A circular target design and detection method for multi-task scenarios

Abstract

The application provides a circular target design and detection method for a multi-task scene, which comprises image preprocessing, candidate contour extraction, affine transformation correction, circle contour identification, circular arc segment identification, center point positioning and encoding decoding; the target is composed of a center circle, a positioning ring and 12 circular arc segments, 12-bit binary codes are constructed through black and white binary circular arc segments, and rotation invariant identification is realized by adopting a minimum decimal numbering rule; in the identification process, high-precision center point positioning is realized through multi-circle contour weighted fusion, and the circular arc segments are screened according to angle and radius variance characteristics; the application introduces a multi-circle center weighted fusion positioning strategy in the center point calculation process of the circular coding mark target by designing multi-circle structure, including a center circle contour, a positioning ring and an encoding ring, so that the error caused by local contour recognition deviation can be effectively offset, and the center point positioning precision of the target is improved.

Description

Technical Field

[0001] This invention relates to the field of coded target detection technology, and in particular to a circular target design and detection method for multi-task scenarios. Background Technology

[0002] Encoded target detection is widely used in computer vision-related fields such as indoor test measurement, camera calibration, industrial vision inspection, and robot navigation. Especially in well-lit environments, it can provide a stable geometric reference for other tasks by automatically identifying and locating the target's center point. Therefore, obtaining the accurate coordinates of the target's center point is particularly important. To meet the needs of recognition efficiency and stability, the target design not only needs to have good detection characteristics, but also needs to carry distinguishable coded information to achieve multi-target recognition and numbering.

[0003] Currently, common coding marks are mainly divided into three categories: dot, square, and circular. Dot coding marks are generally composed of small solid circles and were initially used in optical 3D measurement and robot recognition scenarios. Some also use disconnected elliptical rings, which use redundant cyclic coding to ensure uniqueness. However, this type of scheme performs better under small viewing angle and close-range shooting conditions and has limited versatility. Square coding marks are more commonly found in augmented reality (AR) and planar machine vision systems. Circular coding marks, due to their good geometric symmetry, high resistance to rotation, and compact structure, are more favored in precision positioning tasks and are the current research focus of circular visual target design.

[0004] Although various encoded and non-encoded target detection methods exist in the present technology, they still have limitations in terms of multi-target recognition, target structure simplification, decoding stability, and production cost control.

[0005] Existing technology 1: Patent application number 202211047328.X, entitled "Target, Information Detection Method, Device, Terminal and Storage Medium", proposes a method for calculating the image center based on the intersection of the long axis of a circular ring and the base of a triangle. This scheme determines the center position of the target in the image by identifying the farthest pixel pair on the inner and outer circumferences and combining the triangular geometric structure, which can improve the efficiency of image data processing and the stability of target detection. However, this scheme has the following disadvantages: since the target used does not have encoding capabilities, it cannot realize multi-target number recognition, and it is only applicable to single target detection scenarios. Moreover, the target structure is relatively complex and the design is not universal, resulting in high production costs and limited promotion in practical applications.

[0006] Prior art 2: Patent application number 202111076814.X, entitled "A Target Plate, Target Pattern Detection Method and Device", discloses a target plate structure with a square blank area in the center and multiple circular or ring-shaped marking points arranged around it, which is suitable for camera geometric calibration tasks. This structure simplifies the calibration process by arranging non-collinear and unequally spaced ring center points. However, this solution is mainly used for overall target plate recognition and cannot accurately and independently identify or encode individual targets. Its applicable scenarios are limited to a single calibration purpose, and the structure is still complex and has a high manufacturing cost, making it unsuitable for large-scale or dynamic detection environments.

[0007] To address this, a circular target design and detection method for multi-task scenarios is proposed. Summary of the Invention

[0008] In view of this, the present invention provides a circular target design and detection method for multi-task scenarios to solve or alleviate the technical problems existing in the prior art, and at least provides a beneficial option.

[0009] The technical solution of this invention is implemented as follows: a method for designing and detecting circular targets for multi-task scenarios, comprising the following steps:

[0010] S1. Perform gamma illumination processing on the image to be recognized to equalize the brightness distribution of the image, and perform grayscale conversion and binarization on the processed image to obtain a binarized image for subsequent recognition.

[0011] The gamma illumination processing maps the brightness of the original image by setting a nonlinear enhancement parameter γ. When there are underexposed areas in the image, the contrast of the dark areas is improved by adjusting γ<1, which is used to enhance the edge contour of the coded target and preserve high-frequency details in the subsequent binarized image.

[0012] S2. Based on the binarized image, the Suzuki contour extraction algorithm is used to extract all contours in the image;

[0013] Candidate contours that conform to the structural features of the coded target are selected from the extracted contours. The candidate contours must have an approximately elliptical outer contour and an internal structure consisting of a central circle contour and multiple arc segment contours.

[0014] For each closed contour, determine whether it contains:

[0015] An external contour that meets the shape characteristics of a near circle or ellipse;

[0016] At least one central circle outline is used to form the center point;

[0017] At least eight circular arc segments with angular intervals close to the set standard are used for encoding and recognition;

[0018] Only when all of the above conditions are met will the closed contour be retained as a candidate contour for the coded marker target, and the remaining contours will be discarded.

[0019] S3. Perform affine transformation correction on the candidate contour to compensate for the distortion of the contour shape caused by the shooting angle or perspective deformation, so that the circular contour and the arc segment contour are restored to the approximate design state.

[0020] During the affine transformation correction of the candidate contour, the major axis, minor axis and tilt angle information of the outer contour point set of the candidate contour are calculated by the least squares ellipse fitting method. An affine transformation matrix is ​​established to perform geometric transformation on the overall contour image, so that the ellipse in the target structure is restored to an approximately standard circle, so as to eliminate the deformation interference caused by the shooting angle and retain the relative angle information of the internal arc structure.

[0021] S4. Based on the analysis of multiple contour features such as roundness, radius consistency and fitting residuals, the candidate contours after affine correction are identified and the contours that meet the set threshold are used as the center circle contour and positioning ring contour of the coded mark target.

[0022] The criteria for identifying the center circle profile and the positioning ring profile include: calculating the roundness of each profile. Where S represents the area enclosed by the contour, P represents the perimeter of the contour, and D represents the roundness of the contour.

[0023] When the roundness D is close to 1, it means that the contour is close to the ideal circle. At the same time, combined with the contour point fitting residual value, if the residual is lower than the preset threshold ε, the contour is determined to be a valid circular contour in the coded mark target structure.

[0024] S5. For each arc segment of the candidate contour, identify it based on the following features:

[0025] The included angle of the center of the arc segment contour is an integer multiple of a fixed angle, wherein the fixed angle in the 12-bit coded target is 30 degrees;

[0026] The outer radius of the arc segment contour is calculated from the distance from the contour point to the target center point, and the variance of the outer radius is less than a set threshold.

[0027] For each contour point, based on the relationship between its distance to the center point of the target, it is determined whether it belongs to the outer arc segment or the inner arc segment, and the valid arc segment contour is selected accordingly.

[0028] A circular arc segment is considered a valid coded circular arc segment only if it simultaneously meets all of the above feature recognition conditions.

[0029] The variance of the outer radius is calculated as follows: for all pixels on the contour of each arc segment, calculate the Euclidean distance from the center point of the target, and obtain the average radius R and variance σ. 2 , when the variance σ 2 When the value is less than the threshold θ, it indicates that the arc segment has a consistent outer contour curvature in space;

[0030] The method used to determine whether a contour point belongs to the outer or inner arc segment is as follows:

[0031] Draw a ray from the center point of the target along the direction of each contour point, and record the two intersection points of the ray and the contour. If the distance between the current point and the center point is greater than the distance of the other intersection point, it is determined to be an outer arc segment.

[0032] When at least 80% of the contour points in the same arc segment contour meet the outer arc segment conditions, the contour is retained as the coded arc segment contour.

[0033] S6. Using the geometric center coordinates of the central circle contour, the positioning ring contour, and the coding ring contour, a weighted fusion algorithm is used to calculate the target center point;

[0034] The center coordinates of the smallest circumcircle of the positioning ring contour and the coding ring contour are calculated respectively. The three are then weighted and averaged according to a preset weighting coefficient to calculate the center point coordinates of the coding mark target.

[0035] S7. For the identified coded arc segment contour, read the color information of each arc segment sequentially from any starting position in a clockwise direction, setting black to represent "1" and white to represent "0", and construct a 12-bit binary code;

[0036] Perform a cyclic shift on the 12-bit binary code to generate all 12 combinations, convert each combination into a decimal number, and select the smallest decimal value as the unique identification code value of the code flag target;

[0037] The order in which the color information of the arc segment is read is determined by sorting the polar angles corresponding to the lines connecting the center point of the target to the center point of each arc segment from smallest to largest; among all the cyclic displacement combinations of the 12-bit code, the corresponding minimum decimal value is selected as the unique number of the final coded target.

[0038] The embodiments of the present invention have the following advantages due to the adoption of the above technical solutions:

[0039] I. This invention designs a multi-ring circular structure in a circular coded target, including a central circular outline, a positioning ring, and a coding ring. By introducing a multi-center weighted fusion positioning strategy during the target center point calculation process, it can effectively offset the error caused by local contour recognition deviation and improve the center point positioning accuracy of the target.

[0040] Second, this invention designs the coding ring as consisting of 12 high-density, equally angularly distributed arc segments, each of which can encode binary values. These segments are combined to form complete coding information. Furthermore, a rotation-invariant minimum decimal coding strategy is employed to achieve unique identification of the target in any orientation. This not only supports simultaneous detection and identification of multiple targets but also ensures the stability and robustness of the target numbering, significantly expanding the practical application capabilities of this target structure in multiple scenarios and tasks.

[0041] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of the steps of the present invention;

[0044] Figure 2 This is an example diagram of the coded target of the present invention. Detailed Implementation

[0045] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0046] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0047] like Figure 1-2 As shown, this embodiment of the invention provides a circular target design and detection method for multi-task scenarios, which mainly includes two core components: one is the structural design of the coded target, and the other is the processing flow of image recognition, information positioning and encoding / decoding of the target.

[0048] In terms of target structure design, the coded target used in this invention has a ring structure, which is composed of two types of geometric elements: a circular structure for center positioning and an arc coding structure for information expression. The coding ring is composed of 12 equally divided arc segments, each arc segment covering a 30-degree angle. The color adopts a black and white binary coding form, where black represents binary "1" and white represents "0". After being arranged in sequence, it forms a 12-bit binary number. Since this kind of ring arrangement may have uncontrollable rotation angle during the shooting process;

[0049] This invention employs a cyclic displacement matching method, selecting the smallest decimal value from 12 possible starting positions as the unique identifier ID. For example, when the code "001100110000" is read, all its rotation combinations will be automatically converted to decimal and the smallest value "102" will be selected as the identification result, ensuring the uniqueness and rotation invariance of the extracted code. The coding ring has a central circle and a positioning ring inside, which are used to assist in locating the geometric center of the target and enhance the positioning accuracy of the center point.

[0050] Moving on to the recognition and decoding section, the entire process consists of seven steps, specifically including:

[0051] Step S1 is image preprocessing. Due to uneven lighting and abnormal grayscale distribution in the captured images, this invention performs brightness equalization processing on the original images using the Gamma correction method. By setting a γ value (usually less than 1), the image is nonlinearly enhanced to make the details in the dark areas more prominent, which is beneficial for edge detection and contour recognition. Subsequently, the adjusted image is converted into a grayscale image and binarized using an adaptive or fixed threshold method to obtain a high-contrast black and white image, providing clear basic data for subsequent contour extraction.

[0052] Step S2 is candidate contour extraction. This step uses the boundary tracking algorithm proposed by Suzuki for contour detection. This algorithm is based on pixel adjacency and scans all closed regions in the image layer by layer to obtain a complete contour list. Since there are various noises or non-coded graphic interferences in the scene, candidate coded target contours need to be selected from all contours. The selection criteria include: the outer contour is close to a circle, the internal structure contains a central circle contour, and there are at least 8 regularly distributed arc segments. Contours that do not meet the above characteristics will be directly excluded.

[0053] Step S3 is affine transformation correction. This step uses ellipse fitting and affine matrix correction. Specifically, ellipse least squares fitting is performed on the outer contour to obtain the parameters of the major axis, minor axis, and inclination angle. Then, based on these parameters, an affine transformation matrix is ​​established to correct the entire candidate target area, so that the arcs and circular structures in the contour are closer to their original design shape after correction, thereby improving the geometric consistency and accuracy of subsequent recognition.

[0054] Step S4 is the circle contour recognition. This step identifies the central circle and positioning ring contours in the target structure from the corrected candidate contours. The main recognition indicators include:

[0055] The roundness of the profile (calculated using the formula D = 4πS / P) 2 The fit residual value (i.e., the average distance from the contour point to the smallest fitted circle) and the contour radius are consistent. When a contour simultaneously meets the conditions that the roundness is close to 1, the fit residual is lower than the set threshold, and the radius variance is stable, it can be identified as a valid circular contour.

[0056] Step S5 is the identification of the arc segment contour. The criteria for judging the structural features of the arc segment in this step are as follows:

[0057] 1) The included angle of each arc segment must be an integer multiple of a fixed value, which is 30 degrees for a 12-bit target;

[0058] 2) The variance of the outer radius should be stable, that is, the set of distances from the center point of the target to each pixel on the arc segment should have a small variance;

[0059] 3) The outline should have good continuity, that is, there should be no large breaks or discrete points;

[0060] 4) The outer / inner side of each arc segment must be definitive, that is, the outer arc and the inner arc can be distinguished by the distance between the two intersection points of the ray drawn from the center to the contour;

[0061] 5) The number of arc segments should be consistent with the number of bits in the code, and the angular arrangement should meet the requirements of polar coordinate symmetry;

[0062] Arc segments that meet all the conditions will be recorded and sorted as input for encoding.

[0063] Step S6 is the target center point localization. The present invention adopts a multi-circle center fusion strategy, that is, calculates the coordinates of the minimum circumscribed circle center of the center circle contour, the positioning ring contour and the coding ring contour respectively, and performs a weighted average according to a preset weight. Since the single contour recognition is offset by local noise or deformation, by fusing multi-layer geometric center information, the error is effectively suppressed, a more accurate target geometric center is obtained, and the recognition stability and positioning accuracy are improved.

[0064] Step S7 is the decoding of the encoded value. After the arc segment recognition and sorting are completed, starting from any arc segment in a clockwise direction, its color value is read in sequence, and black is set to "1" and white is set to "0", forming a complete 12-bit binary number. Then, 12 combination forms are generated in sequence through cyclic displacement, and each is converted into a decimal number. Finally, the smallest value is selected as the unique encoded value of the target. The biggest advantage of this encoding method is that it has natural invariance to rotation transformation. No matter what the image shooting angle is, the extracted encoded value remains logically consistent.

[0065] Application example:

[0066] Suppose a camera captures an image, and after Gamma correction, a clear image is obtained. After Suzuki contour extraction, a candidate contour containing a central circle and 12 arc segments is selected. Affine transformation restores the contour structure to a standard shape. Contour detection identifies the central circle contour, the positioning ring contour, and the 12 equally spaced arc segments. Further, the arc segments are selected and encoded according to five geometric features. At this point, the encoding is "100100110001". Twelve combinations are generated in a loop, and the corresponding decimal values ​​are calculated. The minimum value "73" is selected as the unique identification number, and the entire identification process is completed.

[0067] The following is a description of some of the technical terms and key algorithms involved in this embodiment:

[0068] Gamma illumination processing: Gamma illumination processing is a non-linear image brightness transformation method used to adjust the grayscale contrast of bright and dark areas in an image.

[0069] The Suzuki contour extraction algorithm is a boundary tracking method for contour extraction in binary images. Proposed by Suzuki and Abe, it recursively constructs boundary layers through pixel adjacency relationships, which can completely extract the contour structure of closed regions. It is often used in the contour recognition module of computer vision libraries such as OpenCV.

[0070] Affine transformation: Affine transformation is a two-dimensional geometric transformation that preserves the linearity and parallelism between points. It is used to perform translation, rotation, scaling and tilting. In this invention, it is used to restore the elliptical distortion caused by the shooting angle to a circular outline.

[0071] Ellipse least squares fitting: used to fit the geometric structure of an ellipse to a set of contour points, minimizing the sum of squared errors to obtain the major axis, minor axis, center, and inclination parameters of the ellipse. The results are used to generate an affine transformation matrix to achieve contour affine correction.

[0072] Contour fitting residual: Used to measure the average distance error between contour points and their fitted curve (such as a circle). The smaller the residual, the closer the fitting result is to the actual structure. It is used to help determine the effectiveness of the circular contour.

[0073] Outer radius variance: This parameter is calculated by taking the distance from all pixels on the arc segment to the target center and calculating its variance. It characterizes the consistency of the arc segment's curvature in space. A small variance indicates that the shape is regular and stable, and it is the basis for selecting effective arc segment contours.

[0074] Polar Angle Sorting: During the encoding and reading process, the polar angle (polar coordinate angle) values ​​of the center point of each arc segment relative to the center point of the target are sorted from smallest to largest to standardize the clockwise order of encoding extraction and avoid encoding errors caused by disordered order.

[0075] Cyclic displacement matching: The decoding strategy designed in this invention to solve the target rotation problem is to perform cyclic displacement of all 12 starting bits on the 12-bit binary code read, and select the minimum value as the unique number after converting it to decimal.

[0076] Grayscale threshold determination: During the color extraction process of the arc segment, the grayscale of local area pixels is statistically analyzed using the average grayscale or adaptive threshold method. Based on the threshold, the color of the current arc segment is determined to be "black" (1) or "white" (0) to construct a complete binary code.

[0077] Weighted fusion algorithm: During the center point localization process, multiple circles (center circle, positioning ring, encoding ring) are used to fit the center point coordinates and each is assigned a preset weight. The weighted average is then used to obtain the final fused center point position, suppressing the overall positioning error caused by local contour offset.

[0078] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for designing and detecting circular targets in multi-task scenarios, characterized in that: Includes the following steps: S1. Perform gamma illumination processing on the image to be recognized to equalize the brightness distribution of the image, and perform grayscale conversion and binarization on the processed image to obtain a binarized image for subsequent recognition. S2. Based on the binarized image, the Suzuki contour extraction algorithm is used to extract all contours in the image; Candidate contours that conform to the structural features of the coded target are selected from the extracted contours. The candidate contours must have an approximately elliptical outer contour and an internal structure consisting of a central circle contour and multiple arc segment contours. S3. Perform affine transformation correction on the candidate contour to compensate for the distortion of the contour shape caused by the shooting angle or perspective deformation, so that the circular contour and the arc segment contour are restored to the approximate design state. S4. Based on the analysis of multiple contour features such as roundness, radius consistency and fitting residuals, the candidate contours after affine correction are identified and the contours that meet the set threshold are used as the center circle contour and positioning ring contour of the coded mark target. S5. For each arc segment of the candidate contour, identify it based on the following features: The included angle of the center of the arc segment contour is an integer multiple of a fixed angle, wherein the fixed angle in the 12-bit coded target is 30 degrees; The outer radius of the arc segment contour is calculated from the distance from the contour point to the target center point, and the variance of the outer radius is less than a set threshold. For each contour point, based on the relationship between its distance to the center point of the target, it is determined whether it belongs to the outer arc segment or the inner arc segment, and the valid arc segment contour is selected accordingly. A circular arc segment is considered a valid coded circular arc segment only if it simultaneously meets all of the above feature recognition conditions. S6. Using the geometric center coordinates of the central circle contour, the positioning ring contour, and the coding ring contour, a weighted fusion algorithm is used to calculate the target center point; S7. For the identified coded arc segment contour, read the color information of each arc segment sequentially from any starting position in a clockwise direction, setting black to represent "1" and white to represent "0", and construct a 12-bit binary code; Perform a cyclic shift on the 12-bit binary code to generate all 12 combinations, convert each combination into a decimal number, and select the smallest decimal value as the unique identification code value of the encoding target.

2. The method for designing and detecting circular targets in multi-task scenarios according to claim 1, characterized in that: In step S1, the gamma illumination processing performs brightness mapping on the original image by setting a nonlinear enhancement parameter γ. When there are underexposed areas in the image, the contrast of the dark areas is improved by adjusting γ<1, which is used to enhance the edge contour of the coded target and preserve high-frequency details in the subsequent binarized image.

3. The method for designing and detecting circular targets in multi-task scenarios according to claim 1, characterized in that: In step S2, for each closed contour, it is determined whether it simultaneously contains: An external contour that meets the shape characteristics of a near circle or ellipse; At least one central circle outline is used to form the center point; At least eight circular arc segments with angular intervals close to the set standard are used for encoding and recognition; Only when all of the above conditions are met will the closed contour be retained as a candidate contour for the coded marker target, and the remaining contours will be discarded.

4. The method for designing and detecting circular targets in multi-task scenarios according to claim 1, characterized in that: In step S3, during the affine transformation correction of the candidate contour, the major axis, minor axis and tilt angle information of the outer contour point set of the candidate contour are calculated by the least squares ellipse fitting method. An affine transformation matrix is ​​established to perform geometric transformation on the overall contour image, so that the ellipse in the target structure is restored to an approximately standard circle, so as to eliminate the deformation interference caused by the shooting angle and retain the relative angle information of the internal arc structure.

5. The method for designing and detecting circular targets in multi-task scenarios according to claim 1, characterized in that: In step S4, the criteria for identifying the central circle contour and the positioning ring contour include: calculating the roundness D of each contour D = 4πS / P2, where S represents the area enclosed by the contour, P represents the perimeter of the contour, and D represents the roundness of the contour. When the roundness D is close to 1, it means that the contour is close to the ideal circle. At the same time, combined with the residual value of the contour point fitting, if the residual is lower than the preset threshold ε, the contour is determined to be a valid circular contour in the coded mark target structure.

6. The method for designing and detecting circular targets in multi-task scenarios according to claim 1, characterized in that: In step S5, the variance of the outer radius is calculated as follows: for all pixels on the contour of each arc segment, the Euclidean distance from the pixel to the center point of the target is calculated, and the average radius R and variance σ are obtained. 2 , when the variance σ 2 When the value is less than the threshold θ, it indicates that the arc segment has a consistent outer contour curvature in space.

7. The method for designing and detecting circular targets in multi-task scenarios according to claim 1, characterized in that: In step S5, the method for determining whether a contour point belongs to an outer or inner arc segment is as follows: Draw a ray from the center point of the target along the direction of each contour point, and record the two intersection points of the ray and the contour. If the distance between the current point and the center point is greater than the distance of the other intersection point, it is determined to be an outer arc segment. When at least 80% of the contour points in the same arc segment contour meet the outer arc segment conditions, the contour is retained as the coded arc segment contour.

8. The method for designing and detecting circular targets in multi-task scenarios according to claim 1, characterized in that: In step S6, the center coordinates of the minimum circumcircle of the positioning ring profile and the coding ring profile are calculated respectively, and the three are weighted and averaged according to a preset weighting coefficient to calculate the center point coordinates of the coding mark target.

9. The method for designing and detecting circular targets in multi-task scenarios according to claim 1, characterized in that: In step S7, the order in which the color information of the arc segment is read is determined by sorting the polar angles corresponding to the lines connecting the center point of the target to the center point of each arc segment from smallest to largest.

10. The method for designing and detecting circular targets in multi-task scenarios according to claim 1, characterized in that: In step S7, the minimum decimal value among all the cyclic shift combinations of the 12-bit code is selected as the unique number of the final encoded flag target.

Images