A binocular camera geometric calibration method

Abstract

This invention relates to the field of computer vision and camera calibration technology, specifically a geometric calibration method for a binocular camera, comprising the following steps: Step S1: Constructing a binocular camera system, the binocular camera system comprising two cameras of the same model and a high-definition LCD screen, with lenses of the same focal length mounted on the front of the two cameras, keeping the two cameras relatively fixed and the optical axes of the lenses parallel to each other, so that the two cameras have a common field of view and a common depth of field; pre-designing a phase-shift circular fringe array pattern, the phase-shift circular fringe array pattern comprising several circular fringe of the same size but different phase shifts. In this invention, the calibration efficiency is high, requiring only 3 phase-shift circular fringe array images to complete the calibration, without the need for additional auxiliary encoded images, greatly reducing the number of calibration images acquired and reducing the calibration workload.

Description

Technical Field

[0001] This invention relates to the field of computer vision and camera calibration technology, and in particular to a geometric calibration method for binocular cameras. Background Technology

[0002] Binocular cameras mimic the principle of binocular parallax in biological systems, acquiring depth information by calculating the differences between two images to achieve 3D spatial perception. With advantages such as low cost, high speed, and ease of integration, this technology has been widely applied in fields such as assisted driving, industrial manufacturing, biomedicine, and consumer entertainment. Geometric calibration is a crucial step in achieving 3D reconstruction, aiming to accurately calculate the internal parameters (such as focal length, principal point, distortion coefficients, etc.) and external parameters (such as rotation matrix, translation vector, etc.) of the binocular camera. Its calibration accuracy directly affects the 3D reconstruction results of the binocular camera system. High-precision targets are the core tool for camera calibration. Based on dimensionality, they can be divided into one-dimensional, two-dimensional, and three-dimensional targets. Among them, two-dimensional targets have the advantages of low cost and high accuracy, and are the most widely used in practice. Commonly used two-dimensional targets include checkerboard patterns and circular arrays, but their feature point extraction accuracy is easily affected by defocus blur. During calibration, the two-dimensional target needs to be placed within the common field of view and common depth of field of the binocular camera, which limits the flexibility of calibration. Unlike traditional two-dimensional targets, phase targets have anti-blur properties, making them more suitable for calibration scenarios involving out-of-focus cameras. Commonly used phase targets include horizontal / vertical phase-shifting fringes and phase-shifting circular fringes arrays.

[0003] At least six fringe images are required for horizontal / vertical phase-shifted fringes, including one set of horizontal and one set of vertical phase-shifted fringes. The horizontal and vertical truncated phases are calculated using the phase-shifting method, and the zero-phase intersection points are extracted as calibration feature points. At least three fringe images are required for the phase-shifted circular fringe array. The radial truncated phase is calculated using the phase-shifting method, and zero-phase points are extracted to fit a zero-phase ellipse. The center of the ellipse is then used as the calibration feature point. However, when the binocular cameras cannot acquire complete target images, it is difficult to establish a unified world coordinate system to simultaneously calibrate the left and right cameras, leading to calibration task failure.

[0004] Invention patents 2024112684161 and 202511446030X propose using auxiliary coded images to distinguish various calibration feature points, thereby establishing a unified world coordinate system for calibrating binocular cameras. However, these methods require multiple auxiliary coded images, increasing the calibration workload. Therefore, exploring an efficient, flexible, and robust binocular camera calibration method is of great significance for promoting practical applications with small common fields of view or narrow depth of field. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and to propose a geometric calibration method for binocular cameras.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a geometric calibration method for a binocular camera, comprising the following steps: Step S1: Construct a binocular camera system, which includes two cameras of the same model and a high-definition LCD screen. Lenses with the same focal length are installed at the front of the two cameras. The two cameras are kept relatively fixed and the optical axes of the lenses are parallel to each other, so that the two cameras have a common field of view and a common depth of field. A phase-shift circular stripe array pattern is pre-designed, which includes several circular stripes of the same size but different phase shifts. Step S2: Display three phase-shifted circular stripe array patterns sequentially on the LCD screen. Place the LCD screen randomly in front of two cameras and trigger the LCD screen and the two cameras simultaneously to capture the three phase-shifted circular stripe array patterns sequentially, thus obtaining the corresponding three phase-shifted circular stripe array images. Step S3: Calculate the truncated phase and mask region of the three phase-shifted circular stripe array images using the three-step phase-shifting algorithm. Then, extract the mask region and truncated phase corresponding to each circular stripe using the connected component labeling algorithm. Finally, unwrap the truncated phase to obtain the absolute phase corresponding to each circular stripe. Step S4: Calculate the centroid position of the mask region for each circular stripe in sequence, extract the absolute phase value of the centroid position, and then decode the actual order of the corresponding circular stripes. Step S5: Sequentially extract the mask regions with phase values ​​less than or equal to π from the absolute phase of each circular stripe, then use the edge detection algorithm to extract the edge pixels of the mask region, then use the least squares method to fit the edge pixels into an ellipse, and extract the center of the ellipse as the corresponding center feature point. Step S6: Repeat steps S2-S5 three or more times to acquire phase-shifted circular stripe array images from three or more different orientations using a binocular camera, and extract the corresponding center feature points and actual order. Step S7: Based on the detection results of the center feature points, establish the geometric projection equations of the world coordinate system and its image coordinate system, and use Zhang Zhengyou's calibration method to calculate the internal and external parameters of the binocular camera.

[0007] Preferably, in step S1, the self-encoded phase-shifted circular stripe array pattern It can be represented as: In the formula Represents pixel coordinates; Indicates the actual order of each circular stripe; Indicates the radial period of the circular stripes; Indicates the first The radius of each circular stripe is expressed mathematically as follows: In the formula Indicates the first The centers of the circular stripes are spaced at the same horizontal or vertical intervals to form a uniformly distributed array of circular stripes.

[0008] Preferably, in step S2, the acquired phase-shifted circular fringe array image It can be represented as: In the formula Indicates background intensity. Indicates modulation intensity. This represents the truncated phase to be solved.

[0009] Preferably, in step S3, the truncated phase is calculated using a three-step phase shift algorithm. and mask area The mathematical expression is as follows: In the formula This represents the arctangent operation. This represents the Otsu thresholding function; its principle and implementation can be found in MATLAB. function.

[0010] Preferably, in step S3, the mask area corresponding to each circular stripe It can be represented as: In the formula Indicates the mask area The serial number of the marker. The function represents the connected component labeling algorithm; its principle and implementation can be found in MATLAB. function.

[0011] Preferably, in step S3, the absolute phase corresponding to each circular stripe It can be represented as: In the formula This represents the dot product operator. This represents the phase expansion function; its principle and implementation can be found in MATLAB. function.

[0012] Preferably, in step S4, the mask area of ​​the circular stripes center of gravity The calculation formula is as follows: , .

[0013] Preferably, in step S4, the actual order of the circular stripes... It can be represented as: .

[0014] Preferably, in step S5, the phase value is less than or equal to mask area It can be represented as: .

[0015] Preferably, in step S5, the edge detection algorithm is recommended to be the Sobel edge detection algorithm, the principle and implementation of which can be found in MATLAB. function.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the calibration efficiency is high, requiring only 3 phase-shifted circular fringe array images to complete the calibration, without the need for additional auxiliary encoded images, which greatly reduces the number of calibration images to be acquired and reduces the calibration workload; It has strong anti-interference ability and high robustness. Based on the phase shift method, it extracts phase information and realizes feature point detection. It has the anti-defocus blur characteristics of phase targets and can achieve high-precision calibration under defocus blur state. It has high calibration flexibility and wide applicability. It achieves unique encoding of feature points through self-encoded phase-shifted circular fringes. Even if the binocular camera only captures an incomplete target image, it can accurately decode the actual order of the circular fringes and establish a unified world coordinate system to complete the calibration. It is suitable for complex application scenarios with small common field of view and narrow depth of field. The feature point extraction accuracy is high. By extracting zero-phase edges through phase information and fitting ellipses to obtain the center feature points, the feature point extraction accuracy is higher than that of traditional checkerboard and circular array targets, thereby improving the calibration accuracy of the binocular camera. Attached Figure Description

[0017] Fig. 1 This is a three-image phase-shifted circular fringe array; Fig. 2The diagram shows the intermediate processing results of the present invention, where (a) is a truncated phase diagram, (b) is a mask region diagram, and (c) is an absolute phase diagram. Fig. 3 The image shows the feature point extraction results of the present invention, where (a) is a mask region image with a phase value less than or equal to π, (b) is an edge pixel image, and (c) is a feature point detection result image. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the 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 merely illustrative and not intended to limit the invention.

[0019] Please see Figs. 1-3 This invention provides a technical solution: a geometric calibration method for a binocular camera, comprising the following steps: Step S1: Construct a binocular camera system, which includes two cameras of the same model and a high-definition LCD screen. Lenses with the same focal length are installed at the front of the two cameras. The two cameras are kept relatively fixed and the optical axes of the lenses are parallel to each other, so that the two cameras have a common field of view and a common depth of field. A phase-shift circular stripe array pattern is pre-designed, which includes several circular stripes of the same size but different phase shifts. In step S1, the self-encoded phase-shifted circular stripe array pattern It can be represented as: In the formula Represents pixel coordinates; Indicates the actual order of each circular stripe; Indicates the radial period of the circular stripes; Indicates the first The radius of each circular stripe is expressed mathematically as follows: In the formula Indicates the first The centers of the circular stripes are spaced at the same horizontal or vertical intervals to form a uniformly distributed array of circular stripes. Step S2: Display three phase-shifted circular stripe array patterns sequentially on the LCD screen. Place the LCD screen randomly in front of two cameras and simultaneously trigger the LCD screen and the two cameras to capture the three phase-shifted circular stripe array patterns sequentially, thus obtaining the corresponding three phase-shifted circular stripe array images. In step S2, the acquired phase-shifted circular stripe array image It can be represented as: In the formula Indicates background intensity. Indicates modulation intensity. This represents the truncated phase to be solved; Step S3: Calculate the truncated phase and mask region of the three phase-shifted circular stripe array images using the three-step phase-shifting algorithm. Then, extract the mask region and truncated phase corresponding to each circular stripe using the connected component labeling algorithm. Finally, unwrap the truncated phase to obtain the absolute phase corresponding to each circular stripe. In step S3, the truncated phase is calculated using a three-step phase shift algorithm. and mask area The mathematical expression is as follows: In the formula This represents the arctangent operation. This represents the Otsu thresholding function; its principle and implementation can be found in MATLAB. function; In step S3, each circular stripe corresponds to a mask area. It can be represented as: In the formula Indicates the mask area The serial number of the marker. The function represents the connected component labeling algorithm; its principle and implementation can be found in MATLAB. function; In step S3, the absolute phase corresponding to each circular stripe It can be represented as: In the formula This represents the dot product operator. This represents the phase expansion function; its principle and implementation can be found in MATLAB. function; Step S4: Calculate the centroid position of the mask region for each circular stripe in sequence, extract the absolute phase value of the centroid position, and then decode the actual order of the corresponding circular stripes. In step S4, the mask area of ​​the circular stripes center of gravity The calculation formula is as follows: , ; In step S4, the actual order of the circular stripes It can be represented as: ; Step S5: Sequentially extract the mask regions with phase values ​​less than or equal to π from the absolute phase of each circular stripe, then use the edge detection algorithm to extract the edge pixels of the mask region, then use the least squares method to fit the edge pixels into an ellipse, and extract the center of the ellipse as the corresponding center feature point. In step S5, the phase value is less than or equal to mask area It can be represented as: ; In step S5, the Sobel edge detection algorithm is recommended. Its principle and implementation can be found in MATLAB. function; Step S6: Repeat steps S2-S5 three or more times to acquire phase-shifted circular stripe array images from three or more different orientations using a binocular camera, and extract the corresponding center feature points and actual order. Step S7: Based on the detection results of the center feature points, establish the geometric projection equations of the world coordinate system and its image coordinate system, and use the Zhang Zhengyou calibration method to calculate the internal and external parameters of the binocular camera. The principle of the Zhang Zhengyou calibration method can be found in the reference A flexible new technique for camera calibration [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22 (11):1330-1334. Its implementation method can be found in the calibrateCamera function in MATLAB.

[0020] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for geometric calibration of a binocular camera, characterized in that, Includes the following steps: Step S1: Construct a binocular camera system, which includes two cameras of the same model and a high-definition LCD screen. Lenses with the same focal length are installed at the front of the two cameras. The two cameras are kept relatively fixed and the optical axes of the lenses are parallel to each other, so that the two cameras have a common field of view and a common depth of field. A phase-shift circular stripe array pattern is pre-designed, which includes several circular stripes of the same size but different phase shifts. Step S2: Display three phase-shifted circular stripe array patterns sequentially on the LCD screen. Place the LCD screen randomly in front of two cameras and trigger the LCD screen and the two cameras simultaneously to capture the three phase-shifted circular stripe array patterns sequentially, thus obtaining the corresponding three phase-shifted circular stripe array images. Step S3: Calculate the truncated phase and mask region of the three phase-shifted circular stripe array images using the three-step phase-shifting algorithm. Then, extract the mask region and truncated phase corresponding to each circular stripe using the connected component labeling algorithm. Finally, unwrap the truncated phase to obtain the absolute phase corresponding to each circular stripe. Step S4: Calculate the centroid position of the mask region for each circular stripe in sequence, extract the absolute phase value of the centroid position, and then decode the actual order of the corresponding circular stripes. Step S5: Sequentially extract the mask regions with phase values ​​less than or equal to π from the absolute phase of each circular stripe, then use the edge detection algorithm to extract the edge pixels of the mask region, then use the least squares method to fit the edge pixels into an ellipse, and extract the center of the ellipse as the corresponding center feature point. Step S6: Repeat steps S2-S5 three or more times to acquire phase-shifted circular stripe array images from three or more different orientations using a binocular camera, and extract the corresponding center feature points and actual order. Step S7: Based on the detection results of the center feature points, establish the geometric projection equations of the world coordinate system and its image coordinate system, and use Zhang Zhengyou's calibration method to calculate the internal and external parameters of the binocular camera.

2. The geometric calibration method for a binocular camera according to claim 1, characterized in that: In step S1, the self-encoded phase-shifted circular stripe array pattern It can be represented as: In the formula Represents pixel coordinates; Indicates the actual order of each circular stripe; Indicates the radial period of the circular stripes; Indicates the first The radius of each circular stripe is expressed mathematically as follows: In the formula Indicates the first The centers of the circular stripes are spaced at the same horizontal or vertical intervals to form a uniformly distributed array of circular stripes.

3. The geometric calibration method for a binocular camera according to claim 1, characterized in that: In step S2, the acquired phase-shifted circular stripe array image It can be represented as: In the formula Indicates background intensity. Indicates modulation intensity. This represents the truncated phase to be solved.

4. The geometric calibration method for a binocular camera according to claim 1, characterized in that: In step S3, the truncated phase is calculated using a three-step phase shift algorithm. and mask area The mathematical expression is as follows: In the formula This represents the arctangent operation. This represents the Otsu thresholding function; its principle and implementation can be found in MATLAB. function.

5. The geometric calibration method for a binocular camera according to claim 1, characterized in that: In step S3, each circular stripe corresponds to a mask area. It can be represented as: In the formula Indicates the mask area The serial number of the marker. The function represents the connected component labeling algorithm; its principle and implementation can be found in MATLAB. function.

6. The geometric calibration method for a binocular camera according to claim 1, characterized in that: In step S3, the absolute phase corresponding to each circular stripe It can be represented as: In the formula This represents the dot product operator. This represents the phase expansion function; its principle and implementation can be found in MATLAB. function.

7. The method for geometric calibration of a binocular camera according to claim 1, characterized in that: In step S4, the mask area of ​​the circular stripes center of gravity The calculation formula is as follows: ， 。 8. The geometric calibration method for a binocular camera according to claim 1, characterized in that: In step S4, the actual order of the circular stripes It can be represented as: 。 9. The geometric calibration method for a binocular camera according to claim 1, characterized in that: In step S5, the phase value is less than or equal to mask area It can be represented as: 。 10. The geometric calibration method for a binocular camera according to claim 1, characterized in that: In step S5, the Sobel edge detection algorithm is recommended. Its principle and implementation can be found in MATLAB. function.

Images