Full-automatic unpacking detection method and system for waste steel packing blocks based on machine vision

Abstract

The invention belongs to the technical field of scrap steel automatic grade judgment, and provides a scrap steel packing block full-automatic unpacking detection method and system based on machine vision, and the system comprises an unpacking detection model and a PLC auxiliary recognition module; constructing an unpacking detection model; the PLC auxiliary identification module monitors a PLC signal, obtains a video stream, and judges and records the state of a packing block through an unpacking detection model; the claw state of the bale breaker is tracked, and the bale breaking behavior is judged and recorded; and tracking the claw state of the bale breaker, judging the bale breaking state, and stopping video stream interception. Based on machine vision and the PLC, the state of the bale breaker is detected in a high-precision and full-process mode, the problem of full-automatic process vacancy in the waste steel packing block bale breaking detection process is solved, and therefore the detection method for full-automatic waste steel packing block bale breaking operation is achieved, all application scenes can be covered, and the detection efficiency is improved. And the influence on the subsequent task process is avoided.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of automatic scrap steel grading, and in particular relates to a full-automatic unpacking detection method and system for scrap steel bales based on machine vision. BACKGROUND

[0002] In the early years, the automatic scrap steel grading technology mainly used machine learning models, and in recent years, some enterprises have realized the fusion application of machine vision and big data. At present, there are still problems such as scene adaptability defects, the inability of bulk material recognition technology to handle the internal structure of baled blocks (such as closed container detection), and the identification accuracy fluctuation reaching plus or minus 15% under complex lighting conditions.

[0003] In the prior art, there is a significant technical fault line that bulk material detection is mature, but baled block detection is weak. Patent document “Automatic recognition method for scrap steel unloading change area based on deep learning” (CN114049543A) discloses a method based on deep learning, which uses Mask R-CNN algorithm to construct a scrap steel carriage positioning model, combines YOLO-v4 algorithm to construct a grabber tracking model, and uses Gaussian mixture model for background modeling to realize automatic recognition and image saving of the scrap steel carriage change area. However, its detection method is only for bulk materials, which is also the method used by most intelligent detection systems, and is completely unsuitable for baled block scenes. Similarly, patent document “Scrap steel grading method, device and system” (CN119941670A) realizes the function of automatic grading of scrap steel associated with license plates, but its image processing module is still limited to the surface feature recognition of bulk scrap steel. Patent document “Automatic detection and evaluation method and system for scrap steel” (CN119991601A) introduces fuzzy boundary analysis technology, but its defect evaluation system cannot penetrate the internal structure of baled blocks to detect hidden dangers such as closed containers.

[0004] The current mainstream technical solution has three major technical barriers. First, the bulk material recognition technology relies on a two-dimensional image analysis model constructed by Mask R-CNN algorithm, and the recognition accuracy of the three-dimensional overlapping structure of baled blocks is less than 65%. Second, the existing sorting equipment can only complete the primary screening of surface impurities on the conveying line, and lacks a deep composition analysis module after unpacking. Third, the density detection method can indirectly evaluate the material composition, but cannot locate the specific spatial distribution of dangerous objects inside the baled blocks.

[0005] For example, patent document “Scrap steel baling and sorting identification equipment and sorting identification method” (CN119346443A) discloses a combination of an identification system and a sorting device to realize automatic sorting of closed containers in scrap steel materials, and uses an edge computing terminal and an image acquisition device to identify the types of scrap steel briquettes and manage information. Although it involves baled block processing, it focuses on the sorting link rather than the unpacking link.

[0006] This technological gap severely restricts the automation upgrade process of scrap steel processing and distribution centers. Therefore, there is a need for a method that can accurately detect the status of the unpacking machine, solve the problem of the lack of a fully automated process in the unpacking and detection of scrap steel bales, and thus realize a fully automated unpacking and detection method for scrap steel bales. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this invention is to provide a fully automated unpacking and inspection method and system for scrap steel bales based on machine vision.

[0008] The present invention provides a fully automated unpacking and inspection method for scrap steel bales based on machine vision, comprising:

[0009] Step S1: Construct a packet unpacking detection model;

[0010] Step S2: Listen to the PLC signal, acquire the video stream, determine the status of the packaged block through the unpacking detection model, and record it;

[0011] Step S3: Track the status of the unpacking machine's gripper, determine the unpacking behavior, and record it;

[0012] Step S4: Track the status of the unpacking machine's gripper, determine the unpacking status, and stop capturing the video stream.

[0013] Repeat steps S2 to S4 until unpacking is complete.

[0014] Preferably, step S1 includes:

[0015] Step S1.1: Collect sample photos and construct a training set;

[0016] Step S1.2: Preprocess the training set data and train the target detection model to obtain the packet splitting detection model.

[0017] The sample photos were extracted from frames captured by the camera at the unpacking machine station during the on-site unpacking operation. The photos included images of the unpacking machine's claw, the unpacking table, unopened packing blocks, and unopened packing blocks.

[0018] The preprocessing includes data augmentation, resolution unification, and histogram equalization.

[0019] Preferably, step S1.1 includes:

[0020] Step S1.1.1: Take video recordings of the unpacking operation taken by the camera at different time periods at the unpacking operation site;

[0021] Step S1.1.2: Extract and filter frames from the unpacking operation video to obtain usable unpacking operation images;

[0022] Step S1.1.3: Mark the target detection boxes for the unpacking machine claw, unpacking table, unopened packing blocks and opened packing blocks in the unpacking operation image to obtain the training set.

[0023] The images of the unpacking operation are frame-by-frame images of unopened and opened packing blocks under various lighting conditions and in various positions of the packing blocks.

[0024] Preferably, step S2 includes:

[0025] Step S2.1: Listen to the PLC signal. When the unpacking machine's gripper is in running state for a continuously set time, send a start flag message to the unpacking machine's workstation camera.

[0026] Step S2.2: The camera at the unpacking machine station receives the start marker information and immediately acquires the real-time video stream from the site.

[0027] Step S2.3: Determine whether there are unopened packaged blocks on the unpacking platform using the unpacking detection model, and track and record the status of the packaged blocks in real time;

[0028] If there are unopened packing blocks, the PLC returns a closing signal to the unpacking machine gripper and executes step S3; if there are no unopened packing blocks, then step S2.1 is executed.

[0029] Preferably, in step S3, the status of the unpacking machine's gripper is tracked in real time. If the position of the gripper reaches a preset threshold, it is determined that unpacking behavior has occurred, the unpacking behavior flag is recorded, and step S4 is executed. If the position of the gripper does not reach the preset threshold, step S2.2 is executed.

[0030] In step S4, the unpacking machine claw is tracked in real time to determine its position. If it is not far from the set threshold, step S2.2 is executed. If it is far from the preset threshold, the PLC signal is continuously monitored to determine the unpacking machine's operating status.

[0031] If the unpacking machine remains running, proceed to step S2.2. If the unpacking machine remains stationary for the set time, the unpacking is considered complete, and the camera at the unpacking machine's workstation stops acquiring real-time video streams.

[0032] The present invention provides a fully automated unpacking and detection system for scrap steel bales based on machine vision, which specifically includes: a PLC-assisted recognition module and an unpacking and detection model.

[0033] Construct a package unpacking detection model.

[0034] The PLC-assisted identification module listens to PLC signals, acquires video streams, and determines and records the status of packaged blocks through the unpacking detection model.

[0035] Track the status of the unpacking machine's gripper, determine the unpacking behavior, and record it;

[0036] Track the status of the unpacking machine's gripper, determine the unpacking status, and stop capturing the video stream.

[0037] Preferably, the construction of the unpacking detection model includes:

[0038] Module M1.1: Collect sample photos and construct the training set;

[0039] Module M1.2 preprocesses the training set data and trains the target detection model to obtain the packet splitting detection model.

[0040] The sample photos were extracted from frames captured by the camera at the unpacking machine station during the on-site unpacking operation. The photos included images of the unpacking machine's claw, the unpacking table, unopened packing blocks, and unopened packing blocks.

[0041] The preprocessing includes data augmentation, resolution unification, and histogram equalization.

[0042] Preferably, module M1.1 includes:

[0043] Module M1.1.1: Captures unpacking operation videos taken by cameras at different time periods at the unpacking operation site;

[0044] Module M1.1.2: Extract frames and filter unpacking operation videos to obtain usable unpacking operation images;

[0045] Module M1.1.3 performs target detection box annotation on the unpacking machine claw, unpacking table, unopened packing blocks and opened packing blocks in the unpacking operation image to obtain the training set.

[0046] The images of the unpacking operation are frame-by-frame images of unopened and opened packing blocks under various lighting conditions and in various positions of the packing blocks.

[0047] Preferably, the PLC auxiliary identification module continuously monitors the PLC signal, and when the unpacking machine claw is in operation for a set time, it sends a start flag message to the unpacking machine station camera.

[0048] Once the camera at the unpacking machine station receives the start signal, it immediately acquires a real-time video stream from the site.

[0049] The unpacking detection model determines whether there are unopened packing blocks on the unpacking platform, and tracks and records the status of the packing blocks in real time.

[0050] If there are unopened packing blocks, the PLC returns a closing signal to the unpacking machine's gripper and continues to track the status of the gripper. If there are no unopened packing blocks, the PLC continues to monitor the signals.

[0051] Preferably, the PLC-assisted identification module tracks the status of the unpacking machine's gripper in real time. If there is a gripper closure signal and the gripper position reaches a preset threshold, it is determined that unpacking behavior has occurred, the unpacking behavior flag is recorded, and the gripper position is tracked again. If the gripper position does not reach the preset threshold or there is no gripper closure signal, the video stream is reacquired.

[0052] The position of the unpacking machine's gripper is determined. If it is not far from the set threshold, the video stream is reacquired. If it is far from the preset threshold, the PLC signal is continuously monitored to determine the unpacking machine's operating status.

[0053] If the unpacking machine remains running, it will reacquire the video stream. If the unpacking machine remains stationary for the set time, it will be determined that the unpacking is complete, and the camera at the unpacking machine's workstation will stop acquiring the real-time video stream.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] 1. This invention solves the problem of the lack of fully automated process in the unpacking and inspection of scrap steel bales, thereby realizing a fully automated unpacking and inspection method for scrap steel bales.

[0056] 2. This invention fully integrates visual recognition technology with PLC signals, using PLC to assist in visual recognition, thereby achieving higher accuracy in automatic unpacking and detection.

[0057] 3. Based on machine vision and programmable logic controller (PLC), this invention provides high-precision and accurate full-process detection of the unpacking machine status, covering all application scenarios and avoiding impact on subsequent task progress. Attached Figure Description

[0058] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0059] Figure 1 This is a schematic diagram of a fully automated unpacking and inspection method for scrap steel bales. Detailed Implementation

[0060] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0061] The present invention provides a fully automated unpacking and inspection method for scrap steel bales based on machine vision, which... Figure 1 For example, the specific process includes:

[0062] Step S1: Collect sample photos of the unpacking machine's claw, unpacking table, unopened packing blocks, and opened packing blocks to construct a training set.

[0063] The sample photos were obtained by taking daily video recordings of the unpacking operations at the unpacking machine station using a camera near the machine. Frames were extracted from the video recordings and labeled to obtain the training set.

[0064] Specifically, step S1 includes the following steps:

[0065] Step S1.1: Collect video recordings of the unpacking operation at different time periods.

[0066] Step S1.2: Extract frames from the video recording obtained in step S1.1 to obtain usable unpacking operation images.

[0067] The images used for the unpacking operation should cover as many scenes as possible, specifically, various lighting environments and different locations of the packaged blocks, including both unpacked and unpacked blocks. A rich collection of sample images can significantly improve the generalization ability of the model after training, while also increasing the model's accuracy.

[0068] Step S1.3: Mark the target detection boxes for the unpacking machine claw, unpacking table, unopened package blocks and opened package blocks in the image obtained in step S1.2.

[0069] Step S2: Train the unpacking detection model on the training set. This will result in a model capable of detecting unpacking machine claws, unpacking tables, unopened packages, and opened packages.

[0070] Step S2 includes the following steps:

[0071] Step S2.1: Perform data augmentation, resolution unification, and histogram equalization on the training set from step S1.

[0072] Data augmentation can enhance data diversity, improve model generalization ability, and prevent overfitting. Through geometric transformations (rotation, flipping) and color adjustments (brightness, contrast), it generates diverse samples, alleviating the problem of insufficient data. It forces the model to learn more discriminative features rather than relying on specific scene attributes (such as fixed lighting or object positions), thereby adapting to changes in the testing environment, reducing the model's over-reliance on training samples, and lowering the risk of overfitting due to small dataset size or single sample.

[0073] Histogram equalization can enhance image contrast, improve detail visibility and lighting adaptability, adjust the image grayscale distribution, and make it more robust under different lighting conditions, especially suitable for low-contrast scenes.

[0074] Standardizing the input resolution can standardize the input format, ensure input consistency, and improve computational efficiency. Standardizing the image size can avoid performance fluctuations or structural adaptation problems caused by differences in input resolution. Fixed input resolution simplifies the batch processing process and optimizes memory and computing resource allocation.

[0075] Step S2.2: Train an object detection model using the training set data processed in step S2.1.

[0076] For unpacking blocks of different sizes, the distance the unpacking machine's claw moves closer is not consistent when manually controlled. It is difficult to cover all scenarios and accurately determine the unpacking status using only visual recognition technology. Therefore, it is necessary to use PLC to assist visual recognition in order to achieve extremely high recognition accuracy.

[0077] Step S3: Continuously monitor the PLC signal. When the unpacking machine's gripper moves, notify the system to start acquiring real-time video streams from the on-site cameras. Use the model from Step S2 to determine if there are any unopened packing blocks on the unpacking table, track the status of these packing blocks in real time, and record it.

[0078] Step S3 includes the following steps:

[0079] Step S3.1: Listen to the PLC signal and when the unpacking machine's claw is in a continuous running state for a set period of time, notify it to start acquiring the video stream.

[0080] Step S3.2: After receiving the flag information that started in step S3.1, the camera at the unpacking machine station immediately begins to acquire the real-time video stream from the on-site camera.

[0081] Step S3.3: Determine whether there are unopened packing blocks on the unpacking platform using the model from step S2, and track and record the status of the packing block in real time.

[0082] If there are unopened packing blocks, the PLC returns a closing signal to the unpacking machine gripper; if there are no unopened packing blocks, step S4 is executed.

[0083] By fully integrating visual recognition technology with PLC signals, and using PLC-assisted visual recognition, higher accuracy in automatic unpacking and detection can be achieved.

[0084] Step S4: Simultaneously with step S3, the status of the unpacking machine's gripper is tracked in real time. When the position of the unpacking machine's gripper continues to approach the preset threshold and the PLC returns that the unpacking machine's gripper is closed, it is determined that an unpacking behavior has occurred, and the unpacking behavior flag is recorded.

[0085] Step S5: Continue to track the unpacking machine's gripper in real time. When the gripper moves away from the preset threshold again, continue to listen to the PLC signal. When the unpacking machine stops for a period of time, it is determined that the unpacking is complete and the acquisition of the real-time video stream from the on-site camera is stopped.

[0086] Since unpacking detection requires detecting a series of states from the packing block being placed on the packing table, the start of unpacking, the end of unpacking, the packing block leaving the packing block and being placed on the packing table of the next packing block, real-time feedback from PLC signals minimizes the misjudgment rate of unpacking detection models that rely solely on visual recognition when determining unpacking actions, improves detection accuracy, and avoids the highly probable impact on subsequent states that would result from misjudgment.

[0087] Repeat steps S3, S4, and S5 until all unpacking tasks are completed and the unpacking machine stops.

[0088] The present invention also provides a fully automated unpacking and detection system for scrap steel baled blocks based on machine vision. The fully automated unpacking and detection system for scrap steel baled blocks based on machine vision can be implemented by executing the process steps of the fully automated unpacking and detection method for scrap steel baled blocks based on machine vision. That is, those skilled in the art can understand the fully automated unpacking and detection method for scrap steel baled blocks based on machine vision as a preferred embodiment of the fully automated unpacking and detection system for scrap steel baled blocks based on machine vision.

[0089] The present invention provides a fully automated unpacking and detection system for scrap steel bales based on machine vision, which specifically includes: a PLC-assisted recognition module and an unpacking and detection model.

[0090] Construct a package unpacking detection model.

[0091] The PLC-assisted identification module listens to PLC signals, acquires video streams, and determines and records the status of packaged blocks through the unpacking detection model.

[0092] Track the status of the unpacking machine's gripper, determine the unpacking behavior, and record it;

[0093] Track the status of the unpacking machine's gripper, determine the unpacking status, and stop capturing the video stream.

[0094] In more preferred embodiments, the construction of the unpacking detection model includes:

[0095] Module M1.1: Collect sample photos and construct the training set;

[0096] Module M1.2 preprocesses the training set data and trains the target detection model to obtain the packet splitting detection model.

[0097] The sample photos were extracted from frames captured by the camera at the unpacking machine station during the on-site unpacking operation. The photos included images of the unpacking machine's claw, the unpacking table, unopened packing blocks, and unopened packing blocks.

[0098] The preprocessing includes data augmentation, resolution unification, and histogram equalization.

[0099] In more preferred embodiments, module M1.1 includes:

[0100] Module M1.1.1: Captures unpacking operation videos taken by cameras at different time periods at the unpacking operation site;

[0101] Module M1.1.2: Extract frames and filter unpacking operation videos to obtain usable unpacking operation images;

[0102] Module M1.1.3 performs target detection box annotation on the unpacking machine claw, unpacking table, unopened packing blocks and opened packing blocks in the unpacking operation image to obtain the training set.

[0103] The images of the unpacking operation are frame-by-frame images of unopened and opened packing blocks under various lighting conditions and in various positions of the packing blocks.

[0104] In more preferred embodiments, the PLC auxiliary identification module continuously monitors the PLC signal and sends a start flag message to the unpacking machine station camera when the unpacking machine claw is in operation for a set time.

[0105] Once the camera at the unpacking machine station receives the start signal, it immediately acquires a real-time video stream from the site.

[0106] The unpacking detection model determines whether there are unopened packing blocks on the unpacking platform, and tracks and records the status of the packing blocks in real time.

[0107] If there are unopened packing blocks, the PLC returns a closing signal to the unpacking machine's gripper and continues to track the status of the gripper. If there are no unopened packing blocks, the PLC continues to monitor the signals.

[0108] In more preferred embodiments, the PLC-assisted identification module tracks the status of the unpacking machine's gripper in real time. If there is a gripper closure signal and the gripper position reaches a preset threshold, it is determined that unpacking behavior has occurred, the unpacking behavior flag is recorded, and the gripper position is tracked again. If the gripper position does not reach the preset threshold or there is no gripper closure signal, the video stream is reacquired.

[0109] The position of the unpacking machine's gripper is determined. If it is not far from the set threshold, the video stream is reacquired. If it is far from the preset threshold, the PLC signal is continuously monitored to determine the unpacking machine's operating status.

[0110] If the unpacking machine remains running, it will reacquire the video stream. If the unpacking machine remains stationary for the set time, it will be determined that the unpacking is complete, and the camera at the unpacking machine's workstation will stop acquiring the real-time video stream.

[0111] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0112] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A fully automated unpacking and inspection method for scrap steel bales based on machine vision, characterized in that, include: Step S1: Construct a packet unpacking detection model; Step S2: Listen to the PLC signal, acquire the video stream, determine the status of the packaged block through the unpacking detection model, and record it; Step S3: Track the status of the unpacking machine's gripper, determine the unpacking behavior, and record it; Step S4: Track the status of the unpacking machine's gripper, determine the unpacking status, and stop capturing the video stream; Repeat steps S2 to S4 until unpacking is complete.

2. The fully automated unpacking and inspection method for scrap steel bales based on machine vision according to claim 1, characterized in that, Step S1 includes: Step S1.1: Collect sample photos and construct a training set; Step S1.2: Preprocess the training set data and train the target detection model to obtain the packet splitting detection model; The sample photos were obtained by extracting frames from the video stream of the unpacking operation captured by the camera at the unpacking machine station, including photos of the unpacking machine claw, the unpacking table, unopened packing blocks, and opened packing blocks; The preprocessing includes data augmentation, resolution unification, and histogram equalization.

3. The fully automated unpacking and inspection method for scrap steel bales based on machine vision according to claim 2, characterized in that, Step S1.1 includes: Step S1.1.1: Take video recordings of the unpacking operation taken by the camera at different time periods at the unpacking operation site; Step S1.1.2: Extract and filter frames from the unpacking operation video to obtain usable unpacking operation images; Step S1.1.3: Mark the target detection boxes for the unpacking machine claw, unpacking table, unopened packing blocks and opened packing blocks in the unpacking operation image to obtain the training set; The images of the unpacking operation are frame-by-frame images of unopened and opened packing blocks under various lighting conditions and in various positions of the packing blocks.

4. The fully automated unpacking and inspection method for scrap steel bales based on machine vision according to claim 1, characterized in that, Step S2 includes: Step S2.1: Listen to the PLC signal. When the unpacking machine's gripper is in running state for a continuously set time, send a start flag message to the unpacking machine's workstation camera. Step S2.2: Upon receiving the start signal information, the camera at the unpacking machine station immediately acquires the real-time video stream from the site. Step S2.3: Determine whether there are unopened packaged blocks on the unpacking platform using the unpacking detection model, and track and record the status of the packaged blocks in real time; If there are unopened packing blocks, the PLC returns a closing signal to the unpacking machine gripper and executes step S3; if there are no unopened packing blocks, then step S2.1 is executed.

5. The fully automated unpacking and inspection method for scrap steel bales based on machine vision according to claim 4, characterized in that, In step S3, the status of the unpacking machine's gripper is tracked in real time. If the position of the unpacking machine's gripper reaches a preset threshold, it is determined that unpacking behavior has occurred, the unpacking behavior flag is recorded, and step S4 is executed. If the position of the unpacking machine's gripper does not reach the preset threshold, step S2.2 is executed. In step S4, the unpacking machine claw is tracked in real time to determine its position. If it is not far from the set threshold, step S2.2 is executed. If it is far from the preset threshold, the PLC signal is continuously monitored to determine the unpacking machine's operating status. If the unpacking machine remains running, proceed to step S2.

2. If the unpacking machine remains stationary for the set time, the unpacking is considered complete, and the camera at the unpacking machine's workstation stops acquiring real-time video streams.

6. A fully automated unpacking and inspection system for scrap steel bales based on machine vision, characterized in that, include: Unpacking detection model and PLC-assisted identification module; Construct an unpacking detection model; The PLC-assisted identification module listens to PLC signals, acquires video streams, and determines and records the status of packaged blocks through the unpacking detection model. Track the status of the unpacking machine's gripper, determine the unpacking behavior, and record it; Track the status of the unpacking machine's gripper, determine the unpacking status, and stop capturing the video stream.

7. The fully automated unpacking and inspection system for scrap steel bales based on machine vision according to claim 6, characterized in that, The construction of the packet unpacking detection model includes: Module M1.1: Collect sample photos and construct the training set; Module M1.2: Preprocess the training set data and train the target detection model to obtain the packet splitting detection model; The sample photos were obtained by extracting frames from the video stream of the unpacking operation captured by the camera at the unpacking machine station, including photos of the unpacking machine claw, the unpacking table, unopened packing blocks, and opened packing blocks; The preprocessing includes data augmentation, resolution unification, and histogram equalization.

8. The fully automated unpacking and inspection system for scrap steel bales based on machine vision according to claim 7, characterized in that, The module M1.1 includes: Module M1.1.1: Captures unpacking operation videos taken by cameras at different time periods at the unpacking operation site; Module M1.1.2: Extract frames and filter unpacking operation videos to obtain usable unpacking operation images; Module M1.1.3: Mark the target detection boxes for the unpacking machine claw, unpacking table, unopened packing blocks and opened packing blocks in the unpacking operation images to obtain the training set; The images of the unpacking operation are frame-by-frame images of unopened and opened packing blocks under various lighting conditions and in various positions of the packing blocks.

9. The fully automated unpacking and inspection system for scrap steel bales based on machine vision according to claim 6, characterized in that, The PLC auxiliary identification module continuously monitors the PLC signal. When the unpacking machine claw is in running state for a set time, it sends a start flag message to the unpacking machine station camera. Once the unpacking machine's workstation camera receives the start signal, it immediately acquires a real-time video stream of the scene. The unpacking detection model determines whether there are unopened packaged blocks on the unpacking platform, and tracks and records the status of the packaged blocks in real time. If there are unopened packing blocks, the PLC returns a closing signal to the unpacking machine's gripper and continues to track the status of the gripper. If there are no unopened packing blocks, the PLC continues to monitor the signals.

10. The fully automated unpacking and inspection system for scrap steel bales based on machine vision according to claim 9, characterized in that, The PLC-assisted identification module tracks the status of the unpacking machine's gripper in real time. If there is a gripper closure signal and the gripper position reaches a preset threshold, it determines that unpacking has occurred, records the unpacking behavior flag, and continues to track the gripper position. If the gripper position does not reach the preset threshold or there is no gripper closure signal, it reacquires the video stream. The position of the unpacking machine's gripper is determined. If it is not far from the set threshold, the video stream is reacquired. If it is far from the preset threshold, the PLC signal is continuously monitored to determine the unpacking machine's operating status. If the unpacking machine remains running, it will reacquire the video stream. If the unpacking machine remains stationary for the set time, it will be determined that the unpacking is complete, and the camera at the unpacking machine's workstation will stop acquiring the real-time video stream.

Images