Steel ladle cover feeding hole detection method and system for steel ladle lining detection and medium

Abstract

The invention discloses a steel ladle cover feeding hole detection method and system for steel ladle lining detection and a medium. The detection method comprises the steps that a real-time image of a steel ladle cover feeding hole is collected, and contour features are extracted; pixel matching is carried out on the plurality of contour features and a template image of the steel ladle cover feeding hole, threshold judgment is carried out on the similarity of pixel points in the plurality of similarity point sets, the ratio of the number of pixel points larger than a threshold in the number of pixel points in the template image is calculated, and the similarity point set with the highest ratio is extracted as a target contour feature; extracting a position corresponding to the highest similarity from the target contour features as a central position of the target contour features; extracting at least three feature points with the highest similarity from the target contour features, and matching standard points corresponding to the feature points in the template image; and performing affine transformation on the feature points, and controlling the robot to drive a probe to extend into the steel ladle from the steel ladle cover feeding hole to perform lining detection according to the central position and the pose information.

Description

Technical Field

[0001] The invention relates to the technical field of ladle thermal repair, and in particular to a ladle cover feeding hole detection method, system and medium for ladle lining detection. Background Art

[0002] During the hot repair of the ladle, the detection of the ladle lining mainly relies on traditional manual visual observation. When performing the detection, the ladle cover needs to be opened. Due to the high temperature at the ladle mouth, the observer needs to stand at a distance from the ladle mouth for observation. Therefore, it is impossible to perform a comprehensive and accurate detection of the ladle lining. There are also some methods of using a high-temperature probe to detect the ladle lining. By connecting the high-temperature probe with a longer connecting rod, the observer can hold the connecting rod and extend the high-temperature probe from the ladle cover feeding hole into the ladle for lining detection. However, the manual holding of the connecting rod cannot ensure the smooth movement of the high-temperature probe, thereby affecting the detection accuracy of the high-temperature probe. With the development of Industry 4.0, the manual holding of the connecting rod can be replaced by a robot. However, the position and angle of the ladle will change during the hot repair of the ladle. Therefore, it is necessary to manually mark the position of the ladle cover feeding hole in real time so that the robot can detect the ladle lining according to the marked position, resulting in the robot being unable to perform the fully automated operation of the ladle lining detection, affecting the efficiency of the ladle lining detection. In view of this, the present invention is proposed. Summary of the invention

[0003] In order to solve the problem that the existing ladle lining detection cannot realize full-automatic operation, the present invention provides a ladle cover feeding hole detection method, system and medium for ladle lining detection.

[0004] In order to solve the above technical problems, the present invention provides a method for detecting a feeding hole of a ladle cover for detecting a ladle lining, the detection method comprising the steps of:

[0005] Collect the real-time image of the feeding hole of the ladle cover and extract the contour features contained in the real-time image;

[0006] Perform pixel matching on the multiple contour features and the template image of the ladle cover feeding hole to obtain a similarity point set of the multiple contour features;

[0007] Perform threshold judgment on the similarity of pixels in multiple similarity point sets, calculate the ratio of the number of pixels greater than the threshold to the number of pixels in the template image, and extract the similarity point set with the highest ratio as the target contour feature;

[0008] The position corresponding to the highest similarity is extracted from the target contour feature as the center position of the target contour feature;

[0009] Extract at least three feature points with the highest similarity from the target contour features, and match the corresponding standard points in the template image;

[0010] Combining the affine transformation with the coordinates of the standard points, the feature points are affine transformed to obtain the position and posture information of the transformed target contour features;

[0011] The robot is controlled to drive the probe from the feeding hole of the ladle cover into the interior of the ladle for lining inspection according to the center position and posture information.

[0012] In an embodiment of the present invention, in the step of performing affine transformation on the feature points in combination with the affine transformation and the coordinates of the standard points to obtain the position and posture information of the transformed target contour features, the method further comprises:

[0013] Performing affine transformation on at least three feature points with the highest similarity and corresponding standard points using an affine transformation matrix, and calculating the affine transformation matrix;

[0014] The target contour features are affine transformed using an affine transformation matrix, and the position information and angle information of the transformed target contour features relative to the template image are obtained.

[0015] In an embodiment of the present invention, the at least three extracted feature points with the highest similarity include a center point of a target contour feature and at least two edge points of the target contour feature.

[0016] In an embodiment of the present invention, the affine transformation formula is:

[0017]

[0018] Wherein, M is a 2x3 affine transformation matrix, and the affine transformation of the target contour feature includes operations of rotation, translation and scaling of the target contour feature.

[0019] In an embodiment of the present invention, similarity point sets of multiple contour features are calculated respectively by using a calculation formula of a normalized mutual correlation coefficient, and the calculation formula of the normalized mutual correlation coefficient is:

[0020]

[0021] Where T(i, j) is the pixel value of the template image, I(x+i, y+j) is the local pixel value of the contour feature at position (x, y) in the real-time image, and are the average values ​​of template image and contour features respectively.

[0022] In an embodiment of the present invention, the step of performing a threshold judgment on the similarity of pixels in a plurality of similarity point sets, calculating the ratio of the number of pixels greater than the threshold to the number of pixels in the template image, and extracting the similarity point set with the highest ratio as the target contour feature, further comprises the steps of:

[0023] The similarity of the pixels in each similarity point set is compared with the threshold value, and the number of pixel similarities greater than the threshold value is screened out in each similarity point set;

[0024] Calculate the proportion of the number of pixel similarities in the corresponding similarity point set;

[0025] The similarity point set with the highest proportion is extracted as the target contour feature.

[0026] In an embodiment of the present invention, in the step of extracting the contour features contained in the real-time image, the step of:

[0027] Use Gaussian filtering to perform Gaussian smoothing on real-time images;

[0028] Convolution of the real-time image with a two-dimensional Gaussian is performed to remove noise from the real-time image;

[0029] The Sobel operator is used to calculate the horizontal and vertical gradients of the real-time image, and the non-maximum suppression method is used to remove non-edge points in the real-time image;

[0030] The double threshold processing method is used to detect the strong edge and weak edge of the real-time image, and the contour features in the real-time image are detected by combining the strong edge and the weak edge.

[0031] In an embodiment of the present invention, in the step of controlling the robot to drive the probe from the ladle cover feeding hole into the ladle to perform lining detection according to the center position and posture information, the step further includes the following steps:

[0032] Convert the center position to the target coordinates in robot base coordinates;

[0033] Control the robot to generate motion trajectory according to target coordinates and posture information;

[0034] The robot is controlled to drive the probe to extend from the feeding hole of the ladle cover into the interior of the ladle according to the motion trajectory, and drive the probe to perform rotational scanning on the inner lining of the ladle.

[0035] In order to solve the above technical problems, the present invention also provides a ladle cover feeding hole detection system for ladle lining detection, the detection system comprising:

[0036] An image acquisition module is configured to acquire a real-time image of the charging hole of the ladle cover and extract contour features contained in the real-time image;

[0037] A pixel matching module; configured to perform pixel matching on a plurality of contour features and a template image of a ladle cover feeding hole to obtain a similarity point set of the plurality of contour features;

[0038] A threshold judgment module is configured to perform a threshold judgment on the similarity of the pixels in the plurality of similarity point sets, calculate the ratio of the number of pixels greater than the threshold to the number of pixels in the template image, and extract the similarity point set with the highest ratio as the target contour feature;

[0039] A center extraction module is configured to extract a position corresponding to the highest similarity in the target contour feature as the center position of the target contour feature;

[0040] A feature matching module is configured to extract at least three feature points with the highest similarity from the target contour features and match the corresponding standard points in the template image;

[0041] A posture transformation module is configured to perform affine transformation on the feature points in combination with the affine transformation and the coordinates of the standard points to obtain the posture information of the transformed target contour features;

[0042] The motion control module is configured to control the robot to drive the probe from the feeding hole of the ladle cover into the interior of the ladle for lining detection according to the center position and posture information.

[0043] In order to solve the above technical problems, the present invention also provides a computer-readable storage medium. The computer-readable storage medium stores one or more programs, and one or more of the programs can be executed by one or more processors to implement the above detection method.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] By collecting the image of the feeding hole of the ladle cover and detecting the posture information and center position of the feeding hole of the ladle cover from the image, the posture information and center position of the feeding hole of the ladle cover can be accurately detected no matter how the position and / or angle of the ladle changes, so that the robot can extend the probe into the ladle for lining inspection according to the posture information and center position of the feeding hole of the ladle cover, without manual intervention in the inspection operation, thereby realizing automatic identification of the feeding hole of the ladle cover in the ladle lining inspection and automatic inspection of the ladle lining, reducing labor costs and improving inspection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0047] Figure 1 is a flow chart of the steps of the method for detecting the feeding hole of the ladle cover provided by an embodiment of the present invention;

[0048] Figure 2 It is a process diagram of detecting the feeding hole of a ladle cover by the method for detecting the feeding hole of a ladle cover provided by an embodiment of the present invention;

[0049] Figure 3 is a flowchart of further steps in step S6 of the oil cylinder installation method provided in an embodiment of the present invention;

[0050] Figure 4 is a flowchart of further steps in step S3 of the oil cylinder installation method provided in an embodiment of the present invention;

[0051] Figure 5 is a flowchart of further steps in step S1 of the oil cylinder installation method provided in an embodiment of the present invention;

[0052] Figure 6 It is a further step flow chart of step S7 in the oil cylinder installation method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0054] See also Figure 1-Figure 2 In order to solve the problems in the prior art, the present invention provides a method for detecting a charging hole of a ladle cover for detecting a ladle lining, the detection method comprising the steps of:

[0055] S1: Collect the real-time image of the feeding hole of the ladle cover and extract the contour features contained in the real-time image;

[0056] S2: performing pixel matching on the multiple contour features and the template image of the ladle cover feeding hole to obtain a similarity point set of the multiple contour features;

[0057] S3: Perform threshold judgment on the similarity of pixels in multiple similarity point sets, calculate the ratio of the number of pixels greater than the threshold to the number of pixels in the template image, and extract the similarity point set with the highest ratio as the target contour feature;

[0058] S4: extract the position corresponding to the highest similarity in the target contour feature as the center position of the target contour feature;

[0059] S5: extract at least three feature points with the highest similarity from the target contour features, and match the corresponding standard points in the template image;

[0060] S6: performing affine transformation on the feature points by combining the affine transformation and the coordinates of the standard points to obtain the position and posture information of the transformed target contour features;

[0061] S7: Control the robot to drive the probe from the feeding hole of the ladle cover into the inside of the ladle to perform lining inspection according to the center position and posture information.

[0062] In step S1, a real-time image is collected by an image acquisition device, and the real-time image includes the ladle cover feeding hole, so that the characteristics of the ladle cover feeding hole can be identified when the real-time image is subsequently processed, and then when the ladle lining scanning operation is performed, the robot can drive the probe into the ladle cover feeding hole according to the identified characteristics of the ladle cover feeding hole to scan the lining. The image acquisition device includes any one of a 3D camera, a depth camera, and a binocular camera to capture a real-time image including the ladle cover feeding hole.

[0063] In step S2, the feature of the charging hole of the ladle cover is identified by contour similarity comparison, that is, after all the contour features contained in the real-time image are extracted in step S1, each contour feature is pixel-matched with the template image of the charging hole of the ladle cover to obtain the similarity between each pixel point in each contour feature and the pixel point in the template image, and then a similarity point set is generated according to the similarity of the pixel points and the contour features in which they are located. For example, the pixel points obtained by pixel matching of contour feature A with the template image are A1, A2, A3..., and the pixel points obtained by pixel matching of contour feature B with the template image are B1, B2, B3..., therefore, a similarity point set of contour feature A and contour feature A can be generated, that is, the similarity point set of contour feature A is A=(A1, A2, A3...), and the similarity point set of contour feature B is B=(B1, B2, B3...), so as to facilitate the subsequent identification of the target contour feature, that is, the contour of the charging hole of the ladle cover, from the similarity point set under each contour feature.

[0064] In step S3, first, a threshold judgment is performed on the similarity of pixels in multiple similarity point sets to improve the accuracy of pixel matching. For example, the threshold is set to 0.5. When the pixel similarity is higher than 0.5, the detection result is considered reliable. When the pixel similarity is lower than 0.5, it is considered to be a false detection or failure to match the target. Then, the proportion of the number of pixels greater than the threshold to the number of pixels in the template image is calculated, and the similarity point set with the highest proportion is extracted as the target contour feature. For example, the number of pixels in the template image is 3000, and the number of pixels obtained by similarity point set A after threshold judgment is 1000, and the number of pixels obtained by similarity point set B after threshold judgment is 2500. Therefore, it can be calculated that the proportion of similarity point set A is 30%, and the proportion of similarity point set B is about 83.3%. At this time, similarity point set B can be extracted as the target contour feature.

[0065] Subsequently, in step S4, the position corresponding to the highest similarity is extracted from the target contour feature as the center position of the target contour feature, that is, the center position of the feeding hole of the ladle cover, so that the robot can extend the probe into the ladle from the center position and perform lining detection to avoid the probe from touching the edge of the ladle cover.

[0066] In step S5 and step S6, at least three feature points with the highest similarity are extracted from the target contour features, and the corresponding standard points are matched in the template image to calculate the association between the target contour features and the template image. When calculating the association, the feature points are affine transformed in combination with the affine transformation and the coordinates of the standard points to obtain the pose information of the transformed target contour features, that is, the pose information of the feeding hole of the ladle cover is obtained.

[0067] In step S7, the center position and posture information of the component are sent to the robot, so that the robot can drive the probe to move according to the center position and posture information, and extend the probe from the center position of the feeding hole of the ladle cover into the ladle for lining detection.

[0068] In the above-mentioned ladle lining detection process, by collecting the image of the ladle cover feeding hole and detecting the posture information and center position of the ladle cover feeding hole from the image, no matter how the position and / or angle of the ladle changes, the posture information and center position of the ladle cover feeding hole can be accurately detected, and then the robot extends the probe into the ladle for lining detection according to the posture information and center position of the ladle cover feeding hole, without manual intervention in the detection operation, thereby realizing the automatic identification of the ladle cover feeding hole in the ladle lining detection and the automatic detection of the ladle lining, reducing labor costs and improving detection efficiency.

[0069] See also Figure 3In an embodiment of the present invention, in the step of performing affine transformation on the feature points in combination with the affine transformation and the coordinates of the standard points to obtain the position and posture information of the transformed target contour features, the step further includes:

[0070] S601: Performing affine transformation on at least three feature points with the highest similarity and corresponding standard points using an affine transformation matrix, and calculating the affine transformation matrix;

[0071] S602: Performing an affine transformation on the target contour feature using an affine transformation matrix, and obtaining position information and angle information of the transformed target contour feature relative to the template image.

[0072] In one example, the three feature points with the highest similarity in the target contour feature are feature point A, feature point B, and feature point C, and their corresponding coordinates are A=(x A ,y A ), B=(x B ,y B ), C=(x C ,y C ), feature point A, feature point B, feature point C correspond to standard point A', standard point B', standard point C' in the template image, and their corresponding coordinates are A'=(x A ′,y A ′), B′=(x B ′,y B ′), C′=(x C ′,y C ′), at this time, affine transformation can be performed on A and A′, B and B′, C and C′ respectively, and the affine transformation matrix can be calculated. Then, the affine transformation matrix can be used to perform affine transformation on the target contour features, so as to obtain the position information and angle information of the target contour features relative to the template image.

[0073] By using the affine transformation formula to perform affine transformation on at least three feature points in the target contour feature, and then using the affine transformation matrix to perform affine transformation on the target contour feature, the position and angle information of the target contour feature can be accurately adjusted, thereby achieving accurate recognition of the ladle cover feeding hole in different positions and postures, further improving the accuracy and reliability of the detection method, and ensuring that the robot can accurately insert the probe into the ladle cover feeding hole for lining detection.

[0074] In an embodiment of the present invention, at least three feature points with the highest similarity extracted include a center point of a target contour feature and at least two edge points of the target contour features, so that the extracted feature points better represent the shape and position characteristics of the feeding hole of the ladle cover, wherein the center point can reflect the approximate position center of the feeding hole of the ladle cover, and the edge point helps to determine the contour shape and direction of the feeding hole of the ladle cover, thereby improving the accuracy of subsequent affine transformation and posture calculation, and making the detection result more reliable.

[0075] When performing an affine transformation, the affine transformation formula used is:

[0076]

[0077] Wherein, M is a 2x3 affine transformation matrix, and the affine transformation of the target contour feature includes operations of rotation, translation and scaling of the target contour feature.

[0078] The rotation, scaling and translation of the target contour features are controlled by the matrix elements in the affine transformation formula. Therefore, after the target contour features change at any position and angle, the position information of the feeding hole of the ladle cover can be accurately detected by adjusting the matrix parameters, thereby improving the adaptability of the detection of the feeding hole of the ladle cover.

[0079] In an embodiment of the present invention, similarity point sets of multiple contour features are calculated respectively by using a calculation formula of a normalized cross-correlation coefficient (NCC), and the calculation formula of the normalized cross-correlation coefficient is:

[0080]

[0081] Where T(i, j) is the pixel value of the template image, I(x+i, y+j) is the local pixel value of the contour feature at position (x, y) in the real-time image, and are the average values ​​of template image and contour features respectively.

[0082] The normalized cross-correlation coefficient calculation formula is used to calculate the pixel similarity between multiple contour features and template images to reduce interference items in multiple contour features and improve the reliability of target contour feature detection, that is, to improve the detection reliability of the feeding hole of the ladle cover.

[0083] See also Figure 4 In an embodiment of the present invention, in the step of performing a threshold judgment on the similarity of the pixels in the plurality of similarity point sets, calculating the ratio of the number of pixels greater than the threshold to the number of pixels in the template image, and extracting the similarity point set with the highest ratio as the target contour feature, the step further includes:

[0084] S301: Compare the similarity of the pixels in each similarity point set with a threshold value, and select the number of pixel similarities greater than the threshold value in each similarity point set;

[0085] S302: Calculate the proportion of the number of pixel similarities in the corresponding similarity point set;

[0086] S303: Extract the similarity point set with the highest proportion as the target contour feature.

[0087] By comparing the similarity with the threshold and calculating the ratio to extract the target contour features, the most reliable similarity point set can be selected from multiple similarity point sets. In the screening of the similarity point set, not only the similarity value of a single pixel in the similarity point set is considered, but also the ratio of the number of pixels in the entire similarity point set that is greater than the threshold is considered, thereby avoiding misselection due to individual abnormally high similarity values, improving the reliability of target contour feature extraction, and thus improving the detection accuracy of target contour features.

[0088] See also Figure 5 In an embodiment of the present invention, in the step of extracting the contour features contained in the real-time image, the step further includes:

[0089] S101: Performing Gaussian smoothing on the real-time image using Gaussian filtering;

[0090] S102: Convolve the real-time image using a two-dimensional Gaussian to remove noise in the real-time image;

[0091] S103: using the Sobel operator to calculate the horizontal gradient and the vertical gradient of the real-time image, and using the non-maximum suppression method to remove non-edge points in the real-time image;

[0092] S104: using a double threshold processing method to detect strong edges and weak edges of the real-time image, and combining the strong edges and weak edges to detect contour features in the real-time image.

[0093] The real-time image is preprocessed by the Canny algorithm. First, the real-time image is smoothed by Gaussian filtering. By selecting a suitable Gaussian kernel function (such as a 3×3 or 5×5 Gaussian kernel, etc., determined according to the image noise situation), each pixel in the image is weighted averaged to remove the noise in the image and make the image smoother. For example, for a pixel in the image, the weighted average of the surrounding pixels is calculated according to the Gaussian kernel function as the new value of the pixel, thereby reducing the impact of noise on image quality.

[0094] Then, a two-dimensional Gaussian kernel is used to perform a convolution operation on the processed image to further remove residual noise. The two-dimensional Gaussian kernel slides on the image, performs convolution calculation on each pixel, and performs weighted summation on the pixels according to the value of the kernel function to adjust the pixel value to make the image clearer.

[0095] For each pixel in the image, the Sobel operator calculates the grayscale difference between its adjacent pixels in the horizontal and vertical directions to obtain horizontal gradient and vertical gradient images. Then the non-maximum suppression method is used to remove non-edge points in the image. In the gradient image, the gradient value of each pixel is compared with that of its adjacent pixels. If the gradient value of the pixel is not the local maximum, it is judged as a non-edge point and removed, and the edge points are retained, thereby extracting the edge information of the image.

[0096] For the edge contour of the image, the double threshold processing method is used to set two thresholds T low and T high Points with gradient intensity greater than the high threshold are considered "strong edges", points less than the low threshold are ignored, and points in between are considered "weak edges", thereby eliminating false edges caused by noise, retaining important edges, and finally connecting edges, taking strong edges as the basis of edges, starting from these strong edges, if a weak edge is connected to a strong edge, it is considered an edge, otherwise it is suppressed to ensure the coherence of the edge, thereby realizing denoising of the real-time image and improving the accuracy of subsequent recognition of the feeding hole of the ladle cover.

[0097] In one example, after Gaussian smoothing using Gaussian filtering, a 3x3 Gaussian kernel with a standard deviation of σ=1.2 is selected, noise is removed by convolution of the Gaussian kernel with the image, and then the image is further convolved with a two-dimensional Gaussian kernel to remove noise, and then the Sobe operator is used to calculate the horizontal and vertical gradients. After obtaining the gradient image, the non-maximum suppression method is used to remove non-edge points, and a double threshold (such as a high threshold T high =0.8, low threshold T low =0.3) detects strong edges and weak edges, and then combines the two to detect contour features in the real-time image.

[0098] See also Figure 6 In an embodiment of the present invention, in the step of controlling the robot to drive the probe from the ladle cover feeding hole into the ladle for lining detection according to the center position and posture information, the step further includes the following steps:

[0099] S701: Convert the center position into the target coordinates under the robot base coordinates;

[0100] S702: Control the robot to generate a motion trajectory according to the target coordinates and posture information;

[0101] S703: Control the robot to drive the probe from the feeding hole of the ladle cover into the ladle according to the motion trajectory and drive the probe to rotate and scan the ladle lining.

[0102] After detecting the posture information of the target contour, the center position is converted into the target coordinates under the robot base coordinates through the coordinate transformation matrix, and the robot is controlled to generate a motion trajectory according to the target coordinates and posture information, so that the robot moves according to the motion trajectory and drives the probe from the center position into the ladle, that is, drives the probe from the feeding hole of the ladle cover into the ladle, so that the probe scans the inside of the ladle, and at the same time, by controlling the probe to rotate, the probe can perform an all-round scan of the inside of the ladle, so as to comprehensively detect the ladle lining and obtain more complete lining information, thereby improving the comprehensiveness and effectiveness of the ladle lining detection, and helping operators to promptly discover problems with the ladle lining and repair them.

[0103] In order to solve the above technical problems, the present invention also provides a ladle cover feeding hole detection system for ladle lining detection, the detection system comprising:

[0104] An image acquisition module is configured to acquire a real-time image of the charging hole of the ladle cover and extract contour features contained in the real-time image;

[0105] A pixel matching module; configured to perform pixel matching on a plurality of contour features and a template image of a ladle cover feeding hole to obtain a similarity point set of the plurality of contour features;

[0106] A threshold judgment module is configured to perform a threshold judgment on the similarity of the pixels in the plurality of similarity point sets, calculate the ratio of the number of pixels greater than the threshold to the number of pixels in the template image, and extract the similarity point set with the highest ratio as the target contour feature;

[0107] A center extraction module is configured to extract a position corresponding to the highest similarity in the target contour feature as the center position of the target contour feature;

[0108] A feature matching module is configured to extract at least three feature points with the highest similarity from the target contour features and match the corresponding standard points in the template image;

[0109] A posture transformation module is configured to perform affine transformation on the feature points in combination with the affine transformation and the coordinates of the standard points to obtain the posture information of the transformed target contour features;

[0110] The motion control module is configured to control the robot to drive the probe from the feeding hole of the ladle cover into the interior of the ladle for lining detection according to the center position and posture information.

[0111] The real-time image of the feeding hole of the ladle cover is collected by the image acquisition module, and multiple similarity point sets are extracted from the real-time image by the pixel matching module, so that the threshold judgment module can extract the target contour feature from the multiple similarity point sets, and the center extraction module can extract the center position of the target contour feature from the target contour feature. Subsequently, the feature matching module can extract the feature points and standard points corresponding to the template image from the target contour feature, so that the posture transformation module can calculate the posture information of the target contour feature, and finally the motion control module can control the robot to drive the probe from the feeding hole of the ladle cover into the inside of the ladle for lining detection according to the posture information. The entire ladle lining detection process does not require human intervention, and realizes the automatic recognition of the feeding hole of the ladle cover and the automatic detection of the ladle lining in the ladle lining detection, thereby reducing labor costs and improving detection efficiency.

[0112] In order to solve the above technical problems, the present invention also provides a computer-readable storage medium. The computer-readable storage medium stores one or more programs, and one or more of the programs can be executed by one or more processors to implement the above detection method, so that when performing the ladle lining detection operation, by collecting the image of the ladle cover feeding hole and detecting the posture information and center position of the ladle cover feeding hole from the image, the posture information and center position of the ladle cover feeding hole can be accurately detected after the position and / or angle of the ladle changes, so that the robot can extend the probe into the ladle for lining detection according to the posture information and center position of the ladle cover feeding hole, without manual intervention in the detection operation, thereby realizing the automatic identification of the ladle cover feeding hole in the ladle lining detection and the automatic detection of the ladle lining, reducing labor costs and improving detection efficiency.

[0113] In the present invention, unless otherwise clearly specified and limited, the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. The first feature being "up" or "down" the second feature may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, the first feature being "above", "above" and "above" the second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. The first feature being "below", "below" and "below" the second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature, and therefore cannot be understood as limiting the present invention.

[0114] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited thereto. Within the technical concept of the present invention, the technical solution of the present invention can be subjected to a variety of simple modifications, including the combination of various specific technical features in any appropriate manner. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations. However, these simple modifications and combinations should also be regarded as the contents disclosed by the present invention and belong to the protection scope of the present invention.

[0115] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for detecting a feeding hole of a ladle cover for detecting a ladle lining, characterized in that: The detection method comprises the steps of: Collect the real-time image of the feeding hole of the ladle cover and extract the contour features contained in the real-time image; Perform pixel matching on the multiple contour features and the template image of the ladle cover feeding hole to obtain a similarity point set of the multiple contour features; Perform threshold judgment on the similarity of pixels in multiple similarity point sets, calculate the ratio of the number of pixels greater than the threshold to the number of pixels in the template image, and extract the similarity point set with the highest ratio as the target contour feature; The position corresponding to the highest similarity is extracted from the target contour feature as the center position of the target contour feature; Extract at least three feature points with the highest similarity from the target contour features, and match the corresponding standard points in the template image; Combining the affine transformation with the coordinates of the standard points, the feature points are affine transformed to obtain the position and posture information of the transformed target contour features; The robot is controlled to drive the probe from the feeding hole of the ladle cover into the interior of the ladle for lining inspection according to the center position and posture information.

2. The method for detecting the feeding hole of the ladle cover for detecting the ladle lining according to claim 1, characterized in that: In the step of performing affine transformation on the feature points by combining the affine transformation and the coordinates of the standard points to obtain the position and posture information of the transformed target contour features, the following is further included: Performing affine transformation on at least three feature points with the highest similarity and corresponding standard points using an affine transformation matrix, and calculating the affine transformation matrix; The target contour features are affine transformed using an affine transformation matrix, and the position information and angle information of the transformed target contour features relative to the template image are obtained.

3. The method for detecting the feeding hole of the ladle cover for detecting the ladle lining according to claim 2, characterized in that: The at least three extracted feature points with the highest similarity include a center point of a target contour feature and at least two edge points of the target contour feature.

4. The method for detecting the feeding hole of the ladle cover for detecting the ladle lining according to claim 2, characterized in that: The affine transformation formula is: Wherein, M is a 2x3 affine transformation matrix, and the affine transformation of the target contour feature includes operations of rotation, translation and scaling of the target contour feature.

5. The method for detecting the feeding hole of the ladle cover for detecting the ladle lining according to claim 1, characterized in that: The similarity point sets of multiple contour features are calculated respectively by the calculation formula of the normalized mutual correlation coefficient. The calculation formula of the normalized mutual correlation coefficient is: Where T(i, j) is the pixel value of the template image, I(x+i, y+j) is the local pixel value of the contour feature at position (x, y) in the real-time image, and are the average values ​​of template image and contour features respectively.

6. The method for detecting the feeding hole of the ladle cover for detecting the ladle lining according to claim 1, characterized in that: In the step of performing a threshold judgment on the similarity of the pixels in the plurality of similarity point sets, calculating the ratio of the number of pixels greater than the threshold to the number of pixels in the template image, and extracting the similarity point set with the highest ratio as the target contour feature, further comprising the steps of: The similarity of the pixels in each similarity point set is compared with the threshold value, and the number of pixel similarities greater than the threshold value is screened out in each similarity point set; Calculate the proportion of the number of pixel similarities in the corresponding similarity point set; The similarity point set with the highest proportion is extracted as the target contour feature.

7. The method for detecting the feeding hole of the ladle cover for detecting the ladle lining according to claim 1, characterized in that: In the step of extracting the contour features contained in the real-time image, the steps are further included: Use Gaussian filtering to perform Gaussian smoothing on real-time images; Convolution of the real-time image with a two-dimensional Gaussian is performed to remove noise from the real-time image; The Sobel operator is used to calculate the horizontal and vertical gradients of the real-time image, and the non-maximum suppression method is used to remove non-edge points in the real-time image; The double threshold processing method is used to detect the strong edge and weak edge of the real-time image, and the contour features in the real-time image are detected by combining the strong edge and the weak edge.

8. The method for detecting the feeding hole of the ladle cover for detecting the ladle lining according to claim 1, characterized in that: In the step of controlling the robot to drive the probe from the ladle cover feeding hole into the ladle to perform lining detection according to the center position and posture information, the step further includes the following steps: Convert the center position to the target coordinates in robot base coordinates; Control the robot to generate motion trajectory according to target coordinates and posture information; The robot is controlled to drive the probe to extend from the feeding hole of the ladle cover into the interior of the ladle according to the motion trajectory, and drive the probe to perform rotational scanning on the inner lining of the ladle.

9. A ladle cover feeding hole detection system for ladle lining detection, characterized in that: The detection system comprises: An image acquisition module is configured to acquire a real-time image of the charging hole of the ladle cover and extract contour features contained in the real-time image; A pixel matching module; configured to perform pixel matching on a plurality of contour features and a template image of a ladle cover feeding hole to obtain a similarity point set of the plurality of contour features; A threshold judgment module is configured to perform a threshold judgment on the similarity of the pixels in the plurality of similarity point sets, calculate the ratio of the number of pixels greater than the threshold to the number of pixels in the template image, and extract the similarity point set with the highest ratio as the target contour feature; A center extraction module is configured to extract a position corresponding to the highest similarity in the target contour feature as the center position of the target contour feature; A feature matching module is configured to extract at least three feature points with the highest similarity from the target contour features and match the corresponding standard points in the template image; A posture transformation module is configured to perform affine transformation on the feature points in combination with the affine transformation and the coordinates of the standard points to obtain the posture information of the transformed target contour features; The motion control module is configured to control the robot to drive the probe from the feeding hole of the ladle cover into the interior of the ladle for lining detection according to the center position and posture information.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the detection method according to any one of claims 1 to 8.

Images