Steel coil loosening recognition method and device based on semantic segmentation and ellipse fitting

Abstract

The application provides a steel coil loosening recognition method and device based on semantic segmentation and ellipse fitting. The steel coil loosening recognition method based on semantic segmentation and ellipse fitting comprises the following steps: acquiring a steel coil image of a steel coil end face in real time; using a semantic segmentation model EfficientNet model to extract a steel coil loosening area mask of the steel coil image, wherein the steel coil loosening area mask comprises a head area mask and an inner ring area mask; extracting a head area point set and an inner ring area point set in the head area mask and the inner ring area mask; performing ellipse fitting on the head area point set and the inner ring area point set respectively to reconstruct a head arc line and an inner ring arc line respectively, and calculating interval information according to the head arc line and the inner ring arc line; and performing threshold value judgment on the interval information to determine whether steel coil loosening occurs.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of feature detection of ladle, in particular to a steel coil loosening recognition method and device based on semantic segmentation and ellipse fitting. BACKGROUND

[0002] Steel coil refers to the coil-shaped product processed by hot rolling, cold rolling, galvanizing and other processes. It is an important raw material for manufacturing various metal products. Steel coil has a wide range of applications, including construction, automobile manufacturing, home appliances, machinery manufacturing and other industries.

[0003] During the processing process, due to improper tension setting during coiling, coiling failure causing strip steel deformation on the roller, and factors such as loose binding, the steel coil may not be tightly coiled, at this time there is a gap between the layers of the steel coil, which makes the steel coil easy to loosen during further processing, resulting in safety accidents. Therefore, it is necessary to quickly detect the defects of the steel coil loosening and timely adjust the steel coil. Therefore, the steel coil loosening detection is an important part of the quality detection process in the factory.

[0004] Image recognition is commonly used in steel coil loosening discrimination. The steps of data collection, feature extraction, model training, and testing and discrimination are adopted. However, the direction of steel coil loosening is complex, that is, steel coil loosening has diversity. When the steel coil loosens, the strip steel has serious wave shape or other deformation, and the tightness of the steel coil fitting mode is different. It is difficult to extract the steel coil loosening area from the steel coil point cloud / pixel using simple geometric features. The error rate of conventional connected component analysis is high. In simple terms, the conventional data collection cannot cover all steel coil loosening scenarios, making subsequent feature extraction and other steps more difficult.

[0005] One method commonly used in image recognition to solve the above problem is to extract features through deep learning, trying to use transfer learning, and the other is to use multi-scale feature extraction. For the former, due to the particularity of the steel coil, the owner cannot bear the risk of obtaining samples for training in practice. For the latter, the development cost and efficiency are low.

[0006] Based on the above, the present application is proposed. SUMMARY

[0007] In the above image recognition application, due to the diversity and complexity of steel coil loosening, the conventional data collection, feature extraction and model training steps may not accurately determine the loosening area of the steel coil. Therefore, these methods have disadvantages in this specific problem, that is. In order to overcome these disadvantages and improve accuracy and reliability.

[0008] According to a first aspect of the present application, the present application provides a steel coil loosening recognition method based on semantic segmentation and ellipse fitting, the steel coil loosening recognition method comprising: acquiring a steel coil image of a steel coil end face in real time;

[0009] The semantic segmentation model EfficientNet model is used to extract a steel coil loosening region mask of the steel coil image, wherein the steel coil loosening region mask comprises a head band region mask and an inner ring region mask;

[0010] The head band region point set and the inner ring region point set in the head band region mask and the inner ring region mask are extracted;

[0011] Ellipse fitting is performed on the head band region point set and the inner ring region point set respectively to reconstruct a head band arc line and an inner ring arc line respectively, and interval information is calculated according to the head band arc line and the inner ring arc line;

[0012] Threshold judgment is performed on the interval information to determine whether the steel coil loosening occurs.

[0013] In a further scheme of the present application, the semantic segmentation model EfficientNet model is used to extract a steel coil loosening region mask of the steel coil image, wherein the steel coil loosening region mask comprises a head band region mask and an inner ring region mask, comprising:

[0014] The steel coil image is input into the EfficientNet model, and a mask of each pixel point output by the model is acquired;

[0015] Threshold processing and morphological operation are performed on the mask of each pixel point to obtain the head band region mask and the inner ring region mask.

[0016] In a further scheme of the present application, ellipse fitting is performed on the head band region point set and the inner ring region point set respectively to reconstruct a head band arc line and an inner ring arc line respectively, comprising:

[0017] A first geometric center of the inner ring contour is fitted according to the inner ring contour point set;

[0018] The first geometric center is taken as an initial long diameter parameter and a short diameter parameter of a growing ellipse, and an area of the ellipse is acquired at the same time;

[0019] When the area of the ellipse can surround all the inner ring contour point sets, the inner ring ellipse with the smallest area is extracted as an inner ring ellipse;

[0020] An inner diameter of the inner ring contour is acquired according to the mean value of the long diameter parameter and the short diameter parameter of the inner ring ellipse at this time;

[0021] The inner ring arc line is reconstructed according to the inner diameter, the geometric center, the long diameter parameter, the short diameter parameter and a rotation angle of the inner ring ellipse.

[0022] In a further aspect of the application, when the area of the ellipse can encompass all the inner ring contour point set, the inner ring ellipse is extracted as the ellipse with the smallest area, comprising:

[0023] The distance of each point in the inner ring contour point set to the ellipse is calculated;

[0024] The distance of each point to the ellipse is compared with the long diameter parameter of the ellipse;

[0025] When the distance of all points to the ellipse is less than or equal to the long diameter parameter of the ellipse, it is determined that the area of the ellipse can encompass all the inner ring contour point set, and the inner ring ellipse is extracted as the ellipse with the smallest area.

[0026] In a further aspect of the application, the band head arc line and the inner ring arc line are respectively reconstructed by ellipse fitting from the band head area point set and the inner ring area point set, comprising:

[0027] The second geometric center of the band head contour is fitted according to the band head contour point set;

[0028] The long diameter parameter and the short diameter parameter of the continuously growing ellipse are increased with the second geometric center as the initial value, and the area of the ellipse is obtained;

[0029] When the area of the ellipse can encompass all the band head contour point set, the band head ellipse is extracted as the ellipse with the smallest area;

[0030] The inner diameter of the band head ellipse is obtained according to the mean value of the long diameter parameter and the short diameter parameter of the band head ellipse at this time;

[0031] The band head arc line is reconstructed according to the inner diameter, the geometric center, the long diameter parameter, the short diameter parameter and the rotation angle of the inner ring ellipse.

[0032] In a further aspect of the application, the interval information is calculated according to the band head arc line and the inner ring arc line, comprising:

[0033] The nearest neighbor point of each coordinate point on the band head arc line on the inner ring arc line is calculated in a traversal manner;

[0034] The point distance of each coordinate point on the band head arc line and the nearest neighbor point on the inner ring arc line is obtained to obtain the interval information of the band head arc line and the inner ring arc line.

[0035] In a further aspect of the application, the interval information is threshold judged to determine whether the coil is loose, comprising:

[0036] The abnormal coordinate points with an interval greater than the interval threshold in the interval information are counted;

[0037] When the number of abnormal coordinate points exceeds the first threshold, it is judged that the coil is loose.

[0038] In a further aspect of the present application, in the case of steel coil loosening, the alarm is controlled to alarm.

[0039] In a further aspect of the present application, in the case of steel coil loosening, the alarm is controlled to alarm.

[0040] In a further aspect of the present application, in the case of steel coil loosening, the alarm is controlled to alarm.

[0041] In a further aspect of the present application, in the case of steel coil loosening, the alarm is controlled to alarm.

[0042] The features and advantages of the embodiments of the present application will be illustrated in the following detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the description of the embodiments or the prior art. Obviously, the drawings described below are some embodiments of the present application, and the drawings collected by those skilled in the art without creative labor can also be obtained.

[0044] Figure 1 The flow chart of the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application;

[0045] Figure 2 The specific flow chart of step S200 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application;

[0046] Figure 3 The specific flow chart of step S400 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application;

[0047] Figure 4 The specific flow chart of step S403 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application;

[0048] Figure 5 Another specific flow chart of step S400 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application;

[0049] Figure 6 The line schematic diagram of the reconstructed head arc and inner ring arc provided by the embodiment of the present application is exemplarily provided;

[0050] Figure 7 The further flow chart of step S400 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application;

[0051] Figure 8 The specific flow chart of step S500 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application; and

[0052] Figure 9 The module schematic diagram of the steel coil loosening recognition device based on semantic segmentation and ellipse fitting provided by the embodiment of the present application.

[0053] Label description

[0054] 100, steel coil loosening recognition device;

[0055] 10, image acquisition equipment; 20, upper computer;

[0056] 30. An alarm. DETAILED DESCRIPTION

[0057] In order to make the above and other features and advantages of the present application more comprehensible, the present application will be further described below with reference to the drawings. It should be understood that the specific embodiments given herein are by way of example only and are not intended to limit the present application.

[0058] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. It will be apparent, however, to one skilled in the art that the present application can be practiced without the specific details given herein. In other instances, well-known steps or operations have not been described in detail in order to avoid obscuring the present application.

[0059] Please first refer to Figure 1 , Figure 1 The flowchart of the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiments of the present application.

[0060] In one general inventive concept of the embodiments of the present application, a general flowchart of a steel coil loosening recognition method based on semantic segmentation and ellipse fitting is provided, comprising:

[0061] Step S100, acquiring a steel coil image of a steel coil end face in real time;

[0062] Step S200, extracting a steel coil loosening area mask of the steel coil image by using a semantic segmentation model EfficientNet model, wherein the steel coil loosening area mask comprises a head area mask and an inner ring area mask;

[0063] Step S300, extracting a head area point set and an inner ring area point set in the head area mask and the inner ring area mask;

[0064] Step S400, performing ellipse fitting on the head area point set and the inner ring area point set respectively to reconstruct a head arc line and an inner ring arc line respectively, and calculating interval information according to the head arc line and the inner ring arc line;

[0065] Step S500, performing threshold value judgment on the interval information to determine whether steel coil loosening occurs.

[0066] In the steel coil production line, the end face image of the steel coil can be obtained in real time through corresponding sensing devices or camera devices such as a CCD camera, a CMOS camera and the like, and the steel coil image is stored in a library and will be used as input for subsequent steps. Then, the EfficientNet model is used for semantic segmentation to extract the steel coil loose area mask of the steel coil image. It can be understood that the EfficientNet model can be trained and adjusted by weight to segment the loose area and other areas in the steel coil image. The segmentation result includes a “head area mask” and an “inner ring area mask”, which are respectively used to indicate the position of the steel coil loose area, for example:

[0067] The inference result of the EfficientNet model is a single-channel pixel image, each pixel indicating whether the model estimates it to belong to the inner coil loose category, for example: the head area is represented by 2, the inner ring area is represented by 1, and 0 is the background category. Therefore, the connected region of the pixel with 1 in the pixel image represents the inner ring area estimated by the model, and the connected region of the pixel with 2 represents the head area estimated by the model. The inner coil loose area mask is actually the mask of the two areas.

[0068] Then, the contour point set of the steel coil loose area mask is extracted to obtain the head area point set and the inner ring area point set: the mask with the class label of 1 and 2 is extracted from the steel coil loose area mask to obtain the contour point set with continuity. The head contour and the inner ring contour are two of them, which can represent the outer ring and the inner ring of the loose coil part of the steel coil, respectively, for the subsequent ellipse fitting process.

[0069] The inner ring ellipse is further fitted from the head area point set and the inner ring area point set, the diameter length information is obtained by fitting, and then the interval information is calculated according to the diameter length information, which is used to judge whether the steel coil is loose; by fitting the ellipse and obtaining the diameter length information, the deformation degree or loose condition of the steel coil can be quantified more accurately. By calculating the interval information, the uneven deformation of the steel coil surface can be understood more carefully. Finally, the threshold value of the interval information is judged to determine whether the steel coil is loose.

[0070] Through the above, the steel coil loose identification method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application can accurately locate and mark the loose area of the steel coil by obtaining the steel coil image of the steel coil end face in real time, extracting the steel coil loose area mask of the steel coil image by using the semantic segmentation model EfficientNet, automatically obtaining the head contour and inner ring contour of the steel coil from the image by extracting the mask and extracting the contour point set, fitting the inner ring ellipse from the head contour and inner ring contour and calculating the diameter length information, which can adapt to different types of steel coil loose and obtain more detailed steel coil surface deformation. Then, the interval information is calculated according to the diameter length information to provide a more in-depth analysis and judgment basis. This method has high reliability and generalization ability.

[0071] The following detailed description introduces each step in the above general inventive concept:

[0072] Please continue to refer to Figure 2 , Figure 2 The specific flow chart of step S200 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application is as follows.

[0073] In step S200, the semantic segmentation model EfficientNet model is used to extract the steel coil loosening area mask of the steel coil image, wherein the steel coil loosening area mask includes the head area mask and the inner ring area mask, which includes:

[0074] In step S201, the steel coil image is input into the EfficientNet model, and the mask of each pixel point output by the model is obtained.

[0075] In step S202, threshold processing and morphological operation are performed on the mask of each pixel point to obtain the head area mask and the inner ring area mask.

[0076] It can be understood that the steel coil image is input into the pre-trained EfficientNet model, the EfficientNet model processes each pixel point in the steel coil image, and outputs the corresponding mask, that is, the mask of each pixel point is obtained. Threshold processing is performed on the mask of each pixel point, and the threshold processing is to divide the pixel points in the mask into binary (for example, 0 and 1) or multi-value (for example, 0, 1 and 2) pixel points according to a specific threshold. Subsequently, morphological operations are performed, including dilation, erosion, etc. These operations can remove noise, connect regions, fill holes, etc. The head area mask and the inner ring area mask are obtained.

[0077] Then, the head area point set and the inner ring area point set in the head area mask and the inner ring area mask are extracted, which can usually be operated using functions in an image processing library (such as OpenCV). Classification is performed using the head area mask and the inner ring area mask, so as to be classified into the corresponding point set.

[0078] For example, the point set and the coordinates of the head area and the inner ring area can be extracted according to the following steps: traversing the head area mask and the inner ring area mask. For the head area mask, the coordinates of the pixel points with non-0 are added to the head area point set. For the inner ring area mask, the coordinates of the pixel points with non-0 are added to the inner ring area point set. In this process, the inner ring area point set is the position of the pixel point in the image, which is usually represented by (x, y), where x is the horizontal position and y is the vertical position. By traversing the pixel points and adding their coordinates to the corresponding point set, that is, the head area point set and the inner ring area point set.

[0079] Please continue to refer to Figure 3 , Figure 3 The specific flowchart of step S400 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application is as follows.

[0080] The step S400 of fitting an ellipse from the strip head contour and the inner ring contour respectively to obtain the diameter information respectively, and calculating the interval information of the strip head contour and the inner ring contour according to the diameter information includes:

[0081] Step S401, fitting the first geometric center of the inner ring contour according to the inner ring contour point set;

[0082] Step S402, taking the first geometric center as the initial long diameter parameter and short diameter parameter of the growing ellipse, and simultaneously obtaining the area of the ellipse;

[0083] Step S403, when the area of the ellipse can surround all the inner ring contour point set, extracting the ellipse with the smallest area as the inner ring ellipse;

[0084] Step S404, obtaining the inner diameter of the inner ring contour according to the mean value of the long diameter parameter and the short diameter parameter of the inner ring ellipse at this time;

[0085] Step S405, reconstructing the inner ring arc line according to the inner diameter, the geometric center, the long diameter parameter, the short diameter parameter and the rotation angle of the inner ring ellipse.

[0086] It can be understood that in step S401, the first geometric center can be obtained according to the centroid method, such as the center horizontal coordinate o1=(x1+x2+...+xn) / n; the center vertical coordinate o2=(y1+y2+...+yn) / n

[0087] Wherein, (x1, y1), (x2, y2),..., (xn, yn) are the coordinates of each point in the inner ring contour point set, and n is the number of points in the inner ring contour point set.

[0088] Then an ellipse equation is established: ((x-o1) / a)2+((y-o2) / b)2=1; wherein (o1, o2) is the center point coordinate of the ellipse, i.e. the first geometric center obtained above, and a and b are the lengths of the long diameter and the short diameter of the ellipse respectively. Then the long diameter parameter a and the short diameter parameter b of the inner ring ellipse at this time are initialized to be smaller non-zero values. Starting from the initial point, the long diameter parameter a and the short diameter parameter b are gradually increased, and the area thereof is calculated at the same time through the ellipse area formula.

[0089] It should be noted that the long diameter parameter a and the short diameter parameter b are not set to be synchronously increased in steps, but preferably are increased asynchronously, that is, the long diameter parameter a and the short diameter parameter b are alternately increased, and the step length of the increase can be specifically configured. Generally speaking, the shorter the step length, the higher the precision. However, the efficiency will also be reduced, so it can be debugged according to specific needs.

[0090] Please continue to refer to Figure 4 , Figure 4 A specific flowchart of step S403 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application.

[0091] In one way of judging the surrounding of all inner ring contour point sets:

[0092] Step S403 includes extracting the smallest ellipse as the inner ring ellipse when the area of the ellipse can surround all the inner ring contour point sets, including:

[0093] Step S4031, traversing to calculate the distance of each point in the inner ring contour point set to the ellipse;

[0094] Step S4032, comparing the distance of each point to the ellipse with the long diameter parameter of the ellipse;

[0095] Step S4033, in the case that all the distances of the points to the ellipse are less than or equal to the long diameter parameter of the ellipse, it is determined that the area of the ellipse can surround all the inner ring contour point sets, and at this time the smallest ellipse is extracted as the inner ring ellipse.

[0096] Wherein, the distance calculation formula can use the standard distance formula of the point to the ellipse to calculate; in the case that all the distances of the points to the ellipse are less than or equal to the long diameter parameter a of the ellipse, it is determined that all the inner ring contour point sets are surrounded.

[0097] Further, the values of the diameter parameter a and the short diameter parameter b are averaged to obtain an average value, that is, (a+b) / 2, which can be considered as the inner diameter of the inner ring ellipse. Further, the inner ring arc line is reconstructed according to the geometric center coordinates (o1, o2) of the inner ring, the inner diameter, the rotation angle of the inner ring ellipse and other parameters.

[0098] It can be understood that the rotation angle of the inner ring ellipse is obtained when the ellipse is fitted, and the rotation angle represents the rotation angle of the ellipse relative to the horizontal direction.

[0099] Please continue to refer to Figure 5 , Figure 5 Another specific flowchart of step S400 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application.

[0100] Ellipse fitting is performed on the band head contour and the inner ring contour respectively in step S400 to obtain radial length information respectively, and interval information of the band head contour and the inner ring contour is calculated according to the radial length information, including:

[0101] In step S401', a second geometric center of the band head contour is fitted according to the band head contour point set;

[0102] In step S402', the long diameter parameter and the short diameter parameter of the continuously growing ellipse are initialized with the second geometric center, and the area of the ellipse is obtained at the same time;

[0103] In step S403', when the area of the ellipse can enclose all the band head contour point set, the ellipse with the smallest area is extracted as the band head ellipse;

[0104] In step S404', the inner diameter of the band head ellipse is obtained according to the mean value of the long diameter parameter and the short diameter parameter of the band head ellipse at this time;

[0105] In step S405', the band head arc is reconstructed according to the inner diameter, the geometric center, the long diameter parameter, the short diameter parameter and the rotation angle of the inner ring ellipse.

[0106] The second radial length information of the band head arc is calculated, and the above-mentioned calculation of the inner ring arc can be similarly analogized, and will not be repeated again. Finally, as shown in Figure 6 .

[0107] Please continue to refer to Figure 7 , Figure 7 The further flowchart of the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application.

[0108] Step S400 and the interval information of the band head contour and the inner ring contour is calculated according to the radial length information, including:

[0109] In step S407, each coordinate point on the band head arc is traversed to obtain the nearest neighbor point on the inner ring arc.

[0110] In step S408, the point distance between each coordinate point on the band head arc and the nearest neighbor point on the inner ring arc is obtained to obtain the interval information of the band head arc and the inner ring arc.

[0111] Taking the inner ring arc as an example, after reconstructing the inner ring arc, the coordinate points of the reconstructed inner ring arc need to be obtained. For example, the inner ring arc is divided into several equal points, and the appropriate number of equal parts can be selected according to the needs. For example, the arc is divided into 360 points to cover the entire arc length, and the coordinates of each equal point are calculated. The parameter equation of the ellipse can be used to calculate the x and y coordinates of each point. The specific parameter equation is:

[0112] x1 = o1 + (d / 2) * cos(to1eta) * cos(po1i) - (f / 2) * sin(to1eta) * sin(po1i)

[0113] y1 = o2 + (d / 2) * cos(to1eta) * sin(po1i) + (f / 2) * sin(to1eta) * cos(po1i)

[0114] where (x1, y1) is the coordinate of the point on the inner ring arc, (o1, o2) is the geometric center coordinate of the inner ring arc, d is the inner diameter length, f is the minor axis length of the inner ring, to1eta is the rotation angle, and po1i is the parameter angle (0 to 2π) on the arc.

[0115] It can be understood that, by traversing all points on the inner ring arc for each coordinate point on the head arc, the distance from the point to all points on the inner ring arc is calculated, and the point on the inner ring arc with the minimum distance is found, and the interval information of the head arc and the inner ring arc is obtained. The method of calculating the distance can be selected according to the specific application scenario. Common methods of calculating the distance include Euclidean distance, Manhattan distance, etc. The selection of the appropriate method of calculating the distance depends on the characteristics and requirements of the coordinate points.

[0116] Please continue to refer to Figure 8 , Figure 8 The specific flowchart of step S500 in the steel coil loosening recognition method based on semantic segmentation and ellipse fitting provided by the embodiment of the present application.

[0117] In step S500, whether the steel coil loosening occurs is determined according to the interval information.

[0118] Step S501, count the abnormal coordinate points with an interval greater than the interval threshold in the interval information.

[0119] Step S502, when the number of abnormal coordinate points exceeds the first threshold, it is determined that the steel coil loosening occurs.

[0120] Step S500 is a threshold judgment on the interval information. Since the head arc and the inner ring arc are continuous, when the number of abnormal coordinate points is greater than the first threshold, it is proved that the steel coil loosening occurs.

[0121] In a further manner of the embodiment of the present application, the steel coil loosening recognition method further comprises:

[0122] Step S600, in the case of steel coil loosening, controlling the alarm to alarm.

[0123] Please refer to Figure 9 , Figure 9A module schematic diagram of the steel coil loosening recognition device 100 based on semantic segmentation and ellipse fitting provided by the embodiment of the present application.

[0124] A steel coil loosening recognition device 100 based on semantic segmentation and ellipse fitting, comprising:

[0125] An image acquisition device 10 for acquiring a steel coil image of a steel coil end face in real time;

[0126] A host computer 20, which is electrically connected with the image acquisition device and is configured to:

[0127] acquire a steel coil image of a steel coil end face in real time;

[0128] extract a steel coil loosening area mask of the steel coil image by using a semantic segmentation model EfficientNet model, wherein the steel coil loosening area mask comprises a band head area mask and an inner ring area mask;

[0129] extract contour point sets from the steel coil loosening area mask to obtain a band head contour and an inner ring contour respectively;

[0130] perform ellipse fitting on the band head contour and the inner ring contour respectively to obtain diameter length information respectively, and calculate interval information of the band head contour and the inner ring contour according to the diameter length information;

[0131] determine whether the steel coil loosening occurs according to the interval information.

[0132] Further, the steel coil loosening recognition device further comprises:

[0133] An alarm 30 electrically connected with the host computer 20;

[0134] The host computer 20 is electrically connected with the alarm 30 and is configured to:

[0135] control the alarm to alarm in the case of the steel coil loosening.

[0136] Further, it should be understood by those skilled in the art that if all or part of the sub-modules involved in each product of the brightness contrast adjusting device provided by the embodiment of the present application are combined, replaced, transformed, etc. by means of condensation, simple change, mutual transformation, etc., such as the movement of the positions of each component; or the integral setting of the product formed thereby; or the detachable design; as long as the combined components can form a device / apparatus / system with a specific function, the device / apparatus / system can replace the corresponding components of the present application and still fall within the protection scope of the present application.

[0137] It should be understood that the specific features, operations and details described herein in connection with the method of the present application can be similarly applied to the device and system of the present application, or vice versa. In addition, each step of the method of the present application described above can be performed by the corresponding component or unit of the device or system of the present application.

[0138] The present application provides a computer readable storage medium, and computer program instructions are stored on the computer readable storage medium, and the computer program instructions are executed by a processor to implement the brightness contrast adjustment method described above.

[0139] Those skilled in the art can understand that the steps of the method of the present application can be instructed by a computer program to complete the relevant hardware such as a measuring device or a processor, and the computer program that can implement the brightness contrast adjustment method described above can be stored in a non-transitory computer readable storage medium, and the computer program is executed to cause the steps of the present application to be executed. According to the circumstances, any reference to memory, storage or other medium herein can include non-volatile or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid state disk, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0140] The technical features described above can be combined arbitrarily. Although all possible combinations of these technical features are not described, any combination of these technical features should be considered to be covered by the present specification, as long as such a combination does not contradict.

[0141] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A steel coil loosening recognition method based on semantic segmentation and ellipse fitting, characterized in that, The method comprises the following steps: real-time acquisition of a coil image of a coil end face; extracting a coil loose area mask of the coil image by using a semantic segmentation model EfficientNet model, wherein the coil loose area mask comprises a head area mask and an inner ring area mask; extracting a head area point set and an inner ring area point set in the head area mask and the inner ring area mask; performing ellipse fitting on the head area point set and the inner ring area point set respectively to reconstruct a head arc line and an inner ring arc line respectively, and calculating interval information according to the head arc line and the inner ring arc line; threshold judgment on the interval information to determine whether coil loosening occurs; the ellipse fitting on the head area point set and the inner ring area point set respectively to reconstruct the head arc line and the inner ring arc line comprises: fitting a first geometric center of the inner ring contour according to the inner ring contour point set; taking the first geometric center as the long diameter parameter and the short diameter parameter of the continuously growing ellipse, and simultaneously obtaining the area of the ellipse; when the area of the ellipse can enclose all the inner ring contour point set, extracting the smallest ellipse as the inner ring ellipse; obtaining the inner diameter of the inner ring contour according to the mean value of the long diameter parameter and the short diameter parameter of the inner ring ellipse at this time; reconstructing the inner ring arc line according to the inner diameter, the geometric center, the long diameter parameter, the short diameter parameter and the rotation angle of the inner ring ellipse; the ellipse fitting on the head area point set and the inner ring area point set respectively to reconstruct the head arc line and the inner ring arc line comprises: fitting a second geometric center of the head contour according to the head contour point set; taking the second geometric center as the long diameter parameter and the short diameter parameter of the continuously growing ellipse, and simultaneously obtaining the area of the ellipse; when the area of the ellipse can enclose all the head contour point set, extracting the smallest ellipse as the head ellipse; obtaining the inner diameter of the head ellipse according to the mean value of the long diameter parameter and the short diameter parameter of the head ellipse at this time; and reconstructing the head arc line according to the inner diameter, the geometric center, the long diameter parameter, the short diameter parameter and the rotation angle of the inner ring ellipse.

2. The coil slack identification method according to claim 1, characterized by, the extraction of the coil loose area mask of the coil image by using the semantic segmentation model EfficientNet model, wherein the coil loose area mask comprises the head area mask and the inner ring area mask, comprises: inputting the coil image into the EfficientNet model to obtain the mask of each pixel point output by the model; performing threshold processing and morphological operation on the mask of each pixel point to obtain the head area mask and the inner ring area mask.

3. The coil slack identification method of claim 1, wherein when the area of the ellipse can enclose all the inner ring contour point set, extracting the smallest ellipse as the inner ring ellipse comprises: iteratively calculating the distance of each point in the inner ring contour point set to the ellipse; comparing the distance of each point to the ellipse with the long diameter parameter of the ellipse; in the case that the distance of all points to the ellipse is less than or equal to the long diameter parameter of the ellipse, it is determined that the area of the ellipse can enclose all the inner ring contour point set, and the smallest ellipse is extracted as the inner ring ellipse at this time.

4. The coil slack identification method of claim 1, wherein the calculation of the interval information according to the head arc line and the inner ring arc line comprises: traversing each coordinate point on the head arc, the nearest neighbor point on the inner ring arc is obtained; obtaining the point distance of each coordinate point on the head arc and the nearest neighbor point on the inner ring arc to obtain the interval information of the head arc and the inner ring arc.

5. The coil slack identification method according to claim 4, characterized by, threshold judgment is performed on the interval information to determine whether the coil loosening occurs, including: counting the abnormal coordinate points in the interval information whose interval is greater than the interval threshold value; when the number of abnormal coordinate points exceeds the first threshold value, it is determined that the coil loosening occurs.

6. The coil slack recognition method according to any one of claims 1 to 5, characterized by, The coil loosening identification method further includes: in the case of coil loosening, the alarm is controlled to alarm.

7. A coil looseness recognition device based on semantic segmentation and ellipse fitting, characterized by, including: an image acquisition device for acquiring the coil image of the coil end face in real time; a host computer, the host computer and the image acquisition device are electrically connected, and are configured to execute the coil loosening identification method of any one of claims 1 to 6.

8. The coil slack detection apparatus according to claim 7, characterized by Further comprising: an alarm electrically connected to the host computer; the host computer is electrically connected to the alarm, and is configured to: in the case of coil loosening, the alarm is controlled to alarm.

Images