Design and automatic identification method of control mark for deformation measurement of ultra-long blade

Abstract

The invention belongs to the technical field of wind power blade deformation measurement, and provides a control mark design and automatic identification method for solving the problems of low manual control point processing efficiency and high error rate in super-long blade vision measurement, which comprises the following steps of: designing a composite control mark integrated with a digital number and a positioning mark; the method comprises the following steps of: firstly, realizing mark positioning under a complex working condition based on gradient template matching and a feature grading method; then, through straight line detection screening and intersection point calculation, mark center sub-pixel coordinates are extracted; and finally, segmenting a digital region by using contour circle fitting, and analyzing the serial number by combining adaptive threshold binaryzation and an optical character recognition technology. According to the method, synchronous automatic acquisition of the image point coordinates and numbers of the control points is realized, the manual intervention requirement is remarkably reduced, and the measurement reliability is improved.

Description

Technical Field

[0001] The present invention relates to the field of deformation measurement of wind turbine blades, and particularly provides a method for visual deformation measurement of super-long blades, including control mark design, automatic mark recognition, and automatic control point extraction. Background Art

[0002] Conducting tensile deformation testing on wind turbine blades is an essential measure to ensure blade safety, improve blade efficiency, and extend blade service life. During the test, blade deformation needs to be measured in real time. When using visual methods to measure blade deformation, the existing technology usually uses manual methods to identify control points on the image, and corresponds the actual three-dimensional coordinates of the control points with the image point coordinates on the image. However, the deformation measurement of ultra-long blades requires the deployment of multiple industrial cameras and a large number of control points. The identification of control points and the correspondence between the actual coordinates of the control points and the image point coordinates on the image captured by each camera, as well as the extraction of image point coordinates, greatly increase the workload of deformation measurement and are prone to errors. In response to this situation, the present invention provides a method for control mark design, control mark recognition, and image point coordinate extraction, which can automatically recognize control marks on images, extract the image point coordinates of control mark points, and correspond to the corresponding three-dimensional coordinates, thereby saving the workload of blade deformation measurement and providing strong support for automated blade tensile deformation measurement. Summary of the Invention

[0003] In view of the above technical deficiencies, the present invention proposes a control mark design and automatic recognition method for ultra-long blade deformation measurement.

[0004] A design and automatic identification method of control marks for ultra-long blade deformation measurement specifically includes the following steps: S1. Use Figure 2 The control marks shown are laid out in the order of numbers, and the three-dimensional actual position of the mark center point is observed with a total station; multiple cameras are laid out to observe the entire control field; S2. For each camera’s observation of the control field image, use the gradient-based template matching and feature classification method to locate the mark, and save the center position of the located mark. , and the width and height of the positioning box ; S3. Center point extraction: Automatically extract the coordinates of the center image point of the positioning mark area through Hough line detection, edge screening, line segment merging and longest line intersection; S4. Identification of control mark numbers: The digital numbers of control marks are automatically parsed through circle fitting to segment the digital area, adaptive Otsu threshold binarization and OCR recognition.

[0005] In S3, the method for extracting the center point is:

[0006] The positioning frame area is selected as the region of interest P, and straight line segments are extracted within P. The straight line segments whose two endpoints are very close to the same edge of P are deleted.

[0007] The straight line fitting method is used to merge straight line segments with approximately the same direction and two endpoints close to each other;

[0008] After merging, the straight line segments are sorted by length, the two longest straight line segments are selected, and the intersection of the two straight line segments is calculated. The coordinates of the intersection are the coordinates of the control point image point.

[0009] In S4, the identification of the control flag number specifically includes:

[0010] Move the logo positioning box to the left Unit, locate the control mark number, take the translated positioning frame area as the region of interest Q, and perform contour extraction within Q;

[0011] For the extracted contour line, a random sampling consistency algorithm is used to perform circle fitting to obtain a circular curve M; the grayscale value of the points inside the region of interest Q, on M, and outside M is assigned to 255;

[0012] Calculate the Otsu threshold of the grayscale value of the points in M, and use this Otsu threshold as the threshold to binarize the region Q. Identify the control mark number in the binarized region.

[0013] The present invention provides a control mark design and automatic identification method specifically for ultra-long blade deformation measurement, effectively addressing the need to process massive control points in this field. Through a unique composite mark structure (combining numbering and positioning marks) and a supporting automatic identification process (including positioning, center point extraction, and numbering), the efficiency and accuracy of acquiring control point information during deformation measurement are significantly improved. This method significantly reduces the workload and potential error risks associated with traditional manual identification and coordinate mapping, and improves recognition robustness in complex industrial field environments (such as curved surface reflections and background interference). BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 Flowchart of the present invention; Figure 2 This is the control logo diagram of the present invention; Figure 3 This is the control logo diagram after positioning of the present invention; Figure 4 The region of interest Q of the present invention; Figure 5 This is the control symbol diagram after circle fitting of the present invention. DETAILED DESCRIPTION

[0015] S1. Design of control signs, including:

[0016] like Figure 2 As shown in the figure, the control mark consists of two parts: the left side is the control point number, which is used to distinguish and identify different control points; the right side is the mark used to locate the control point, and the coordinates of the center are the control point coordinates. When measuring the deformation of an extra-long blade, a large number of control marks must be deployed. These control marks are arranged in ascending order according to the control point number to facilitate logical checking of incorrectly identified marks.

[0017] S2. Use grayscale gradient-based template matching and feature classification methods to achieve accurate positioning of control marks. The specific solution is as follows:

[0018] The template matching similarity measurement function based on grayscale gradient is shown as follows: (1).

[0019] In the formula, T represents the template image, O represents the target image, is the position in the template image The gradient direction in radians at is the target image relative to Pan The gradient direction of the position after is expressed in radians. For each template, a list is used to store the position and gradient information, recorded as .

[0020] Extract feature points from a given template T. After extracting the feature points, calculate the feature point strength as follows: (2).

[0021] In the formula is the autocorrelation matrix, defined as: (3).

[0022] and is the gradient of the image in the x and y directions, det and trace represent the determinant and trace of the matrix respectively. k is an empirical constant set to 0.05.

[0023] Two feature strength thresholds are set based on R (here we choose 30 and 60 based on experience). Feature points with strengths below 30 are considered weak features and do not participate in matching, thereby filtering out less significant feature points and reducing the possibility of false matching; feature points with strengths greater than 60 are considered strong features and are used first in the matching process;

[0024] Taking into account the rotation and scale issues of the object to be detected in the target image, the template is rotated every 1 degree between 0 and 360 degrees, and the template is upsampled and downsampled respectively. Template matching and feature point extraction are performed on the rotated, upsampled and downsampled images;

[0025] A gradient-based template matching method is used to find regions in the target image that are similar to the template. The template matching function is used to calculate the similarity between the template and the image patch. This similarity is used to find the initial matching position, and then feature points are used to achieve fine positioning of the test image.

[0026] S3. Use the line intersection method to extract the coordinates of the center point of the mark. The specific solution is:

[0027] For the positioned control mark ( Figure 3 ), let the center coordinates of the red positioning frame be , width and height are , select the area within the red positioning frame as the region of interest P;

[0028] The Hough line detection algorithm is used to detect straight line segments in the region of interest P, and the straight line segments whose two endpoints are less than a certain threshold distance from the same edge of the region of interest P are deleted (here the threshold is set to 1 / 5 of the length of the region of interest P);

[0029] For the remaining detected straight line segments, the angle between the directions is less than a certain threshold (the threshold is set here radians) and where the distance between a pair of endpoints is less than a threshold (here the threshold is set to 1 / 20 of the length of the region of interest P) are merged. The merging method is to perform a straight line fit on the endpoints of the line segments that meet the merging conditions, and use the fitted line as the merged line. The endpoints of the line segments involved in the merger are projected onto the merged line, and the two projected points with the largest distance between them are the two endpoints of the merged line segment.

[0030] Select the two straight line segments with the longest length from the merged straight line segments, calculate the intersection of the two straight line segments, and the coordinates of the intersection are the coordinates of the image point of the center point of the mark.

[0031] S4. Identification of control mark number, the specific scheme is as follows:

[0032] Will Figure 3 The center coordinate c of the middle positioning frame moves to the left To point n, select n as the center, the width and height are The area is the area of ​​interest Q;

[0033] Perform contour detection within the region of interest Q and extract all contour points;

[0034] Apply the random sampling consistency algorithm to perform circle fitting on all contour points and fit the circle M;

[0035] In the region of interest Q, the grayscale values ​​of all points on and outside the circle M are assigned a value of 255;

[0036] In the region of interest Q, the grayscale values ​​of the points falling inside the circle M are counted, and the Otsu threshold of the grayscale values ​​of the points falling inside the circle M is calculated;

[0037] Using the Otsu threshold as the threshold, the region of interest Q is binarized;

[0038] The Tesseract digit recognition library is used to perform digit recognition in the region of interest Q.

[0039] The above description is only 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 in the scope of protection of the present invention.

Claims

1. A design and automatic identification method of control marks for deformation measurement of ultra-long blades, characterized by The following steps are involved: S1. Use the control markers shown in Figure 2 to lay out the control field in numerical order, and use a total station to observe the three-dimensional coordinates of the marker center. Deploy multiple cameras to observe the entire control field. S2. For each camera observing the control field image, use the gradient-based template matching and feature classification method to locate the mark, and save the center position of the located mark , and the width and height of the positioning box ; S3. Center point extraction: Automatically extract the coordinates of the center image point of the positioning mark area through Hough line detection, edge screening, line segment merging and longest line intersection; S4. Identification of control mark numbers: The digital numbers of control marks are automatically parsed through circle fitting to segment the digital area, adaptive Otsu threshold binarization and OCR recognition.

2. A method for designing and automatically identifying control marks for ultra-long blade deformation measurement according to claim 1, characterized in that: In the above S3, the method for extracting the center point is: Select the positioning frame area as the region of interest P, extract straight line segments within P, and delete the straight line segments whose two endpoints are very close to the same edge of P; The straight line fitting method is used to merge straight line segments with approximately the same direction and two endpoints close to each other; After merging, the straight line segments are sorted by length, the two longest straight line segments are selected, and the intersection of the two straight line segments is calculated. The coordinates of the intersection are the coordinates of the control point image point.

3. The method for designing and automatically identifying control marks for ultra-long blade deformation measurement according to claim 1 is characterized by: In the above-mentioned S4, the identification of the control mark number specifically includes: Move the logo positioning box to the left Unit, locate the control mark number, take the translated positioning box area as the region of interest Q, and perform contour extraction within Q; For the extracted contour line, a random sampling consistency algorithm is used to fit the circle to obtain the circular curve M. The grayscale value of the points inside the region of interest Q, on M and outside M is assigned to 255; Calculate the Otsu threshold of the grayscale value of the points in M, and use this Otsu threshold as the threshold to binarize the region. Identify the control mark number in the binarized region.

Images