Online visual inspection method and device for square billets

Abstract

The application discloses an online visual detection method for rhombus deformation of a continuous casting square billet, and relates to the field of continuous casting square billet rhombus deformation detection. The method comprises the following steps: a flame cutting machine cuts the continuous casting billet into square billets with a proper length, and then the square billets are transported to a detection camera through a roller way; the camera shoots the end face of the square billet to obtain a digital image of the end face of the square billet; the square billet end face position in the image is located and segmented through image binarization operation; the positions of each edge line of the square billet end face are calculated, and the intersection points of two adjacent edge lines are obtained, then the diagonal line length of the square billet and the rhombus deformation rate can be calculated; if the calculation result has a low confidence level due to the influence of the image effect, then the above process is performed on digital images obtained under different camera parameter conditions, and the calculation result with the highest confidence level is saved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of billet shape quality detection, and more particularly relates to an online continuous casting billet diamond visual detection method and device. BACKGROUND

[0002] Diamond, commonly known as off-square, refers to a kind of out-of-tolerance defect that square billets or rectangular billets evolve into parallelogram, trapezoid or diamond due to the four corners not being 90°. Diamond is a common defect of small square billets, and its characteristic is that the difference between the opposite sides of the square billet cross section exceeds the standard. The hazards of diamond mainly include the following points: 1. Diamond induces internal cracks of the casting billet, especially diagonal cracks; 2. Diamond accompanied by a certain degree of bulging can cause a leakage accident, affecting the normal production of the casting billet; 3. Diamond can cause difficulty in hoisting the casting billet, and the scattered clamps are easy to damage the equipment and cause safety accidents during hoisting; 4. Diamond can cause the square billet to be difficult to bite into the roll pass, which brings certain difficulties to rolling, and is prone to steel stacking and material pouring accidents, which has a great impact on production and safety.

[0003] At present, the square billet diamond situation is mainly supervised by visual or manual measurement sampling methods. In actual production process, for the hot delivery and loading of square billets, manual measurement is difficult and has a high safety risk, and the visual method is mainly used to evaluate whether the diamond exceeds the standard requirement; for the square billets after the cold bed loading, the manual measurement sampling method is mainly used to monitor whether the diamond exceeds the standard requirement. The above two methods cannot realize accurate monitoring of the diamond of each square billet, and there is a risk of sending the diamond exceeding the standard square billet to the rolling process. SUMMARY

[0004] In view of the above defects and improvement needs of the prior art, the present application provides an online continuous casting billet diamond visual detection method, which effectively improves the diamond detection efficiency, improves the diamond detection accuracy, reduces the measurement difficulty and reduces the safety hazard.

[0005] To achieve the above purpose, according to one aspect of the present application, an online continuous casting billet diamond visual detection method is provided, comprising the following steps:

[0006] (1) A high-resolution industrial camera with a resolution greater than a preset resolution is selected, and the configuration parameters of the industrial camera are adjusted so that the industrial camera can shoot a digital image with clear boundaries of each side of the square billet end face;

[0007] (2) Set the visual algorithm region of interest, when the square billet end face enters the region of interest, perform binary segmentation on the square billet end face digital image, and locate the pixel coordinate point set of each side of the square billet end face;

[0008] (3) According to the pixel coordinate point set provided in step (2), a fitting calculation is performed, boundary transition noise is removed, and a square blank end face boundary function is obtained in a linear fitting manner;

[0009] (4) According to the square blank end face boundary function obtained in step (3), a combination calculation is performed on the square blank end face boundary functions of two adjacent edges, and the focal point coordinates of the four corners of the square blank end face in the image are fitted out;

[0010] (5) According to the focal point coordinates of the four corners of the square blank end face in the image obtained in step (4), the pixel lengths D1 and D2 of the two diagonal lines are calculated, since the distance from the camera to the square blank end face is certain and the field of view angle of the camera is certain, the actual lengths d1 and d2 of the two diagonal lines are calculated, and the diagonal line difference d and the rhombus rate R are calculated according to the formula as follows:

[0011] d = |d1-d2|

[0012]

[0013] (6) Since the image is greatly affected by the environmental light condition, the image effect is corrected by adjusting the camera parameters, one fixed stage value is adjusted each time, and steps (2)-(5) are repeated to calculate the diagonal line difference and the rhombus rate R;

[0014] (7) According to different environmental conditions, the correction times are set, and after the method in step (6), the detection result with the highest confidence is taken as the final result.

[0015] In an embodiment of the present application, in the step (1), in order to eliminate the interference factors of environmental light, a gray image with more detailed square blank and background boundaries is input to the subsequent steps, and an industrial camera with black and white imaging is selected.

[0016] In an embodiment of the present application, in the step (1), in order to make the visual effect of the square blank end face in the image reach one-dimensional form, the camera optical axis is perpendicular to the square blank end face, and the camera position is adjusted so that the square blank end face feature image appears in the center area of the digital image.

[0017] In an embodiment of the present application, in the step (1), in order to obtain a clear image of each edge line of the square blank end face, the camera focal length is adjusted to a clear state perceived by the naked eye.

[0018] In an embodiment of the present application, in the step (2), a visual algorithm region of interest is set, specifically including: the region of interest is a region in which the square blank end face appears in the image and is to be detected and stationary, which is represented as a rectangle, and the pixel coordinate range is [x min ,x max ], [y min ,y max], x min ], x max ], y min ], y max The four coordinates are taken from the pixel point coordinates in the digital image in step (1).

[0019] In an embodiment of the present application, in the step (2), the digital image of the blank end face is binarized and segmented, specifically including: image binarization refers to according to a given threshold, assigning the pixel point gray value greater than or equal to the threshold in the whole image as 1, and assigning the pixel point gray value less than the threshold as 0, thereby obtaining a black and white image with only two gray values of 0 and 1, the binarized image can effectively distinguish the background from the target, and quickly and accurately extract the target region boundary pixel point.

[0020] In an embodiment of the present application, the step (2) specifically includes:

[0021] For the gray scale image in step (1), the pixel gray scale value ranges from x i ∈[0, 255], the image is segmented into two parts of foreground and background, and it is necessary to ensure that the difference between the two parts of pixels is maximum, that is, the variance of the foreground and the background is maximum;

[0022] First, for the mean and variance of the whole image, they are respectively as follows:

[0023]

[0024]

[0025] Wherein, N represents the total number of pixels in the region of interest, x i represents the pixel value of the i-th position after the image is vectorized; similarly, if the image is segmented into two parts of foreground C1 and background C2 by a threshold t, the mean values of the two parts are respectively as follows:

[0026]

[0027]

[0028] Wherein, N1 and N2 respectively represent the number of pixels with a pixel value less than the threshold t and greater than the threshold t;

[0029] Let respectively represent the probability that any pixel in the remaining range of interest belongs to C1 or C2, and P1+P2=1, is represented as:

[0030]

[0031] For the variance σ 2According to a calculation formula of variance of discrete random variables, the following is expressed:

[0032]

[0033] The target is to find the optimal segmentation threshold t * So that the variance between the pixel mean of the foreground part C1 And the pixel mean of the background part C2 Is maximum, that is, the difference between the foreground and the background is maximum:

[0034] σ(t * ) 2 = max 0≤k≤255 σ(t) 2 .

[0035] In an embodiment of the present application, in the step (3), the closer the coordinate values of the pixel coordinate points in the set to the fitting function, the higher the function confidence, and vice versa.

[0036] In an embodiment of the present application, the step (7) specifically comprises:

[0037] After multiple detections, the detection result with the highest confidence is saved, and the detection result data is compared with the warning threshold value; if the diagonal difference and the diamond rate R exceed the threshold value, an alarm prompt is sent through a telecommunication signal and an audible and visual signal to prompt the operator to remove the abnormal square billet.

[0038] According to another aspect of the present application, there is also provided an online continuous casting square billet diamond visual detection device, comprising at least one processor and a memory, the at least one processor and the memory are connected through a data bus, the memory stores instructions executable by the at least one processor, and the instructions are used to complete the online continuous casting square billet diamond visual detection method after being executed by the processor.

[0039] Overall, compared with the prior art, the above technical solutions conceived by the present application have the following beneficial effects:

[0040] (1) The online visual inspection method for rhombic deformation of continuous casting billets provided by this invention involves setting an industrial camera at the end of the conveyor roller. After the fire-cutting machine cuts the billet into billets of appropriate length according to specifications, the billets are transported to the inspection camera via the roller. The camera takes pictures of the end face of the billet to obtain a digital image of the end face. The position of the end face of the billet in the image is located and segmented through image binarization. The position of each edge line of the end face of the billet is calculated by fitting. After obtaining the intersection of each pair of adjacent edge lines, the diagonal length and rhombic deformation rate of the billet can be calculated. If the confidence of the calculation result is low due to the influence of image effect, the above process is performed by obtaining digital images under multiple camera parameter conditions, and the calculation result with the highest confidence is saved. After the inspection is completed, the transport device transports the billet to another location. Compared with the current mainstream manual visual inspection and manual measurement, this method effectively improves the efficiency and accuracy of rhombic deformation detection and reduces safety hazards.

[0041] (2) The present invention enables online detection of square billet rhomboid defects. If an abnormality of square billet rhomboid is detected in the production line, an alarm system can be used to issue an alarm prompt, reminding on-site personnel to re-inspect the abnormal square billet and remove the square billet with rhomboid defects, thereby ensuring the quality of steel output, avoiding the harm of unqualified square billets to subsequent steel rolling processes, and greatly improving the accuracy and efficiency of detection. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the detection method of an online continuous casting square billet rhomboid deformation visual inspection method in an embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram illustrating the principle of an online continuous casting square billet rhomboid deformation visual inspection method in an embodiment of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0045] like Figure 1 The diagram shown is a schematic representation of the detection method for an online continuous casting square billet rhomboid deformation visual inspection method in an embodiment of the present invention.

[0046] The online continuous casting square billet rhombus visual inspection will face the problem of high temperature of the detection environment, so the following equipment needs to be installed and deployed to cooperate with each other to make the key electronic components meet the requirements of accurate observation and dust reduction and cooling, wherein the equipment structure is mainly divided into three parts, namely, a customized support, a single light air cooling straight blowing shield and a stainless steel air pipe.

[0047] The customized support is installed on the ground, mainly supports the single light air cooling straight blowing shield to a suitable height position, so that the industrial camera in the single light air cooling straight blowing shield can face the square billet end face, and secondly can also fix the stainless steel air pipe and the matching cable.

[0048] The single light air cooling straight blowing shield is installed at the end of the customized support arm, the high-temperature-resistant industrial camera is installed in the shield, and the single light air cooling straight blowing shield can prevent dust, collision, heat insulation and cooling for the industrial camera.

[0049] The stainless steel air pipe is inserted into the air hole at the tail of the single light air cooling straight blowing shield at one end, and is inserted into the air supply hole at the other end, the air source directly blows the industrial camera body, and mainly plays a blowing and cooling role.

[0050] As shown in Figure 2 , the present application provides an online continuous casting square billet rhombus visual inspection method, comprising:

[0051] (1) selecting a high-resolution industrial camera with a resolution greater than a preset resolution, adjusting the configuration parameters of the industrial camera, so that the industrial camera can shoot a digital image with clear boundaries of each side of the square billet end face;

[0052] Generally, in the step (1), in order to eliminate the interference factors of ambient light, a gray image with clear boundaries between the square billet and the background is input to the subsequent steps, and a black and white imaging industrial camera is selected;

[0053] Generally, in the step (1), in order to make the visual effect of the square billet end face in the image reach one-dimensional form, the camera optical axis is perpendicular to the square billet end face, and the camera position is adjusted so that the square billet end face feature image appears in the center area of the digital image;

[0054] Generally, in the step (1), in order to obtain a clear image of each side line of the square billet end face, the camera focal length focal point is adjusted to a clear state perceived by the naked eye;

[0055] (2) setting a visual algorithm region of interest, when the square billet end face enters the region of interest, performing binary segmentation on the square billet end face digital image, and positioning to the pixel coordinate point set of each side of the square billet end face;

[0056] Specifically, in step (2), setting the region of interest for the visual algorithm includes: defining the area in the image where the end face of the billet appears and is to be inspected as the region of interest, represented as a rectangle, with pixel coordinates ranging from [x...]. min ,x max ],[y min ,y max ], x min x max y min y max All four coordinates are taken from the pixel coordinates within the digital image described in step (1);

[0057] Specifically, in step (2), the digital image of the billet end face is binarized and segmented, which specifically includes: image binarization means that according to a given threshold, the gray values ​​of pixels in the whole image that are greater than or equal to the threshold are assigned to 1, and the gray values ​​of pixels that are less than the threshold are assigned to 0, so as to obtain a black and white image with only two gray values, 0 and 1. The binarized image can effectively distinguish the background and the target, and quickly and accurately extract the boundary pixels of the target area.

[0058] One specific implementation of step (2) is as follows:

[0059] For the grayscale image described in step (1), the range of its pixel grayscale values ​​is x. i To divide the image into foreground and background parts, the difference between the pixels of the two parts must be maximized, i.e., the variance between the foreground and background must be maximized.

[0060] First, the mean and variance of the entire image are calculated as follows:

[0061]

[0062]

[0063] Where N represents the total number of pixels in the region of interest, x i This represents the pixel value at the i-th position after the image has been vectorized; similarly, if the image is divided into foreground C1 and background C2 by a threshold t, the mean values ​​of the two parts are as follows:

[0064]

[0065]

[0066] Where N1 and N2 represent the number of pixels with pixel values ​​less than threshold t and greater than threshold t, respectively;

[0067] make Pi and P2 represent the probability that any pixel in the remaining range of interest belongs to C1 or C2, respectively, and there exists P1+P2=1, is expressed as:

[0068]

[0069] For two random variables, the variance σ 2 According to the calculation formula of the variance of a discrete random variable, it is expressed as:

[0070]

[0071] So far, the goal is to find the optimal segmentation threshold t * So that the mean of the foreground C1 part of the pixel And the mean of the background part C2 of the pixel) The variance between them is maximized, that is, the difference between the foreground and the background is maximized:

[0072] σ(t * ) 2 = max 0≤k≤255 σ(t) 2 .

[0073] (3) According to the pixel coordinate point set provided in step (2), the fitting calculation is performed, the boundary transition noise is removed, and the square blank end face boundary function is obtained in a linear fitting manner;

[0074] Generally, in the step (3), the closer the coordinate values of each point in the pixel coordinate point set to the fitting function, the higher the function confidence, and vice versa.

[0075] (4) According to the square blank end face boundary function obtained in step (3), the square blank end face boundary function of each adjacent edge is combined and calculated, and the focal point coordinates of the four corners of the square blank end face in the image are fitted out;

[0076] (5) According to the focal point coordinates of the four corners of the square blank end face in the image obtained in step (4), the pixel lengths D1 and D2 of the two diagonal lines are calculated. Since the distance from the camera to the square blank end face is certain and the field of view angle of the camera is certain, the actual lengths d1 and d2 of the two diagonal lines are calculated, and the diagonal line difference d and the rhombus rate R are calculated according to the formula, which is as follows:

[0077] d = |d1-d2|

[0078]

[0079] (6) Since the image is greatly affected by the environmental light conditions, the image effect is corrected by adjusting the camera parameters, one fixed stage value is adjusted each time, and steps (2)-(5) are repeated to calculate the diagonal line difference and the rhombus rate R;

[0080] (7) According to different environmental conditions, the number of corrections is set, and after the method of step (6) is performed, the detection result with the highest confidence is taken as the final result.

[0081] The step (7) specifically comprises: saving the detection result with the highest confidence after multiple detections, and comparing the detection result data with the warning threshold; if the diagonal difference and the rhombus variation rate R exceed the threshold, an alarm prompt is sent through a telecommunication signal and an audible and visual signal to prompt an operator to remove the abnormal square billet. The whole process does not affect normal production.

[0082] Further, the present application provides an online continuous casting square billet rhombus visual detection device, comprising at least one processor and a memory, the at least one processor and the memory are connected through a data bus, the memory stores instructions executable by the at least one processor, and the instructions are used to complete the online continuous casting square billet rhombus visual detection method after being executed by the processor.

[0083] Further, the present application provides an online continuous casting square billet rhombus visual detection system, comprising the following modules:

[0084] The front-end image acquisition module is composed of a high-temperature-resistant industrial camera and a matching customized device, and mainly plays a role of collecting data images and protecting and adjusting equipment;

[0085] The back-end visual algorithm module is composed of an image acquisition card and a calculation module, and mainly plays a role of visual algorithm operation and output of result data;

[0086] The electrical module mainly comprises network equipment, a power supply and a cable, and mainly plays a role of power supply and communication.

[0087] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present application, and is not used to limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An on-line visual inspection method for distortion of a continuously cast square billet, characterized by, It comprises the following steps: (1) Select a high-resolution industrial camera with a resolution higher than a preset resolution, and adjust the configuration parameters of the industrial camera, so that the industrial camera can shoot a digital image with clear boundaries of each side of the billet end face; (2) Set the visual algorithm interest area, when the billet end face enters the interest area, perform binary segmentation on the billet end face digital image, and locate the pixel coordinate point set of each side of the billet end face; (3) According to the pixel coordinate point set provided in step (2), perform fitting calculation, remove boundary transition noise, and obtain the billet end face boundary function in a linear fitting manner; (4) According to the billet end face boundary function obtained in step (3), perform combined calculation on the billet end face boundary functions of each two adjacent sides, and fit the focal point coordinates of the four corners of the billet end face in the image; (5) According to the focal point coordinates of the four corners of the billet end face in the image obtained in step (4), calculate the pixel lengths D1 and D2 of the two diagonal lines, calculate the actual lengths d1 and d2 of the two diagonal lines, and calculate the diagonal line difference d and the rhombus rate R according to the formula as follows: d = |d1-d2| (6) Adjust the camera parameters to correct the image effect, adjust one fixed stage value each time, and repeat steps (2)-(5) to calculate the diagonal line difference and the rhombus rate R; (7) Set the number of corrections according to different environmental conditions, take the detection result with the highest confidence as the final result after step (6).

2. The on-line visual detection method of square billet distortions according to claim 1, characterized in that, Specifically, in the step (1), a black and white imaging industrial camera is selected.

3. The on-line visual inspection method for square billets according to claim 1 or 2, wherein Specifically, in the step (1), the camera optical axis is perpendicular to the billet end face, and the camera position is adjusted so that the billet end face feature image appears in the center area of the digital image, so that the visual effect of the billet end face in the image reaches one-dimensional form.

4. The on-line visual inspection method for square billets according to claim 1 or 2, wherein In the step (1), in order to obtain a clear image of each side line of the billet end face, the camera focal length is adjusted to a clear state perceived by the naked eye.

5. The on-line visual inspection method for square billets according to claim 1 or 2, wherein In the step (2), the visual algorithm interest area is set, which specifically includes: The region in which the square blank end face appears and is to be inspected in the image is the region of interest, which is a rectangle with pixel coordinate ranges of [x min ,x max ] and [y min ,y max ]. The four coordinates are all taken from the pixel point coordinates in the digital image in step (1). min max min max ​​​​ 6. The on-line visual inspection method for square billets according to claim 1 or 2, wherein In the step (2), the billet end face digital image is binarized and segmented, which specifically includes: Image binarization refers to assigning the pixel point gray value greater than or equal to the threshold value in the entire image to 1, and assigning the pixel point gray value less than the threshold value to 0, so as to obtain a black and white image with only two gray values of 0 and 1.

7. The on-line visual inspection method for square billets according to claim 1 or 2, wherein The step (2) specifically includes: For the digital image described in step (1), the pixel gray value range is x i ∈[0, 255], the image is divided into two parts of foreground and background, and it is necessary to ensure that the difference between the two parts of pixels is maximum, that is, the variance of the foreground and background is maximum; First, the mean and variance of the entire image are calculated as follows: where N represents the total number of pixels in the region of interest, x i represents the pixel value at the i-th position after the image is vectorized; similarly, if the image is divided into two parts, foreground C1 and background C2, by a threshold t, the mean values of the two parts are as follows: Wherein, N1 and N2 respectively represent the number of pixels with pixel value less than threshold t and greater than threshold t; Let P1and P2denote the probabilities that any pixel in the remaining range of interest belongs to C1or C2, respectively, and that P1+ P2= 1, is denoted as: For two random variables, the variance σ 2 is given by the formula for the variance of a discrete random variable as follows: The goal is to find the optimal threshold t * such that the variance between the mean of the pixels of the foreground part C1 and the mean of the pixels of the background part C2 is maximum, i.e. the difference between foreground and background is maximum: σ(t * ) 2 = max 0≤k≤255 σ(t) 2 .

8. The on-line visual inspection method for square billets according to claim 1 or 2, wherein In the step (3), the closer the coordinate values of each point in the pixel coordinate point set to the fitting function, the higher the function confidence, and vice versa.

9. The on-line visual inspection method for square billets according to claim 1 or 2, wherein The step (7) specifically includes: After multiple detections, save the detection result with the highest confidence, and compare the detection result data with the warning threshold value. If the diagonal line difference and the rhombus rate R exceed the threshold value, an alarm prompt is sent through a telecommunication signal and an audible and visual signal to prompt the operating personnel to remove the abnormal billet.

10. An online continuous casting billet rhombus visual detection device, characterized in that: The method comprises at least one processor and a memory, the at least one processor and the memory are connected through a data bus, the memory stores instructions executable by the at least one processor, and the instructions are used to complete the online continuous casting square billet rhombus visual inspection method in any one of claims 1-9 after being executed by the processor.

Images