Scrap steel hopper furnace entering posture rapid identification method based on Yolov11-seg model preset anchor

Abstract

The invention relates to a method for rapidly recognizing the furnace entering posture of a scrap steel hopper based on Yolov11-seg model preset anchor. The method comprises the following steps that S1, the position of a camera is determined, and the view field is adjusted; s2, transmitting a video stream collected by a camera to an NVR for video storage, and then making a required data set; s3, adopting a Yolov11-seg detection model to carry out real-time scrap steel bucket detection on the camera; s4, calculating a duty ratio by using the mask, the detection frame and a preset anchor; and S5, event identification is carried out, and whether the scrap steel hopper prepares to be fed in front of the furnace or is fed or the view of the camera is removed after feeding is completed is judged. According to the overall structure provided by the embodiment of the invention, feeding counting and charging waste steel weight calculation can be realized, data support is provided for the subsequent steelmaking process, the waste steel bucket adopting the visual technology can avoid the personal safety problem, the duty ratio can be rapidly calculated based on the mask and the detection frame of the waste steel bucket and by combining the preset anchors at the lower left corner and the upper right corner, and the detection accuracy is improved. And the attitude of the scrap steel bucket can be accurately calculated.

Description

Technical Field

[0001] The present application relates to the field of computer software technology, and in particular to a method for quickly identifying the posture of a scrap bucket entering a furnace based on a preset anchor of a Yolov11-seg model. Background Art

[0002] Steel is one of the most important materials used and relied upon by humanity now and in the future. Lightweight, excellent strength and toughness, and long life are the most crucial properties of steel. Converter steelmaking is a key steelmaking process that uses molten iron, scrap steel, and ferroalloys as its primary raw materials. It oxidizes carbon and other impurities (such as silicon and manganese) in the pig iron to produce steel with improved physical, chemical, and mechanical properties. With the rapid development of deep learning technology, particularly in the field of vision, deep learning vision technology has become a powerful tool for improving the automation of steelmaking processes. In converter steelmaking, traditional methods of counting the number of scrap hopper feeds and calculating the weight of incoming scrap are low in intelligence, inefficient, require significant manual intervention, and pose safety risks. Automating feed processing using vision technology can reduce manual labor while providing objective and accurate information on feed parameters, thereby improving production safety and efficiency.

[0003] Under existing technological conditions, despite the many advantages of deep learning vision technology, in steelmaking environments, due to poor lighting conditions, the trained state recognition model has a low recognition rate for scrap steel feeding status, and the recognition speed is relatively slow. To improve the accuracy of state recognition, this paper proposes a fast scrap steel bucket feeding posture recognition technology based on the preset anchor of the Yolov11-seg model. Compared with other similar algorithms, this technology has significant performance improvements in recognition speed, accuracy, and robustness. To this end, we propose a fast scrap steel bucket feeding posture recognition method based on the preset anchor of the Yolov11-seg model. Summary of the Invention

[0004] This application provides a method for quickly identifying the posture of scrap bucket entering the furnace based on the preset anchor of the Yolov11-seg model to solve the above-mentioned problems.

[0005] This application provides a method for quickly identifying the posture of a scrap bucket entering a furnace based on a preset anchor of the Yolov11-seg model, comprising the following steps:

[0006] S1: Camera position determination and field of view adjustment;

[0007] S2: Transmit the video stream collected by the camera to the NVR for video storage and then generate the required data set;

[0008] S3: Use the Yolov11-seg detection model to perform real-time scrap bucket detection on the camera;

[0009] S4: Calculate the duty cycle using the mask, detection box, and preset anchor;

[0010] S5: Event recognition, determining whether the scrap steel bucket in front of the furnace is preparing to load, loading, or has completed loading and is being removed from the camera's field of view.

[0011] Preferably, the camera position determination and field of view adjustment include the following process:

[0012] S11: The camera position is determined. The position must ensure that the converter port and scrap bucket are within the camera's field of view.

[0013] S12: Adjust the field of view, adjust the camera focal length or the camera up and down angle until the camera can capture 2 / 3 of the scrap bucket when the scrap bucket is loaded.

[0014] Preferably, the process of transmitting the video stream captured by the camera to the NVR for video storage and then generating the required data set further includes the following process:

[0015] S21: Convert the original video into images, and select valid images containing scrap steel buckets that have certain differences;

[0016] S22: Label the scrap bucket image and create the detection data set required for the segmentation model;

[0017] S23: Perform secondary screening on the collected scrap steel bucket images to ensure that the training set contains images without furnace light interference, with strong light interference, and with weak light interference.

[0018] Preferably, the Yolov11-seg detection model is used to perform real-time scrap bucket detection and segmentation on the camera, which includes the following processes:

[0019] S31 pre-training parameters are selected, and the Yolov11-seg model is trained using gradient descent, with the SGD optimizer used. The model input image size is img = (640, 640, 3);

[0020] S32: Model training. By observing the training loss and validation set loss, several sets of better models are determined. The stability of the model is then verified on the test set to determine the final model. If the effect does not meet the expectations, the training parameters need to be changed and training continues until the expected effect is achieved.

[0021] If the number of training times cannot meet the pre-fetch requirement, then increase the training set, adjust the parameters, and iterate the training until the segmentation requirement is finally met.

[0022] S33: In the model inference stage, the segmentation model outputs the scrap bucket's boding box and segmentation mask. The scrap bucket is then cropped from the 2K original image using the box, and the non-scrap bucket area is replaced with 0 pixels using the mask value. The mask calculation formula is as follows:

[0023]

[0024] Where: q ij Confidence in model predictions.

[0025] Preferably, the step of calculating the duty cycle using the mask, the detection frame and the preset anchor is as follows:

[0026] S41: Preset anchors, with the lower left and upper right corners of the binding box as the anchor zero points, and lay out two anchors to verify the proportion of 0 pixels in the area; x1, y1, x2, y2 are the width of the upper left vertex, the height of the upper left vertex, the width of the lower right vertex, and the height of the lower right vertex of the binding box respectively:

[0027] w=x2-x1

[0028] h=y2-y1

[0029] The width and height of the scrap bucket detection box can be calculated based on x1, y1, x2, and y2, where w represents width and h represents height.

[0030] Anchor width w a and high h a The calculation formula is defined as follows:

[0031] w a =0.3w

[0032] h a =0.2h

[0033] After determining the width and height of the anchor, the coordinate mapping of the four anchors can be calculated. For example, the following formula is the real coordinate mapping calculation of one of the anchors in the lower left corner.

[0034]

[0035] in, and The coordinates of the upper left vertex of the preset anchor, and The coordinates of the lower right vertex of the preset anchor;

[0036] S42: Calculate the percentage of pixels with a lower left corner pixel value of 0 based on the binding box and the two preset anchors in the lower left corner, as shown in the following formula:

[0037]

[0038] Among them, Z l1 is the total number of 0 elements, W l1 and H l1 The width and height of the anchor, respectively, l1 It is the ratio of the number of zero pixels of an anchor in the lower left corner to the pixel format of the entire image;

[0039] S43: Calculate the proportion of 0 pixels in the upper right corner based on the binding box and the two preset anchors in the upper right corner;

[0040]

[0041] Among them, Z rl is the total number of 0 elements, W rl and H rl The width and height of the anchor, respectively, rl The ratio of the number of zero pixels of an anchor in the upper right corner to the pixel format of the entire image;

[0042] S44: Calculate the final scrap bucket feeding posture based on the proportion of 0 pixels in the lower left corner and the upper right corner. If the following formula is satisfied, the scrap bucket posture is the feeding posture;

[0043] max(ratio l1 ,ratio l2 )≤0.3

[0044] max(ratio r1 ,ratio r2 )≤0.5

[0045] Preferably, the event recognition is to determine whether the scrap steel bucket is preparing to be charged, is being charged, or is being removed from the camera's field of view after charging, in the following steps:

[0046] S51: The scrap bucket model based on Yolov11-seg will detect the scrap bucket at a frequency of 3 frames per second for 2 consecutive seconds. When the effective frequency of the scrap bucket is greater than 80%, that is, r ≥ 0.8, it is defined as the start of the event and the scrap bucket begins to load. The formula is as follows:

[0047]

[0048] f = bool(Model(x))

[0049] Where x is the current input image, Model is the neural network model, F is the sum of 6 detection results, and r is the effective frequency;

[0050] S52: After the event starts, the duty cycle calculation according to step S4 will be continuously performed until it is detected that the duty cycle meets the feeding conditions. This is defined as a feeding event, and the scrap steel bucket starts to feed into the furnace;

[0051] S53: When no feeding event is detected within 2 consecutive seconds, the scrap bucket detection frequency will be executed for 2 consecutive seconds and 3 frames per second until the frequency of detecting the scrap bucket accounts for less than 10%. Then the task event ends and there is no scrap bucket in the field of view;

[0052] S54: Loop execution: Detection is performed at a scrap bucket detection frequency of 3 frames per second for 2 consecutive seconds until the next event starts.

[0053] The above technical solution provided by the embodiment of the present application has the following advantages compared with the prior art:

[0054] Compared with the prior art, the overall structure provided by the embodiment of the present application achieves:

[0055] 1. It can be used to count incoming materials and calculate the weight of scrap steel entering the furnace, providing data support for subsequent steelmaking processes. Using vision technology to count scrap steel can not only avoid personal safety issues but also save significant labor costs, making it a key step in the transformation of steel companies into intelligent production.

[0056] 2. The Yolov11-seg model is used to detect, segment and identify scrap buckets. Even in very harsh lighting environments, the scrap bucket can be identified and the feeding posture can be calculated quickly and accurately.

[0057] 3. Based on the scrap bucket mask and detection frame, combined with the preset anchors in the lower left and upper right corners, the duty cycle can be quickly calculated, thereby accurately calculating the scrap bucket's posture. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0059] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0060] Figure 1 It is the overall principle diagram of the present invention;

[0061] Figure 2 It is a schematic diagram of the process of the present invention;

[0062] Figure 3 This is a schematic diagram of duty cycle calculation of the present invention;

[0063] Figure 4 This is a schematic diagram of the start, feeding and end of the event of the present invention;

[0064] Figure 5 This is a schematic diagram of the anchor coordinate mapping of the present invention. DETAILED DESCRIPTION

[0065] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0066] The various embodiments of the present application may be presented in the form of a range. It should be understood that the description in the form of a range is merely for convenience and brevity and should not be construed as a rigid limitation on the scope of the present application. Therefore, it should be considered that the range description has specifically disclosed all possible sub-ranges and single numerical values ​​within the range. For example, it should be considered that the range description from 1 to 6 has specifically disclosed sub-ranges, such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6, etc., as well as single numbers within the range, such as 1, 2, 3, 4, 5 and 6, regardless of the range. In addition, whenever a numerical range is indicated in this application, it is intended to include any quoted number (fraction or integer) within the indicated range. Unless otherwise specified, the various raw materials, reagents, instruments and equipment used in this application are all commercially available or can be prepared using existing equipment.

[0067] In this application, unless otherwise specified, the directional words used, such as "upper" and "lower", specifically refer to the directions of the drawings in the accompanying drawings. In addition, in this application, the terms "including", "comprising", etc. mean "including but not limited to". In this application, relational terms such as "first" and "second" are merely 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. In this application, "and / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. Wherein A and B can be singular or plural. In this application, "at least one" means one or more, and "plurality" means two or more. "At least one", "at least one of the following" or similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, "at least one of a, b, or c" or "at least one of a, b and c" can both mean: a, b, c, ab, i.e. a and b, ac, bc or abc, where a, b, c can be single or multiple.

[0068] like Figure 1-Figure 5 As shown: This embodiment of the application provides a method for quickly identifying the posture of a scrap bucket entering a furnace based on a preset anchor of the Yolov11-seg model, comprising the following steps:

[0069] S1: Camera position determination and field of view adjustment;

[0070] S2: Transmit the video stream collected by the camera to the NVR for video storage and then generate the required data set;

[0071] S3: Use the Yolov11-seg detection model to perform real-time scrap bucket detection on the camera;

[0072] S4: Calculate the duty cycle using the mask, detection box, and preset anchor;

[0073] S5: Event recognition, determining whether the scrap steel bucket in front of the furnace is preparing to load, loading, or has completed loading and is being removed from the camera's field of view.

[0074] The camera position determination and field of view adjustment include the following processes:

[0075] S11: The camera position is determined. The position must ensure that the converter port and scrap bucket are within the camera's field of view.

[0076] S12: Adjust the field of view, adjust the camera focal length or the camera up and down angle until the camera can capture 2 / 3 of the scrap bucket when the scrap bucket is loaded.

[0077] The process of transmitting the video stream collected by the camera to the NVR for video storage and then generating the required data set also includes the following steps:

[0078] S21: Convert the original video into images, and select valid images containing scrap steel buckets that have certain differences;

[0079] S22: Label the scrap bucket image and create the detection data set required for the segmentation model;

[0080] S23: Perform secondary screening on the collected scrap steel bucket images to ensure that the training set contains images without furnace light interference, with strong light interference, and with weak light interference.

[0081] The Yolov11-seg detection model is used to perform real-time scrap bucket detection and segmentation on the camera, which includes the following processes:

[0082] S31 pre-training parameters are selected, and the Yolov11-seg model is trained using gradient descent, with the SGD optimizer used. The model input image size is img = (640, 640, 3);

[0083] S32: Model training. By observing the training loss and validation set loss, several sets of better models are determined. The stability of the model is then verified on the test set to determine the final model. If the effect does not meet the expectations, the training parameters need to be changed and training continues until the expected effect is achieved.

[0084] If the number of training times cannot meet the pre-fetch requirement, then increase the training set, adjust the parameters, and iterate the training until the segmentation requirement is finally met.

[0085] S33: In the model inference stage, the segmentation model outputs the scrap bucket's boding box and segmentation mask. The scrap bucket is then cropped from the 2K original image using the box, and the non-scrap bucket area is replaced with 0 pixels using the mask value. The mask calculation formula is as follows:

[0086]

[0087] Where: q ij Confidence in model predictions.

[0088] The steps for calculating the duty cycle using the mask, detection frame, and preset anchor are as follows:

[0089] S41: Preset the anchor, and use the lower left corner and upper right corner of the binding box as the anchor zero points to lay out two anchors to verify the proportion of 0 pixels in the area; x1, y1, x2, y2 are the width of the upper left vertex, the height of the upper left vertex, the width of the lower right vertex, and the height of the lower right vertex of the binding box respectively:

[0090] w=x2-x1

[0091] h=y2-y1

[0092] The width and height of the scrap bucket detection box can be calculated based on x1, y1, x2, and y2, where w represents width and h represents height.

[0093] Anchor width w a and high h a The calculation formula is defined as follows:

[0094] w a =0.3w

[0095] h a =0.2h

[0096] After determining the width and height of the anchor, the coordinate mapping of the four anchors can be calculated. For example, the following formula is the real coordinate mapping calculation of one of the anchors in the lower left corner.

[0097]

[0098] in, and The coordinates of the upper left vertex of the preset anchor, and The coordinates of the lower right vertex of the preset anchor;

[0099] S42: Calculate the percentage of pixels with a lower left corner pixel value of 0 based on the binding box and the two preset anchors in the lower left corner, as shown in the following formula:

[0100]

[0101] Among them, Z l1 is the total number of 0 elements, W l1 and H l1 The width and height of the anchor, respectively, l1 It is the ratio of the number of zero pixels of an anchor in the lower left corner to the pixel format of the entire image;

[0102] S43: Calculate the proportion of 0 pixels in the upper right corner based on the binding box and the two preset anchors in the upper right corner;

[0103]

[0104] Among them, Z rl is the total number of 0 elements, W rl and H rl The width and height of the anchor, respectively, rl The ratio of the number of zero pixels of an anchor in the upper right corner to the pixel format of the entire image;

[0105] S44: Calculate the final scrap bucket feeding posture based on the proportion of 0 pixels in the lower left corner and the upper right corner. If the following formula is satisfied, the scrap bucket posture is the feeding posture;

[0106] max(ratio l1 ,ratio l2 )≤0.3

[0107] max(ratio r1 ,ratio r2 )≤0.5

[0108] The event recognition determines whether the scrap bucket is preparing to load, loading, or removing from the camera's field of view after loading. The steps are as follows:

[0109] S51: The scrap bucket model based on Yolov11-seg will detect the scrap bucket at a frequency of 3 frames per second for 2 consecutive seconds. When the effective frequency of the scrap bucket is greater than 80%, that is, r ≥ 0.8, it is defined as the start of the event and the scrap bucket begins to load. The formula is as follows:

[0110]

[0111] f = bool(Model(x))

[0112] Where x is the current input image, Model is the neural network model, F is the sum of 6 detection results, and r is the effective frequency;

[0113] S52: After the event starts, the duty cycle calculation according to step S4 will be continuously performed until it is detected that the duty cycle meets the feeding conditions. This is defined as a feeding event, and the scrap steel bucket starts to feed into the furnace;

[0114] S53: When no feeding event is detected within 2 consecutive seconds, the scrap bucket detection frequency will be executed for 2 consecutive seconds and 3 frames per second until the frequency of detecting the scrap bucket accounts for less than 10%. Then the task event ends and there is no scrap bucket in the field of view;

[0115] S54: Loop execution: Detection is performed at a scrap bucket detection frequency of 3 frames per second for 2 consecutive seconds until the next event starts.

[0116] The present invention first sets up a gun at a fixed position in front of the converter and ensures that the camera captures an area of ​​about 2 / 3 of the scrap steel bucket; secondly, the scrap steel bucket is masked and the coordinate frame is recognized in real time through the instance segmentation model; then the furnace entry status is recognized by calculating the duty ratio of the lower left corner and the upper right corner; finally, event detection is performed, that is, when the scrap steel bucket detection is performed for 2 consecutive seconds and 3 frames per second, when the frequency of the detected scrap steel bucket is greater than 80%, the event is considered to have started, and then the duty ratio calculation is started until the feeding state is detected, and the feeding event is entered; finally, when the scrap steel bucket detection is performed for 2 consecutive seconds and 3 frames per second, when the proportion of the detected scrap steel bucket is less than 10%, it is marked as the end of the event and the state enters the waiting state;

[0117] That is, the Yolov11-seg segmentation model is used to perform real-time scrap bucket segmentation on the camera video stream. When scrap bucket detection is performed for 2 consecutive seconds and 3 frames per second, and the proportion of scrap steel buckets detected is greater than 80%, it is determined that the feeding event has started, and the duty cycle is calculated until the feeding posture is detected and the feeding event begins. When scrap steel bucket detection is performed for 2 consecutive seconds and 3 frames per second, and the frequency of scrap steel buckets detected is less than 10%, the feeding event ends, and the cycle waits for the next event to start.

[0118] The foregoing is merely a list of specific embodiments of the present application, intended to enable those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but rather is intended to conform to the broadest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for quickly identifying the posture of scrap bucket entering the furnace based on the preset anchor of the Yolov11-seg model, characterized in that: The steps include: S1: Camera position determination and field of view adjustment; S2: Transmit the video stream collected by the camera to the NVR for video storage, and then generate the required data set; S3: Use the Yolov11-seg detection model to perform real-time scrap bucket detection on the camera; S4: Calculate the duty cycle using the mask, detection box, and preset anchor; S5: Event recognition, determining whether the scrap steel bucket in front of the furnace is preparing to load, loading, or has completed loading and is being removed from the camera's field of view.

2. The method for quickly identifying the scrap bucket entering the furnace posture based on the preset anchor of the Yolov11-seg model according to claim 1 is characterized by: The camera position determination and field of view adjustment include the following processes: S11: The camera position is determined. The position must ensure that the converter port and scrap bucket are within the camera's field of view. S12: Adjust the field of view, adjust the camera focal length or the camera up and down angle until the camera can capture 2 / 3 of the scrap bucket when the scrap bucket is loaded.

3. The method for quickly identifying the scrap bucket entering the furnace posture based on the preset anchor of the Yolov11-seg model according to claim 1 is characterized by: The process of transmitting the video stream collected by the camera to the NVR for video storage and then generating the required data set also includes the following steps: S21: Convert the original video into images, and select valid images containing scrap steel buckets that have certain differences; S22: Label the scrap bucket image and create the detection data set required for the segmentation model; S23: Perform secondary screening on the collected scrap steel bucket images to ensure that the training set contains images without furnace light interference, with strong light interference, and with weak light interference.

4. The method for quickly identifying the posture of scrap bucket entering the furnace based on the preset anchor of the Yolov11-seg model according to claim 1 is characterized in that: The Yolov11-seg detection model is used to perform real-time scrap bucket detection and segmentation on the camera, which includes the following processes: S31 pre-training parameters were selected, and the Yolov11-seg model was trained using gradient descent, with the SGD optimizer being the iterative optimizer. The model input image size was img = (640, 640, 3); S32: Model training. By observing the training loss and validation set loss, several sets of better models are determined. The stability of the model is then verified on the test set to determine the final model. If the effect does not meet the expectations, the training parameters need to be changed and training continues until the expected effect is achieved. If the number of training times cannot meet the pre-fetch requirement, then increase the training set, adjust the parameters, and iterate the training until the segmentation requirement is finally met. S33: In the model inference stage, the segmentation model outputs the scrap bucket's boding box and segmentation mask. The scrap bucket is then cropped from the 2K original image using the box, and the non-scrap bucket area is replaced with 0 pixels using the mask value. The mask calculation formula is as follows: Where: q ij Confidence in model predictions.

5. The method for quickly identifying the posture of scrap bucket entering the furnace based on the preset anchor of the Yolov11-seg model according to claim 1 is characterized in that: The steps for calculating the duty cycle using the mask, detection frame, and preset anchor are as follows: S41: Preset anchors, with the lower left corner and upper right corner of the binding box as the anchor zero points, and lay out two anchors to verify the proportion of 0 pixels in the area; x1, y1, x2, y2 are the width of the upper left vertex, the height of the upper left vertex, the width of the lower right vertex, and the height of the lower right vertex of the binding box respectively; w=x2-x1 h=y2-y1 The width and height of the scrap bucket detection box can be calculated based on x1, y1, x2, and y2, where w represents width and h represents height. Anchor width w a and high h a The calculation formula is defined as follows: In a =0.3w <h2 style=";text-align:left;direction:ltr">h<h2 style=";text-align:left;direction:ltr"> a <h2 style=";text-align:left;direction:ltr"> =0.2h After determining the width and height of the anchor, the coordinate mapping of the four anchors can be calculated. For example, the following formula is the real coordinate mapping calculation of one of the anchors in the lower left corner. in, and The coordinates of the upper left vertex of the preset anchor, and The coordinates of the lower right vertex of the preset anchor; S42: Calculate the percentage of pixels with a lower left corner pixel value of 0 based on the binding box and the two preset anchors in the lower left corner, as shown in the following formula: Among them, Z l1 is the total number of 0 elements, W l1 and H l1 The width and height of the anchor, respectively, l1 It is the ratio of the number of zero pixels of an anchor in the lower left corner to the pixel format of the entire image; S43: Calculate the proportion of 0 pixels in the upper right corner based on the binding box and the two preset anchors in the upper right corner; Among them, Z rl is the total number of 0 elements, W rl and H rl The width and height of the anchor, respectively, rl The ratio of the number of zero pixels of an anchor in the upper right corner to the pixel format of the entire image; S44: Calculate the final scrap bucket feeding posture based on the proportion of 0 pixels in the lower left corner and the upper right corner. If the following formula is satisfied, the scrap bucket posture is the feeding posture; max(ratio l1 ,ratio l2 )≤0.3 max(ratio r1 ,ratio r2 )≤0.5 6. The method for quickly identifying the posture of scrap bucket entering the furnace based on the preset anchor of the Yolov11-seg model according to claim 1, characterized in that: The event recognition determines whether the scrap bucket is preparing to load, loading, or removing from the camera's field of view after loading. The steps are as follows: S51: The scrap bucket model based on Yolov11-seg will detect the scrap bucket at a frequency of 3 frames per second for 2 consecutive seconds. When the effective frequency of the scrap bucket is greater than 80%, that is, r ≥ 0.8, it is defined as the start of the event and the scrap bucket begins to load. The formula is as follows: f = bool(Model(x)) Where x is the current input image, Model is the neural network model, F is the sum of 6 detection results, and r is the effective frequency; S52: After the event starts, the duty cycle calculation according to step S4 will be continuously performed until it is detected that the duty cycle meets the feeding conditions. This is defined as a feeding event, and the scrap steel bucket starts to feed into the furnace; S53: When no feeding event is detected within 2 consecutive seconds, the scrap bucket detection frequency will be executed for 2 consecutive seconds and 3 frames per second until the frequency of detecting the scrap bucket accounts for less than 10%. Then the task event ends and there is no scrap bucket in the field of view; S54: Loop execution: Detection is performed at a scrap bucket detection frequency of 3 frames per second for 2 consecutive seconds until the next event starts.

Images