected gradient strength of the image along the translation vector direction. By covering four orthogonal directions and taking the average, the final sharpness evaluation value (G) is obtained. _cl) constitutes a robust estimate of the overall gradient field intensity of the image, thus objectively characterizing the focus state of the image.

[0026] The primary benefit of this step is the realization of an objective, quantitative evaluation of image sharpness, transforming the traditionally subjective human-based judgment of "sharpness" and "blurriness" into a precise, programmable numerical indicator. This fundamentally eliminates the uncertainty and inconsistency caused by human factors. Secondly, this evaluation value serves as a crucial input signal for subsequent steps, laying the foundation for fully automatic focusing and Sham angle adjustment, making closed-loop control possible. Furthermore, the algorithm, due to its multi-directional averaging characteristics, possesses inherent robustness to image noise and edge directions, ensuring the stability and reliability of the evaluation results.

[0027] S2. Resolution Calculation and Camera Calibration Steps: Referring to Figure 3, based on the pixel coordinates of the precision reference point identified in the image and its known physical distance, the image resolution of the camera is calculated, and autofocus is performed using the sharpness evaluation value; the image resolution in the resolution calculation and camera calibration steps is calculated according to the following formula: ; where Scale represents the image resolution, that is, the actual physical size represented by each pixel, in millimeters per pixel. It represents the actual physical distance between two precision reference points, in millimeters. This indicates the center pixel coordinates of the first precision reference point in the image. This indicates the center pixel coordinates of the second precision reference point in the image.

[0028] The autofocus process includes: controlling the camera to move along its optical axis and calculating the sharpness evaluation value of a specified area in the captured image in real time. ,Will The position of the camera when the maximum value is obtained is determined as its optimal focus position.

[0029] This step integrates two major functions: autofocus and physical scale calibration. The mechanism of autofocus lies in applying the sharpness evaluation value (G... cl) is used as the objective function for optimization. By controlling the camera's displacement along the optical axis and sampling G in real time... The cl value essentially involves performing a peak search process in one-dimensional space. When the image plane coincides with the sensor plane, G... The maximum value of _cl is used to precisely locate the optimal focus position. The mechanism for resolution calculation is based on the perspective projection ratio under the pinhole imaging model. Under the premise of precise focus, there is a direct linear ratio between the feature point (precision reference point) with known physical distance in the world coordinate system and its projection point on the image plane. By accurately extracting the image coordinates of the feature points and combining them with their prior physical distance, the image resolution (scale), that is, the actual physical size represented by each pixel, can be directly calculated.

[0030] The main technical benefits of this step lie in automation and precision. First, it fully automates the camera focusing process, significantly improving debugging efficiency and consistency. Second, it provides a measurement benchmark (scale) that accurately correlates the image pixel dimension with the real-world physical dimension, a prerequisite for all subsequent quantitative measurements. This step integrates two key calibration actions into the same hardware platform and workflow, avoiding system errors introduced by tool switching and demonstrating the advantages of method integration.

[0031] S3. Pose-related correction step: Referring to Figure 4, lens distortion correction and / or perspective distortion correction are performed based on the high-frequency checkerboard pattern. The lens distortion correction in the pose-related correction step adopts the following model formula: ; ;in, This represents the normalized image coordinates before distortion occurred (the origin of the coordinates is located at the principal point of the image, and has been normalized by dividing by the focal length). This represents the normalized image coordinates after distortion correction. Represents the normalized image coordinates. Distance to the principal point . This represents the radial distortion coefficient of the lens, used to correct distortion distributed along the radial direction. This represents the tangential distortion coefficient of the lens, used to correct distortion caused by the lens not being parallel to the imaging plane.

[0032] The perspective distortion correction in the pose-related correction step adopts the following perspective transformation model: ;in, This represents the homogeneous coordinates of a point in the corrected image. This represents the homogeneous coordinates of the corresponding point in the source image to be corrected. to Constructing a 3x3 perspective transformation matrix elements, This represents the scale equivalence relation in homogeneous coordinates. The perspective transformation matrix... The solution is obtained by solving for four or more corresponding point pairs, including the point coordinates in the source image and their target coordinates in the corrected image.

[0033] This step aims to eliminate image geometric distortion caused by both camera intrinsic parameters (lens distortion) and extrinsic parameters (photographic pose). Lens distortion correction is based on a parametric mathematical model of optical distortion (such as the Brown-Conrad model). By utilizing a large number of feature points (checkerboard corner points) with known ideal geometric positions on a calibration board, their theoretical projected coordinates are compared with the actual extracted distortion coordinates in the image. An optimization algorithm (such as least squares) is used to inversely calculate a set of optimal distortion coefficients. This set of coefficients fully characterizes the lens's distortion characteristics, allowing the establishment of a pixel mapping relationship from a distorted image to a distortion-free image. Perspective distortion correction is based on perspective transformation theory. By selecting a coplanar quadrilateral feature of known physical size in the image (composed of checkerboard corner points) and its ideal rectangular position in the corrected image, a 3x3 perspective transformation matrix (H) is solved. This matrix describes the projection transformation relationship from a tilted viewpoint to an orthographic viewpoint; applying its inverse transformation corrects perspective distortion caused by non-frontal viewpoints caused by camera pose.

[0034] Referring to Figure 5, the core effect of this step is a significant improvement in the geometric fidelity of the image. Lens distortion correction eliminates the inherent nonlinear errors of the optical system, ensuring high-precision dimensional measurement and positioning. Perspective distortion correction unifies images taken from different viewpoints to the same standard orthophoto perspective, greatly facilitating the fusion and joint analysis of multi-view data, particularly suitable for applications such as sidewall detection and 3D reconstruction. The combination of these two methods ensures the accuracy and reliability of subsequent image processing and analysis results.

[0035] Furthermore, the method is applied to an imaging system comprising a vertical main camera and at least one tilted Sham camera; the core challenge of this system lies in the precise calibration of the Sham camera. According to Sham's law, the angle (Sham angle) and focal length of the tilted Sham camera must be finely adjusted so that its focal plane precisely coincides with the calibration plate plane (i.e., the plane of the object under test). Traditional methods rely entirely on the calibrator to observe the image visually and subjectively judge whether the image center and corner areas are "simultaneously sharp." This process is not only extremely inefficient but also heavily dependent on personal experience, difficult to quantify and reproduce, resulting in poor performance consistency between different devices and becoming a bottleneck restricting system performance and mass application.

[0036] To address the aforementioned challenges, this invention creatively applies clear, quantitative evaluation metrics to the debugging process of the SAM camera, forming a data-driven, precise debugging method. The specific implementation process is as follows: First, precise focusing, resolution calibration, and lens distortion correction are completed for the vertical main camera. Afterward, the spatial position of the integrated calibration board is strictly kept fixed, thereby establishing a unified and stable world coordinate system benchmark for the entire multi-camera system.

[0037] For each SAM camera, multiple regions of interest (ROIs) are scientifically defined in advance on the images it acquires, typically covering the image center and the four corner areas. During debugging, the system calculates and displays the sharpness evaluation value of each ROI in real time. ).

[0038] The debugging objective is transformed from a subjective visual perception into an objective mathematical optimization problem. Debuggers (or automated control systems) fine-tune the tilt angle and focal length of the SAM camera to optimize all preset ROIs. The simultaneous attainment of maximum values, with these peak values ​​being close to each other, represents a clear optimization objective. When this state is achieved, it indicates that the focal plane of the SAM camera has achieved optimal matching with the calibration plate plane in the tilt direction, ensuring uniform sharpness across the entire imaging field of view.

[0039] This application solution has resulted in significant technological advancements, successfully transforming the debugging of SAM cameras from an experience-based "skill" into a quantifiable, reproducible, and automated standard process. This not only greatly improves debugging efficiency and accuracy, but more importantly, it ensures that mass-produced equipment achieves highly consistent imaging performance, laying a solid foundation for the reliability and stability of industrial quality inspection.

[0040] After successfully solving for the lens distortion coefficients and perspective transformation matrix, the method requires that these key parameters be stored non-volatilely as correction parameters. These parameters, along with the corresponding camera identification (such as the serial number), are stored in the device's configuration file or firmware.

[0041] In actual detection or measurement tasks, when the camera acquires a working image, the system will automatically call the pre-stored correction parameters to perform real-time or offline geometric correction on each frame of the image.

[0042] By using embedded processors or high-performance computers and employing pre-computed mapping tables or fast transformation algorithms, video streams or real-time acquired images can be quickly corrected to meet the timeliness requirements of online detection.

[0043] These correction parameters can be applied in post-processing software to perform high-precision geometric reconstruction of the original image data after acquisition and storage.

[0044] This "one-time calibration, long-term application" model forms a complete technical closed loop from calibration to application, which greatly improves the practicality and ease of use of the entire vision system and ensures the continuous accuracy of measurement results throughout the entire system lifecycle.

[0045] Furthermore, the high reliability of the calibration method is based on the high precision and stability of the calibration board (i.e., the self-made checkerboard pattern). Therefore, the optimal selection of its materials and manufacturing processes is the physical basis for the implementation of the method. The substrate of the calibration board is preferably made of ceramic material with extremely high dimensional stability. This is because ceramics have an extremely low coefficient of thermal expansion and excellent mechanical stability, effectively resisting the effects of environmental temperature fluctuations and physical stress, thereby ensuring the long-term dimensional stability and reliability of the calibration board as a metrological reference and avoiding systematic errors introduced by deformation of the reference itself. The high-frequency checkerboard pattern and precision reference dots on the calibration board must be fabricated on the substrate surface using high-precision processing techniques. This typically refers to the use of micro-machining technologies such as ultra-precision photolithography, laser direct writing, or etching. These processes ensure that the geometric dimensions of the pattern (such as the side length of the squares and the center distance of the dots) have extremely high processing accuracy (up to micrometer or even sub-micrometer level), and that the pattern edges are sharp and have high contrast. This provides a prerequisite for image processing algorithms to perform high-precision feature point localization and extraction, ensuring the accuracy of subsequent steps such as resolution calculation and distortion correction from the source.

[0046] A computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the method described herein. The term "computer-readable storage medium" is a broad concept encompassing any physical or digital medium capable of storing program code, such as optical discs, USB flash drives, memory cards, solid-state drives, and server storage space. When the program is executed by a processor (such as the CPU of a computer, embedded system, or dedicated controller), all the aforementioned calibration steps are performed.

[0047] To objectively evaluate the actual performance of this invention, comparative tests were conducted. The tests were performed on a typical "one main camera, four SAM cameras" five-camera imaging system, used for the three-dimensional topography inspection of precision components. The comparison object was the traditional manual calibration method commonly used in the industry, which relies on the visual judgment of engineers. The tests aimed to quantitatively analyze the performance of the new method in three core indicators: calibration efficiency, result consistency, and calibration accuracy.

[0048] Specifically, the physical parameters of the calibration board are as follows: substrate material: alumina ceramic, thickness 2mm, coefficient of thermal expansion 6.5×10⁻⁶. -6 / ℃; Effective area: 35mm×35mm, flatness <5μm; Checkerboard grid specifications: grid size 5mm×5mm, quantity 7×7, positioning accuracy ±1μm; Precision dots: small dot diameter 2mm, large dot diameter 4mm, spacing 5mm, roundness error <1μm, position accuracy ±1μm; Manufacturing process: laser engraving + aluminum oxide coating, surface reflectivity 18%±2%; Installation requirements: level adjustable platform, flatness <5μm, rigid clamp fixation.

[0049] Camera system configuration: Main camera: resolution 4096×3000, pixel size 3.45μm; SAM cameras: 4, resolution 5472*3648, pixel size 2.4μm; Main lens parameters: telecentric lens 0.345x; SAM lenses: 4 perspective lenses, focal length 25mm; Mounting geometry: Main camera working distance: 300mm±10mm; SAM camera angle: 30°±5° (relative to the main camera optical axis); Camera spacing: 90° evenly distributed in a ring; Comparison method: Traditional method: an engineer with 3 years of relevant experience manually adjusts the focus ring and camera angle by visually observing image sharpness.

[0050] The method of this invention: using the sharpness evaluation value Gr_cl described in this invention as a guide, the value is observed in real time through the software interface, and manual or semi-automatic adjustment is performed (the operator adjusts the knob according to the trend of the value change, rather than relying on the software to directly control the motor).

[0051] Test task: Complete the calibration of the entire five-camera system, including: focusing the main camera, focusing the four Sham cameras, and adjusting the Sham angle.

[0052] Evaluation metric: Efficiency: Total time spent completing the debugging of all 5 cameras.

[0053] Consistency: The same operator uses the same method to calibrate the same system three times, and records the standard deviation (Std.Dev.) of the final Gr_cl value of each camera. The smaller the standard deviation, the higher the consistency.

[0054] Accuracy (indirect evaluation): After calibration, use the system to measure a standard gauge block of known size (20.000 mm), repeat the measurement 10 times, and calculate the standard deviation of the measured values ​​to evaluate the overall measurement repeatability accuracy of the system. Higher accuracy indirectly proves that the calibration results are better.

[0055] The test data and results are summarized in the table below:

[0056] The following is a table of resolution statistics:

[0057] The method of this invention significantly outperforms traditional methods in terms of total time consumption. The fundamental reason lies in the fact that traditional methods involve a cycle of "trial and error - observation - further trial and error," while the method of this invention is a targeted optimization process of "adjustment - data observation - approaching the optimal." Clear numerical feedback greatly reduces uncertainties in operations.

[0058] Traditional methods yield Gr_cl values ​​with extremely high dispersion (standard deviation) after each calibration, indicating that the results are heavily dependent on the operator's subjective state and cannot guarantee reproducibility. In contrast, the Gr_cl values ​​obtained by the method of this invention are highly concentrated with minimal standard deviation, demonstrating its significant advantages of objectivity and repeatability. This is crucial for quality consistency control in mass production.

[0059] The final system measurement repeatability accuracy is of decisive significance. Systems calibrated using traditional methods exhibit significant measurement fluctuations (25.3 μm), while systems calibrated using the method of this invention show a substantial improvement in measurement accuracy and stability (8.7 μm). This strongly demonstrates that the method of this invention, through objective and quantitative adjustments, obtains more accurate and superior camera intrinsic and extrinsic parameters and distortion correction parameters, thereby providing higher fundamental accuracy for the entire vision system.

[0060] The test results fully demonstrate the superiority of the calibration method described in this invention: in terms of efficiency, this method reduces the calibration time of complex multi-camera systems by about 60%, significantly improving equipment deployment efficiency.

[0061] In terms of consistency, this method completely eliminates the influence of subjective human factors, ensures a high degree of repeatability of calibration results, and provides technical support for the consistency of mass production of products.

[0062] In terms of accuracy, this method optimizes the calibration parameters through objective data, ultimately improving the repeatability accuracy of the vision measurement system by approximately 65%.

[0063] In summary, the calibration method provided by this invention is significantly superior to traditional methods that rely on human experience in terms of efficiency, consistency, and accuracy. It offers an efficient, reliable, and high-precision advanced solution for addressing the calibration challenges of multi-camera vision systems, especially complex systems containing SAM cameras.

[0064] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for calibrating the sharpness and photographic pose based on a self-made chessboard grid, characterized in that, The self-made checkerboard pattern is a comprehensive calibration board integrating a high-frequency checkerboard pattern and precision reference dots on a substrate. The method includes the following steps: a sharpness quantification evaluation step: based on the captured image of the self-made checkerboard pattern, a quantified sharpness evaluation value is generated by calculating the grayscale difference before and after the image region is shifted by a predetermined number of pixels; a resolution calculation and camera calibration step: based on the pixel coordinates of the precision reference dots identified in the image and their known physical distances, the image resolution of the camera is calculated, and the sharpness evaluation value is used for autofocus; a pose-related correction step: based on the high-frequency checkerboard pattern, lens distortion correction and / or perspective distortion correction are performed.

2. The method for calibrating the sharpness and photographic pose based on a self-made chessboard grid according to claim 1, characterized in that, In the clarity quantification evaluation step, the clarity evaluation value is calculated according to the following formula: ;in, This represents the final sharpness rating; a higher value indicates a sharper image. Indicates the first The average gray value of the absolute difference between the image obtained after translation in each direction and the original image. These correspond to the four translation directions: up, down, left, and right. The calculation formula is: ;in, This represents the original image region to be evaluated. Indicates the image region Along the first Translation in each direction The new image obtained after [number] pixels. It is a positive integer. Represents image region Width (in pixels). Represents image region Height (in pixels). This represents the coordinate index of a pixel in the image.

3. The method for calibrating the sharpness and photographic pose based on a self-made chessboard grid according to claim 1, characterized in that, The image resolution in the resolution calculation and camera calibration steps is calculated using the following formula: ; where Scale represents the image resolution, that is, the actual physical size represented by each pixel, in millimeters per pixel. It represents the actual physical distance between two precision reference points, in millimeters. This indicates the center pixel coordinates of the first precision reference point in the image. This indicates the center pixel coordinates of the second precision reference point in the image.

4. The method for calibrating the sharpness and photographic pose based on a self-made chessboard grid according to claim 1, characterized in that, The lens distortion correction in the pose-related correction step adopts the following model formula: ； ;in, This represents the normalized image coordinates before distortion occurred (the origin of the coordinates is located at the principal point of the image, and has been normalized by dividing by the focal length). This represents the normalized image coordinates after distortion correction. Represents the normalized image coordinates. Distance to the principal point 。 This represents the radial distortion coefficient of the lens, used to correct distortion distributed along the radial direction. This represents the tangential distortion coefficient of the lens, used to correct distortion caused by the lens not being parallel to the imaging plane.

5. The method for calibrating the sharpness and photographic pose based on a self-made chessboard grid according to claim 1, characterized in that, The perspective distortion correction in the pose-related correction step adopts the following perspective transformation model: ;in, This represents the homogeneous coordinates of a point in the corrected image. This represents the homogeneous coordinates of the corresponding point in the source image to be corrected. to Constructing a 3x3 perspective transformation matrix elements, This represents the scale equivalence relation in homogeneous coordinates. The perspective transformation matrix... The solution is obtained by solving for four or more corresponding point pairs, including the point coordinates in the source image and their target coordinates in the corrected image.

6. The method for calibrating the sharpness and photographic pose based on a self-made chessboard grid according to claim 2, characterized in that, The autofocus process includes: controlling the camera to move along its optical axis and calculating the sharpness evaluation value of a specified area in the captured image in real time. ,Will The position of the camera when the maximum value is obtained is determined as its optimal focus position.

7. The method for calibrating the sharpness and photographic pose based on a self-made chessboard grid according to claim 6, characterized in that, The method is applied to an imaging system comprising a vertical main camera and at least one tilting Sham camera; The debugging of the SAM camera includes: after the main camera has completed focusing and the calibration plate position is fixed, fine-tuning the angle and focal length of the SAM camera, and calculating the sharpness evaluation values ​​of multiple preset regions of interest in its image in real time. , in all regions The goal is to achieve maximum values ​​and close to the target values, thus completing the focusing and adjustment of the Sham camera.

8. The method for calibrating the sharpness and photographic pose based on a self-made chessboard grid according to claim 4 or 5, characterized in that, After completing lens distortion correction or perspective distortion correction, the obtained distortion coefficients or perspective transformation matrix are saved as correction parameters and applied to all subsequent working images captured by the camera for real-time or offline correction.

9. The method for calibrating the sharpness and photographic pose based on a self-made chessboard grid according to claim 1, characterized in that, The substrate of the self-made checkerboard pattern is made of dimensionally stable ceramic material, and the high-frequency checkerboard pattern and precision reference dots are machined on its surface with high precision.

10. A computer-readable storage medium storing a computer program, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1 to 9.

Images