Gun returning detection method, device and equipment for continuous casting billet and medium

Abstract

The invention provides a gun returning detection method, device and equipment for a continuous casting billet and a medium, and relates to the technical field of computer vision. By the adoption of the technical scheme, the gun returning steel flower can be rapidly determined, the gun returning phenomenon generated when the flame cutting vehicle cuts the continuous casting billet is further judged, and the gun returning phenomenon can be responded in time. And the probability that the continuous casting billet deviates due to the gun returning phenomenon is reduced, the possibility that the deviated continuous casting billet impacts other equipment to generate safety accidents is reduced, and the production rhythm is guaranteed.

Description

Technical Field

[0001] The present application relates to the field of computer vision technology, and particularly relates to a method, device, equipment and medium for detecting the return of the torch in continuous casting billets. Background Art

[0002] During the steel manufacturing process, a flame cutting vehicle can cut continuous casting billets into individual billets according to specified dimensions. When cutting continuous casting billets, incomplete cutting may occur, resulting in the phenomena of uncut-through and uncut continuous casting billets.

[0003] The return of the torch phenomenon is a phenomenon that occurs when the continuous casting billet is not completely cut. The main feature of this phenomenon is that a large amount of bright steel flowers and sparks are produced during the cutting process and sprayed upward or laterally from the continuous casting billet. Currently, the detection of the return of the torch phenomenon mainly relies on staff to view the monitoring of the continuous casting billet, and the response speed is slow when the return of the torch phenomenon occurs. Summary of the Invention

[0004] In view of this, the present application provides a method, device, equipment and medium for detecting the return of the torch in continuous casting billets, and the main purpose is to be able to quickly detect the return of the torch steel flowers that occur when a flame cutting vehicle cuts a continuous casting billet, so as to facilitate timely response to the return of the torch phenomenon according to the return of the torch steel flowers.

[0005] To achieve the above object, the first aspect of the present application discloses a method for detecting the return of the torch in continuous casting billets, including:

[0006] Obtain a first image when the flame cutting vehicle cuts the continuous casting billet;

[0007] Input the first image into a pre-trained steel flower detection model to output a first detection result on whether the first image contains the return of the torch steel flowers, where the steel flower detection model is obtained by inputting a plurality of first historical images and the steel flower detection labels corresponding to each of the first historical images as a training sample set into a first network to be trained;

[0008] If the first detection result is that the return of the torch steel flowers is included, it is determined that the return of the torch phenomenon is detected during the cutting process of the continuous casting billet.

[0009] Optionally, the obtaining the first image when the flame cutting vehicle cuts the continuous casting billet includes:

[0010] Obtain a second image when the flame cutting vehicle cuts the continuous casting billet, where the image acquisition range of the second image is larger than that of the first image;

[0011] Input the second image into a pre-trained cutting vehicle detection model to output a second detection result indicating whether the flame cutting vehicle is included in the second image, where the cutting vehicle detection model is obtained by inputting a plurality of second historical images and the corresponding flame cutting vehicle detection labels of each of the second historical images into a second network to be trained;

[0012] If the second detection result indicates that the flame cutting vehicle is included, crop the second image according to the position information of the flame cutting vehicle in the second image to obtain the first image including the flame cutting vehicle.

[0013] Optionally, the inputting the first image into a pre-trained steel flower detection model to output a first detection result indicating whether the returned gun steel flower is included in the first image includes:

[0014] Using the pre-trained steel flower detection model, output the first detection result indicating whether the returned gun steel flower is included in the first image based on the following steps:

[0015] Obtain the total number of pixel points in the first image;

[0016] Determine a binarization threshold according to the brightness feature values of each pixel point, and perform binarization division on each pixel point in the first image according to the brightness value to obtain the number of target pixel points, where the target pixel points are pixel points with a brightness value greater than the binarization threshold;

[0017] If the ratio between the number of target pixel points and the total number of pixel points is greater than or equal to a pixel ratio threshold, the returned gun steel flower is included in the first image;

[0018] Output the first detection result indicating that the returned gun steel flower is included in the first image.

[0019] Optionally, the inputting the second image into a pre-trained cutting vehicle detection model to output a second detection result indicating whether the flame cutting vehicle is included in the second image includes:

[0020] Using the pre-trained cutting vehicle detection model, output the second detection result indicating whether the flame cutting vehicle is included in the second image based on the following steps:

[0021] Obtain at least one feature coordinate value corresponding to the flame cutting vehicle in the second image, where the feature coordinate value represents the coordinate values of different components of the flame cutting vehicle in the second image;

[0022] Obtain the preset coordinate value of the flame cutting vehicle, where the preset coordinate value represents the coordinate values of different components of the preset flame cutting vehicle;

[0023] Determine whether the characteristic coordinate value matches the preset coordinate value;

[0024] If the proportion of the characteristic coordinate values that match the preset coordinate values among the characteristic coordinate values is greater than or equal to the coordinate ratio threshold, then the second image contains the flame cutting vehicle;

[0025] Crop the second image to obtain and output a first image containing the flame cutting vehicle.

[0026] Optionally, before inputting the first image into the pre-trained steel flower detection model, it further includes:

[0027] Preprocess the first image using at least one of contrast adjustment, local adaptive threshold adjustment, median filtering, or image sharpening algorithms;

[0028] The inputting the first image into the pre-trained steel flower detection model includes:

[0029] Input the preprocessed first image into the pre-trained steel flower detection model.

[0030] Optionally, the method further includes that the pre-training process of the cutting vehicle detection model includes:

[0031] Obtain a second network to be trained;

[0032] Obtain a plurality of second historical images of the flame cutting vehicle cutting the continuous casting billet under different working conditions;

[0033] Determine the historical coordinate values corresponding to the flame cutting vehicle in each of the second historical images respectively;

[0034] Add the flame cutting vehicle label to the historical coordinate value corresponding to each of the second historical images;

[0035] Use the plurality of second historical images and the flame cutting vehicle labels corresponding to the second historical images as the training data of the second network to be trained, and iterate the second network to be trained until the second network to be trained is trained to completion to obtain the cutting vehicle detection model.

[0036] Optionally, the method further includes that the pre-training process of the steel flower detection model includes:

[0037] Obtain a first network to be trained;

[0038] When the flame cutting vehicle cuts the continuous casting billet and returns gun steel flowers under different working conditions, obtain a plurality of first historical images containing the flame cutting vehicle;

[0039] In each of the first historical images, determine the pixel point positions corresponding to the returned gun steel flowers respectively;

[0040] Add the steel flower detection tags to all the pixel point positions corresponding to each of the first historical images;

[0041] Use the multiple first historical images and the corresponding steel flower detection tags as the training data for the first network to be trained, and iterate the first network to be trained until the first network to be trained is trained completely, so as to obtain the steel flower detection model.

[0042] A second aspect of the present application discloses a returned gun detection device for continuous casting billets, including:

[0043] An acquisition module, configured to acquire a first image when a flame cutting vehicle cuts the continuous casting billet;

[0044] A detection module, configured to input the first image into a pre-trained steel flower detection model to output a first detection result indicating whether the first image contains returned gun steel flowers, where the steel flower detection model is obtained by inputting a plurality of first historical images and the corresponding steel flower detection tags of each first historical image into a first network to be trained;

[0045] A determination module, configured to determine that a returned gun phenomenon is detected during the cutting process of the continuous casting billet if the first detection result indicates that the returned gun steel flowers are included.

[0046] In a third aspect embodiment of the present application, an electronic device is provided, including:

[0047] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method according to any one of the first aspect disclosed herein.

[0048] In a fourth aspect embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the method according to the first aspect is implemented.

[0049] In summary, according to the technical solution disclosed in the present application, first, a first image is obtained when a flame cutting vehicle cuts a continuous casting billet; then, the first image is input into a pre-trained steel flower detection model to output a first detection result indicating whether the first image contains a back-gun steel flower, where the steel flower detection model is obtained by inputting a plurality of first historical images and the steel flower detection labels corresponding to each first historical image as a training sample set into a first network to be trained; if the first detection result indicates the presence of a back-gun steel flower, it is finally determined that a back-gun phenomenon is detected during the cutting process of the continuous casting billet. In the technical solution of the present application, when the flame cutting vehicle cuts the continuous casting billet, the first image is taken, and the steel flower detection model is used to identify the first image. Since the steel flower detection model is trained by the first historical images containing steel flower detection labels, when the first image contains the back-gun steel flower generated when the flame cutting vehicle cuts the continuous casting billet, the steel flower detection model can quickly identify the steel flower and output the first detection result. The state where the flame cutting vehicle does not completely cut off the continuous casting billet can be represented as a back-gun phenomenon. When the back-gun phenomenon occurs, it is accompanied by the generation of back-gun steel flowers. When the first detection result contains back-gun steel flowers, the occurrence of the back-gun phenomenon can be directly determined. By adopting the technical solution of the present application, the determination of back-gun steel flowers can be quickly realized, and further, the back-gun phenomenon of the flame cutting vehicle cutting the continuous casting billet can be judged, so that the occurrence of the back-gun phenomenon can be responded to in a timely manner, reducing the probability of the continuous casting billet shifting due to the occurrence of the back-gun phenomenon, and reducing the possibility of a safety accident caused by the shifted continuous casting billet hitting other equipment, thus ensuring the production rhythm.

[0050] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the specific embodiments of the present application are hereinafter specifically exemplified. Brief Description of the Drawings

[0051] The accompanying drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present application and used together with the specification to explain the principles of the present application.

[0052] 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 accompanying drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0053] Figure 1 It shows a flowchart of a method for detecting the back-gun of a continuous casting billet provided by an embodiment of the present application;

[0054] Figure 2 It shows a schematic diagram of the detection of a flame cutting vehicle provided by an embodiment of the present applicationFigure 1 ;

[0055] Figure 3 Shows the schematic diagram of the detection of the flame cutting vehicle provided by the embodiment of the present application Figure 2 ;

[0056] Figure 4 Shows the schematic diagram of the steel flower detection provided by the embodiment of the present application Figure 1 ;

[0057] Figure 5 Shows the schematic diagram of the steel flower detection provided by the embodiment of the present application Figure 2 ;

[0058] Figure 6 Shows the structure diagram of a back gun detection device for continuous casting billets provided by the embodiment of the present application. Detailed implementation manners

[0059] In order to better understand the technical solutions provided by the embodiments of this specification, the technical solutions of the embodiments of this specification will be described in detail below through the accompanying drawings and specific embodiments. It should be understood that the specific features in the embodiments of this specification and the embodiments are detailed descriptions of the technical solutions of the embodiments of this specification, rather than limitations on the technical solutions of this specification. Without conflict, the technical features in the embodiments of this specification and the embodiments can be combined with each other.

[0060] In this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "including an..." does not exclude the presence of additional identical elements in the process, method, article or device including the element. The term "more than two" includes two or more than two.

[0061] In the modern steel production process, continuous casting technology is a key production link, and its quality directly affects the processing and use of subsequent products. The flame cutting of continuous casting billets is a process of cutting long billets continuously cast into individual billets according to specified dimensions by a flame cutting vehicle. At present, the flame cutting process is operated by a cutting gun installed on the flame cutting vehicle. However, due to factors such as insufficient fuel combustion of the cutting gun, relatively fast cutting speed, and quality problems of the cutting gun, the situation of incomplete cutting often occurs, resulting in the phenomena of uncut-through and uncut continuous casting billets. Once the continuous casting billet is not cut off, the adhered continuous casting billet will shift under the frictional force of rollers with different roller speeds below, and the shifted continuous casting billet may impact other equipment, causing safety accidents.

[0062] The "return gun" phenomenon is a phenomenon that occurs when the continuous casting billet is flame cut but not cut through. The main characteristics of its occurrence are that a large amount of bright steel flowers and sparks are splashed upward or sideways of the continuous casting billet during cutting. The main reason for its occurrence is that the cutting gun fails to cut through the casting billet, and the sparks ejected by the cutting gun and the steel flowers generated by flame cutting the steel billet cannot pass through the continuous casting billet and enter below the roller table, but are rebounded by the uncut-through continuous casting billet, resulting in a large amount of steel flowers and sparks being splashed above or sideways of the continuous casting billet. Since the return gun phenomenon occurs along with the uncut-through of flame cutting, the detection of uncut-through of flame cutting is realized by detecting the return gun phenomenon on the production site.

[0063] At present, the judgment of whether the continuous casting billet is completely cut off depends on the operator to determine by observing the monitor in real time, and the observation efficiency is relatively low. When the phenomenon of incomplete continuous casting billet occurs, the response speed is relatively slow.

[0064] Based on the above problems, this embodiment provides a method for detecting the return gun of continuous casting billets, as Figure 1 shown, including:

[0065] Step 101, obtain a first image when the flame cutting vehicle cuts the continuous casting billet.

[0066] A camera can be set around the flame cutting vehicle. When the flame cutting vehicle cuts the continuous casting billet, the camera can take pictures of the flame cutting vehicle and the continuous casting billet. Obtain the image containing the flame cutting of the continuous casting billet by the flame cutting vehicle from the camera as the first image.

[0067] Step 102, input the first image into a pre-trained steel flower detection model to output a first detection result on whether the first image contains return gun steel flowers, where the steel flower detection model is obtained by inputting a plurality of first historical images and the steel flower detection labels corresponding to each first historical image into a first network to be trained.

[0068] The steel flower detection model is a pre-trained visual model used to detect whether there is a back gun steel flower in an image. When the first image is input into the steel flower detection model, the steel flower detection model can quickly output an identification result indicating whether the first image contains a back gun steel flower. Exemplarily, the steel flower detection model can be a YOLOv8 model.

[0069] If the first image includes a back gun steel flower, the steel flower detection model can quickly identify the back gun steel flower from the first image and output a first identification result indicating that the first image contains a back gun steel flower.

[0070] Step 103, if the first detection result is that a back gun steel flower is included, it is determined that a back gun phenomenon is detected during the cutting process of the continuous casting billet.

[0071] If the first detection result indicates that the first image contains a back gun steel flower, and the back gun steel flower is an accompanying phenomenon when the flame cutting vehicle does not completely cut the continuous casting billet, it can be determined that a back gun phenomenon occurs.

[0072] In the technical solution of this embodiment, when the flame cutting vehicle cuts the continuous casting billet, the first image is taken, and the steel flower detection model is used to identify the first image. Since the steel flower detection model is trained from the first historical images containing steel flower detection labels, when the first image contains the back gun steel flower generated during the cutting of the continuous casting billet by the flame cutting vehicle, the steel flower detection model can quickly identify the steel flower and output the first detection result. The state where the flame cutting vehicle does not completely cut the continuous casting billet can be represented as a back gun phenomenon. When the back gun phenomenon appears, it is accompanied by the generation of back gun steel flowers. When the first detection result contains back gun steel flowers, the generation of the back gun phenomenon can be directly determined. As Figure 2 shown, the first image contains back gun steel flowers. After the first image is input into the steel flower detection model, the steel flower model can identify the back gun steel flowers in the first image, surround the back gun steel flowers in the first image with a border, and add a label to indicate the back gun steel flower (GH). The label can also include the confidence level of the back gun steel flower (shown as 0.97 in the figure). By adopting the technical solution of this application, the determination of back gun steel flowers can be quickly realized, and further, the back gun phenomenon of the flame cutting vehicle cutting the continuous casting billet can be judged, and the appearance of the back gun phenomenon can be responded to in a timely manner. The probability of the continuous casting billet shifting due to the generation of the back gun phenomenon is reduced, and the possibility of a safety accident caused by the shifted continuous casting billet hitting other equipment is lowered, ensuring the production rhythm.

[0073] In some embodiments, the first image is input into the pre-trained steel flower detection model to output a first detection result indicating whether the first image contains a back gun steel flower, including:

[0074] Using the pre-trained steel flower detection model, output the first detection result indicating whether the first image contains a back gun steel flower based on the following steps:

[0075] Obtain the total number of pixel points in the first image;

[0076] Determine the binarization threshold according to the brightness characteristic values of each pixel point, and perform binarization division on each pixel point in the first image according to the brightness value to obtain the number of target pixel points, where the target pixel points are pixel points with a brightness value greater than the binarization threshold;

[0077] If the ratio between the number of target pixel points and the total number of pixel points is greater than or equal to the ratio threshold, the first image contains returning gun steel flowers;

[0078] Output the first detection result that the first image contains returning gun steel flowers.

[0079] Since the returning gun steel flowers are the sparks generated when the cutting gun carried on the flame cutting vehicle cuts the continuous casting billet, the returning gun steel flowers present a form of high brightness. When the steel flower detection model recognizes that there are pixel points with high brightness values in the first image, it can determine that the first image contains returning gun steel flowers.

[0080] In the process of recognizing pixel points with high brightness values in the first image, the steel flower detection model can determine all pixel points in the first image and further determine the brightness value of each pixel point. On this basis, determine the brightness characteristics of the pixel points according to the brightness value of each pixel point and implement the setting of the binarization threshold. The set binarization threshold can be used to divide high-brightness pixel points and low-brightness pixel points, where the target pixel points represent pixel points with high brightness values in the first image. Exemplarily, as Figure 3 shown, Figure 3 the white points in represent high-brightness pixel points, which are used to represent returning gun steel flowers. When the number of high-brightness pixel points is large, it can be considered that there are a large number of high-brightness pixel points in the first image. When the number of high-brightness pixel points is greater than or equal to the pixel ratio threshold, it can be directly determined that the first image contains returning gun steel flowers, and the fact that the first image contains returning gun steel flowers is output as the first detection result.

[0081] The technical solution of this embodiment can divide the pixel points in the first image according to the brightness value to obtain the binarization threshold. Among them, pixel points with a brightness higher than the binarization threshold can be considered to be possibly the steel flowers in the image. When it is determined that the ratio of the high-brightness pixel points to all pixel points in the first image exceeds the preset pixel ratio threshold, it can be determined that the first image contains returning gun steel flowers, and further output the fact that the first image contains returning gun steel flowers as the detection result, realizing the accurate detection of steel flowers in the first image.

[0082] In some embodiments, before inputting the first image into the pre-trained steel flower detection model, it further includes:

[0083] Preprocess the first image using at least one of contrast adjustment, local adaptive threshold adjustment, median filtering, or image sharpening algorithm;

[0084] Input the first image into a pre-trained steel flower detection model, including:

[0085] Input the preprocessed first image into a pre-trained steel flower detection model.

[0086] This embodiment proposes that before inputting the first image into the steel flower detection model, at least one preprocessing method of contrast adjustment, local adaptive threshold adjustment, median filtering, or image sharpening algorithm can also be performed.

[0087] By adjusting the contrast, the difference between the dark and bright parts in the image can be made more obvious, so that the details in the image are clearer and it is convenient to divide the pixels in the image according to brightness.

[0088] By locally adaptively thresholding the first image, a grayscale image can be converted into a binary image (i.e., a black and white image). Different from global threshold adjustment, local adaptive threshold adjustment considers the local area around each pixel in the image instead of using a fixed threshold. This method is particularly effective when processing images with non-uniform illumination or complex backgrounds.

[0089] By median filtering, it can be used to remove noise in the image, especially the extremely bright or extremely dark pixel points randomly distributed in the image. By sorting the pixel values in the neighborhood and selecting the median value as the new value of the central pixel, these outliers can be effectively suppressed, and the edges and details of the image can be better retained.

[0090] By performing an image sharpening algorithm, it is used to enhance the edges and details of the image, making the image look clearer and sharper, enhancing the details in the image, and making the small structures in the image clearer.

[0091] In some embodiments, obtaining the first image when the flame cutting vehicle cuts the continuous casting billet includes:

[0092] Obtain the second image when the flame cutting vehicle cuts the continuous casting billet, where the image acquisition range of the second image is larger than that of the first image;

[0093] Input the second image into a pre-trained cutting vehicle detection model to output a second detection result indicating whether the second image contains a flame cutting vehicle, where the cutting vehicle detection model is obtained by inputting a plurality of second historical images and the corresponding flame cutting vehicle detection labels of each second historical image into a second network to be trained;

[0094] If the second detection result includes a flame cutting vehicle, the second image is cropped according to the position information of the flame cutting vehicle in the second image to obtain a first image including the flame cutting vehicle.

[0095] The first image input to the steel flower detection model can be generated by cutting the second image. The second image can be an image taken when the flame cutting vehicle cuts the continuous casting billet. If the second image is directly input to the steel flower detection model, in addition to the flame cutting vehicle and the continuous casting billet, the second image may also contain other objects. When performing the steel flower detection on the second image, there are many interference factors and high computing power consumption.

[0096] To reduce the interference factors in the picture, in this embodiment, the second image can be cropped to obtain the first image, where the acquisition range of the second image is larger than that of the first image, or it can also be expressed that the second image belongs to the first image. Here, the acquisition range refers to the field or range covered by information acquisition, and the information acquired in the second image is more than that in the first image. The cutting vehicle detection model adopted in this embodiment can crop a first image smaller than the second image from the second image, and the display range included in the first image can be used as the region of interest (ROI). Exemplarily, the cutting vehicle detection model can be the YOLOv8 model. As Figure 4 shown, Figure 4 it includes a strip steel 41 and a flame cutting vehicle 42. The strip steel 41 is conveyed past the flame cutting vehicle 42. Figure 4 Input to the cutting vehicle detection model, the cutting vehicle detection model can detect the flame cutting vehicle 42. As Figure 5 shown, the flame cutting vehicle detected by the cutting vehicle detection model is as Figure 5 shown. Figure 5 The border includes the flame cutting vehicle and outputs a label representing the flame cutting vehicle (QGC). The label can also include the confidence of the flame cutting vehicle (shown as 0.94 in the figure). The image in the border is used as the first image. By inputting the first image into the steel flower detection model, the probability of recognition errors caused by other interference factors in the second image can be reduced. For example, sparks may be generated at a position far from the flame cutting vehicle in the second image, and the sparks may affect the final recognition result of the steel flower.

[0097] When performing the cutting of the second image, in this embodiment, a cutting vehicle detection model is provided. The second image obtained by taking a picture can be input into the cutting vehicle detection model. The cutting vehicle detection model can detect the cutting vehicle in the second image and perform the cutting of the second image to generate the first image. Compared with the second image before the cutting action is performed, the image representing the cutting vehicle in the first image has a higher proportion in the first image. Inputting the first image into the steel flower detection model can reduce the computing power required to identify the returning gun steel flower and effectively reduce the possibility of recognition errors compared to inputting the second image captured by the camera into the steel flower detection model.

[0098] In some embodiments, inputting the second image into a pre-trained cutting vehicle detection model to output a second detection result indicating whether the second image contains a flame cutting vehicle includes:

[0099] Using the pre-trained cutting vehicle detection model, output the second detection result indicating whether the second image contains a flame cutting vehicle based on the following steps:

[0100] Obtain at least one feature coordinate value corresponding to the flame cutting vehicle in the second image, where the feature coordinate value represents the coordinate value position information of different components of the flame cutting vehicle in the second image;

[0101] Obtain the preset coordinate value of the flame cutting vehicle, where the preset coordinate value represents the coordinate values of different components of the preset flame cutting vehicle;

[0102] Determine whether the feature coordinate value matches the preset coordinate value;

[0103] If the proportion of the feature coordinate values that match the preset coordinate value among the feature coordinate values is greater than or equal to the coordinate ratio threshold, then the second image contains a flame cutting vehicle;

[0104] Crop the second image to obtain and output the first image containing the flame cutting vehicle.

[0105] In this embodiment, the specific implementation manner of the cutting vehicle detection model for detecting the flame cutting vehicle is specifically described. The cutting vehicle detection model can recognize each component of the flame cutting vehicle in the image and determine the coordinate points of each component in the second image as the feature coordinate values of the flame cutting vehicle. Compare the respective feature coordinate values obtained from the second image with the preset coordinate values to determine the matching degree between the coordinate values of the features of the flame cutting vehicle in the second image and the preset coordinate values. If the feature coordinate value matches the preset coordinate value, then it can be considered that the second image contains a flame cutting vehicle.

[0106] When it is determined that there is a flame cutting vehicle in the second image, the second image is cropped to obtain a first image containing the flame cutting vehicle, and the generated first image can be input into the steel flower detection model to achieve steel flower detection.

[0107] Before applying the steel flower detection model and the cutting vehicle detection model mentioned in this embodiment, the present embodiment also discloses how to train the steel flower detection model and the cutting vehicle detection model. Specifically, in some embodiments, the model pre-training process of the cutting vehicle detection model is described as follows:

[0108] Obtain a second network to be trained;

[0109] Obtain a plurality of second historical images of the flame cutting vehicle cutting the continuous casting billet under different working conditions;

[0110] Respectively determine the historical coordinate values corresponding to the flame cutting vehicle in each second historical image;

[0111] Add a flame cutting vehicle label to the historical coordinate values of each second historical image;

[0112] Use the multiple second historical images and the corresponding flame cutting vehicle labels as the training data of the second network to be trained, and iterate the second network to be trained until the training of the second network is completed to obtain the cutting vehicle detection model.

[0113] At the continuous casting billet production site, when the flame cutting vehicle cuts the continuous casting billet, image data during the flame cutting process is collected by installing a high-resolution camera. These image data should cover cutting scenarios under different cutting conditions (including different cutting speeds, ambient light, billet sizes, etc.).

[0114] Screen out the images that clearly show the position and operating state of the flame cutting vehicle from the collected image data. The screening criteria include the clarity of the images and the representativeness of the cutting scenarios.

[0115] Use the LabelImg annotation tool to accurately annotate the screened images. After annotation, the image set contains 1500 images and 3000 annotation boxes. Each image annotates the position and state of the cutting vehicle, and the position of the cutting vehicle can be represented in the form of the historical coordinate values corresponding to the flame cutting vehicle.

[0116] Divide the image set into a training set, a test set, and a validation set according to the ratio of 7:2:1 respectively. Ensuring the performance of the model on various data, and at the same time retaining an independent validation set helps to fairly evaluate the generalization ability of the model during the experiment.

[0117] Using the NVIDIA RTX 4090 training platform and the Pytorch deep learning framework, scale the original image resolution of 2560*1440 to 640*640 resolution. Use YOLOv8m as the base model to balance speed and accuracy. Set the Batch Size to 16, set the initial learning rate to 0.001, and adopt the cosine annealing strategy to adjust the learning rate. Train the second network to be trained for 100 rounds.

[0118] After each epoch, use the validation set to evaluate the model performance, monitor the accuracy and loss of the model, and adjust the training parameters to avoid overfitting, and obtain the cutting vehicle detection model with the best performance respectively.

[0119] The scheme of this embodiment can be used to train the cutting vehicle detection model. By using the second historical images containing the flame cutting vehicle labels as the image set and inputting them into the second network to be trained, the second network to be trained can be iteratively updated, so that the trained second network can detect and intercept the flame cutting vehicle.

[0120] In some embodiments, the model pre-training process of the steel flower detection model is described:

[0121] Obtain the first network to be trained;

[0122] When the flame cutting vehicle cuts the continuous casting billet and returns gun steel flowers under different working conditions, obtain multiple first historical images containing the flame cutting vehicle;

[0123] In each first historical image, respectively determine the pixel point positions corresponding to the returned gun steel flowers;

[0124] Add steel flower detection labels to all pixel point positions corresponding to each first historical image;

[0125] Use the multiple first historical images and the steel flower detection labels corresponding to the first historical images as the training data for the first network to be trained, and use them to iterate the first network to be trained until the first network to be trained is trained and the steel flower detection model is obtained.

[0126] Select images from the collected image data that clearly show the position of the flame cutting vehicle and its operating state, and images that clearly show the phenomenon of returned gun steel flowers. The screening criteria include the clarity of the images, the visibility of the steel flowers, and the representativeness of the cutting scenarios.

[0127] At the continuous casting billet production site, when the flame cutting vehicle cuts the continuous casting billet, collect image data during the flame cutting process by installing a high-resolution camera. These image data should cover cutting scenarios under different cutting conditions (including different cutting speeds, ambient light, billet sizes, etc.).

[0128] Use the LabelImg annotation tool to accurately annotate the filtered image data. After annotation, the image set contains 170 images, each with a bounding box annotating the area where the steel flower is ejected.

[0129] Divide the image set into a training set, a test set, and a validation set according to the ratio of 7:2:1. Ensure the performance of the model on various data, and at the same time, retaining an independent validation set helps to fairly evaluate the generalization ability of the model during the experiment.

[0130] Use the NVIDIA RTX 4090 training platform, adopt the Pytorch deep learning framework, scale the original image resolution of 2560*1440 to 640*640 resolution, use YOLOv8m as the basic model to balance speed and accuracy, set the Batch Size to 16, set the initial learning rate to 0.001, and adopt the cosine annealing strategy to adjust the learning rate, and train the first network to be trained for 100 rounds.

[0131] Evaluate the model performance using the validation set after each epoch, monitor the accuracy and loss of the model, adjust the training parameters to avoid overfitting, and obtain the steel flower detection model with the optimal performance respectively.

[0132] Adopting the solution of this embodiment can realize the training of the steel flower detection model. By using the first historical image containing the steel flower detection label as the image set and inputting it into the first network to be trained, and continuously iterating the first network to be trained, the trained steel flower detection model can realize the recognition of the steel flower.

[0133] This embodiment also discloses a back gun detection device for continuous casting billets, as Figure 6 shown, including:

[0134] An acquisition module 61, configured to acquire a first image when the flame cutting vehicle cuts the continuous casting billet;

[0135] A detection module 62, configured to input the first image into a pre-trained steel flower detection model to output a first detection result on whether the first image contains a back gun steel flower, where the steel flower detection model is obtained by using a plurality of first historical images and the steel flower detection label corresponding to each first historical image as a training sample set and inputting it into the first network to be trained;

[0136] A determination module 63, configured to determine that a back gun phenomenon is detected during the cutting process of the continuous casting billet if the first detection result is that the back gun steel flower is included.

[0137] In some embodiments, the acquisition module 61 is configured to:

[0138] Obtain a second image when the flame cutting vehicle cuts the continuous casting billet, wherein the image acquisition range of the second image is larger than that of the first image;

[0139] Input the second image into a pre-trained cutting vehicle detection model to output a second detection result on whether the flame cutting vehicle is included in the second image, wherein the cutting vehicle detection model is obtained by inputting a plurality of second historical images and the flame cutting vehicle detection labels corresponding to each of the second historical images into a second network to be trained;

[0140] If the second detection result is that the flame cutting vehicle is included, crop the second image according to the position information of the flame cutting vehicle in the second image to obtain the first image including the flame cutting vehicle.

[0141] In some embodiments, the detection module 62 is configured to:

[0142] Utilize a pre-trained steel flower detection model to output a first detection result on whether the first image includes back-gun steel flowers based on the following steps:

[0143] Obtain the total number of pixel points in the first image;

[0144] Determine a binarization threshold according to the brightness feature values of each of the pixel points, and perform binarization division on each of the pixel points in the first image according to the brightness value to obtain the number of target pixel points, wherein the target pixel points are pixel points with a brightness value greater than the binarization threshold;

[0145] If the ratio between the number of the target pixel points and the total number of the pixel points is greater than or equal to a pixel ratio threshold, the first image includes the back-gun steel flowers;

[0146] Output a first detection result that the first image includes the back-gun steel flowers.

[0147] In some embodiments, the acquisition module 61 is configured to:

[0148] Utilize a pre-trained cutting vehicle detection model to output a second detection result on whether the flame cutting vehicle is included in the second image based on the following steps:

[0149] Obtain at least one characteristic coordinate value corresponding to the flame cutting vehicle in the second image, where the characteristic coordinate value represents the coordinate values of different components of the flame cutting vehicle in the second image;

[0150] Obtain the preset coordinate value of the flame cutting vehicle, where the preset coordinate value represents the coordinate values of preset different components of the flame cutting vehicle;

[0151] Determine whether the characteristic coordinate value matches the preset coordinate value;

[0152] If the proportion of the characteristic coordinate values that match the preset coordinate value among the characteristic coordinate values is greater than or equal to the coordinate ratio threshold, then the second image contains the flame cutting vehicle;

[0153] Crop the second image to obtain and output a first image containing the flame cutting vehicle.

[0154] In some embodiments, the detection module 62 is further configured to:

[0155] Preprocess the first image by using at least one of contrast adjustment, local adaptive threshold adjustment, median filtering processing, or image sharpening algorithm;

[0156] The inputting the first image into the pre-trained steel flower detection model includes:

[0157] Input the preprocessed first image into the pre-trained steel flower detection model.

[0158] In some embodiments, the back gun detection device further includes: a training module 64, configured to:

[0159] Obtain a second network to be trained;

[0160] Obtain a plurality of second historical images of the flame cutting vehicle cutting the continuous casting billet under different working conditions;

[0161] Respectively determine the historical coordinate values corresponding to the flame cutting vehicle in each of the second historical images;

[0162] Add the flame cutting vehicle label to the historical coordinate value corresponding to each of the second historical images;

[0163] Use the plurality of second historical images and the flame cutting vehicle labels corresponding to the second historical images as the training data for the second network to be trained, and iterate the second network to be trained until the second network to be trained is trained to obtain the cutting vehicle detection model.

[0164] In some embodiments, the back gun detection device further includes: a training module 64, configured to:

[0165] Obtain a first network to be trained;

[0166] When the flame cutting vehicle has back gun steel flowers when cutting the continuous casting billet under different working conditions, obtain a plurality of first historical images containing the flame cutting vehicle;

[0167] In each of the first historical images, determine the pixel point positions corresponding to the returning gun steel flowers respectively;

[0168] Add the steel flower detection tags to all the pixel point positions corresponding to each of the first historical images;

[0169] Use the multiple first historical images and the corresponding steel flower detection tags as the training data for the first network to be trained, and iterate the first network to be trained until the training of the first network to be trained is completed to obtain the steel flower detection model.

[0170] It should be noted that in the above embodiments, the descriptions of the respective embodiments have their own focuses. For the parts not described in detail in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0171] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-readable program code.

[0172] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowcharts and / or block diagrams, and the combination of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0173] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0174] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide for implementing the process Figure 1 in one or more processes and / or boxes Figure 1 the steps of the functions specified in one or more boxes.

[0175] An embodiment of the present application also provides a computer program product, which includes computer software instructions. When the computer software instructions run on a processing device, the processing device is caused to execute the process of the method for detecting the return of a continuous casting billet.

[0176] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the embodiments of the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website, a computer, a server, or a data center to another website, a computer, a server, or a data center in a wired manner (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or a wireless manner (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be stored by a computer or a data storage device such as a server or a data center that includes one or more integrated available media. The available medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)), etc.

[0177] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0178] In several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in electrical, mechanical, or other forms.

[0179] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0180] In addition, each functional unit in various embodiments of this application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0181] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of this application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0182] The above embodiments are only used to illustrate the technical solution of this application, rather than to limit it; although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of various embodiments of this application.

[0183] Although the preferred embodiments of the present specification have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the present specification. Obviously, those skilled in the art can make various changes and variations to the present specification without departing from the spirit and scope of the present specification. Thus, if these modifications and variations of the present specification fall within the scope of the claims of the present specification and their equivalent technologies, the present specification is also intended to include these modifications and variations.

Claims

1. A method for detecting the return of a lance in a continuous casting billet, characterized in that, Including: Obtaining a first image when a flame cutting vehicle cuts the continuous casting billet; Inputting the first image into a pre-trained steel flower detection model to output a first detection result on whether the first image contains a return gun steel flower, where the steel flower detection model is obtained by inputting a plurality of first historical images and the steel flower detection labels corresponding to each of the first historical images into a first network to be trained; If the first detection result is that the return gun steel flower is included, it is determined that a return gun phenomenon is detected during the cutting of the continuous casting billet.

2. The method according to claim 1, wherein The obtaining of the first image when the flame cutting vehicle cuts the continuous casting billet includes: Obtaining a second image when the flame cutting vehicle cuts the continuous casting billet, where the image acquisition range of the second image is larger than that of the first image; Inputting the second image into a pre-trained cutting vehicle detection model to output a second detection result on whether the second image contains the flame cutting vehicle, where the cutting vehicle detection model is obtained by inputting a plurality of second historical images and the flame cutting vehicle detection labels corresponding to each of the second historical images into a second network to be trained; If the second detection result is that the flame cutting vehicle is included, the second image is cropped according to the position information of the flame cutting vehicle in the second image to obtain the first image including the flame cutting vehicle.

3. The method according to claim 1, characterized in that, The inputting of the first image into a pre-trained steel flower detection model to output a first detection result on whether the first image contains a return gun steel flower includes: Using the pre-trained steel flower detection model to output a first detection result on whether the first image contains a return gun steel flower based on the following steps: Obtaining the total number of pixel points in the first image; Determining a binarization threshold according to the brightness feature values of each pixel point, and performing binarization division on each pixel point in the first image according to the brightness value to obtain the number of target pixel points, where the target pixel points are pixel points with a brightness value greater than the binarization threshold; If the ratio between the number of target pixel points and the total number of pixel points is greater than or equal to a pixel ratio threshold, the return gun steel flower is included in the first image; Outputting a first detection result that the return gun steel flower is included in the first image.

4. The method according to claim 2, wherein The inputting of the second image into a pre-trained cutting vehicle detection model to output a second detection result on whether the second image contains the flame cutting vehicle includes: Using the pre-trained cutting vehicle detection model to output a second detection result on whether the second image contains the flame cutting vehicle based on the following steps: Obtaining at least one characteristic coordinate value corresponding to the flame cutting vehicle in the second image, where the characteristic coordinate value represents the coordinate values of different components of the flame cutting vehicle in the second image; Obtaining a preset coordinate value of the flame cutting vehicle, where the preset coordinate value represents the coordinate values of different components of the preset flame cutting vehicle; Determining whether the characteristic coordinate value matches the preset coordinate value; If the proportion of the characteristic coordinate values matching the preset coordinate values among the characteristic coordinate values is greater than or equal to the coordinate ratio threshold, then the second image contains the flame cutting vehicle; Crop the second image to obtain and output a first image containing the flame cutting vehicle.

5. The method according to claim 1, characterized in that, Before inputting the first image into the pre-trained steel flower detection model, it further includes: Preprocess the first image using at least one of contrast adjustment, local adaptive threshold adjustment, median filtering, or image sharpening algorithms; The step of inputting the first image into the pre-trained steel flower detection model includes: Input the preprocessed first image into the pre-trained steel flower detection model.

6. The method according to claim 2, characterized in that The method further includes that the pre-training process of the cutting vehicle detection model includes: Obtain a second network to be trained; Obtain a plurality of second historical images of the flame cutting vehicle cutting the continuous casting billet under different working conditions; Respectively determine the historical coordinate values corresponding to the flame cutting vehicle in each of the second historical images; Add the flame cutting vehicle label to the historical coordinate values corresponding to each of the second historical images; Use the plurality of second historical images and the flame cutting vehicle labels corresponding to the second historical images as the training data of the second network to be trained, and iterate the second network to be trained until the second network to be trained is trained to completion to obtain the cutting vehicle detection model.

7. The method according to claim 6, wherein The method further includes that the pre-training process of the steel flower detection model includes: Obtain a first network to be trained; When the flame cutting vehicle cuts the continuous casting billet and returns gun steel flowers under different working conditions, obtain a plurality of first historical images containing the flame cutting vehicle; In each of the first historical images, respectively determine the pixel point positions corresponding to the returned gun steel flowers; Add the steel flower detection label to all the pixel point positions corresponding to each of the first historical images; Use the plurality of first historical images and the steel flower detection labels corresponding to the first historical images as the training data of the first network to be trained, and iterate the first network to be trained until the first network to be trained is trained to completion to obtain the steel flower detection model.

8. A back gun detection device for continuous casting billets, characterized in that, It includes: An acquisition module for acquiring a first image when the flame cutting vehicle cuts the continuous casting billet; A detection module for inputting the first image into the pre-trained steel flower detection model to output a first detection result indicating whether the first image contains returned gun steel flowers, where the steel flower detection model is obtained by inputting a plurality of first historical images and the steel flower detection labels corresponding to each of the first historical images into the first network to be trained; A determination module for determining that a returned gun phenomenon is detected during the cutting process of the continuous casting billet if the first detection result is that the returned gun steel flowers are included.

9. An electronic device, characterized in that, It includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images