Intelligent detection system for scrap steel processing and sorting

Abstract

The invention discloses an intelligent detection system for scrap steel processing and sorting, and relates to the field of metal processing, according to the intelligent detection system for scrap steel processing and sorting, stored scrap steel is rapidly sorted through an AI visual system, the labor cost is reduced, a visual identification system constructed by adopting a YOLOv5 target detection algorithm is adopted to compare and sort materials through a database, and the detection accuracy is improved. Materials with low recognition degree are subjected to attached drawing and artificial auxiliary judgment, the problem that an existing system training set is insufficient in precision is avoided, the judgment precision of the system is correspondingly improved, an image fast indexing and embedding data set technology is constructed, and a metadata set is subjected to percentage comparison recognition for grade judgment. The problem that AI recognition precision is not enough is solved through an uploading manual recognition mode, rapid recognition of the materials through an index system is achieved, the recognition precision and the learning ability of the system are extended in combination with a manual mode, and the problem that an existing system is low in recognition precision is avoided.

Description

Technical Field

[0001] This invention relates to a metal processing inspection system, specifically to a smart inspection system for scrap steel processing and sorting. Background Technology

[0002] Reprocessing scrap steel is a major source of metal materials. Through the in-depth application of technologies such as AI visual recognition, laser positioning, and intelligent algorithms, a leapfrog upgrade from manual quality inspection to intelligent grading has been achieved, greatly improving production line efficiency.

[0003] However, existing intelligent systems still have many problems. First, the cost of precision systems is high, and the maintenance of high-end equipment is complex, resulting in high operating costs, operational difficulty, and maintenance costs. Second, the basic training volume of the system is insufficient, and the existing database is difficult to cover all scrap steel types, which limits the generalization ability of the model.

[0004] Therefore, those skilled in the art propose an intelligent detection system for scrap steel processing and sorting. By combining an intelligent system with manual methods, and considering the obvious deficiencies of both manual and intelligent sorting systems, the system learns in a way that coordinates manual processes, ensuring both recognition accuracy and sorting efficiency. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent detection system for scrap steel processing and sorting. By combining an intelligent system with manual sorting, it not only replaces manual labor in performing tedious sorting tasks, thereby increasing work efficiency and reducing labor costs, but also compensates for the limited datasets of existing systems that cannot cover all material types by combining them with manual sorting systems. For materials with low identification accuracy, manual identification is used as an adjunct, achieving a synergy between efficiency and cost.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a smart detection system for scrap steel processing and sorting, comprising the following identification and sorting steps:

[0007] S1. Scan the vehicles, classify them according to the criteria, and input the information and allocate the unloading area;

[0008] S2. Multi-angle image capture of the unloading area is used to remove impurities through AI visual recognition.

[0009] S3. Images with low recognition accuracy are uploaded manually, while simultaneously being sorted by the corresponding equipment;

[0010] S4. Perform secondary identification on sorted materials with low identification accuracy, integrate and finally judge impurities, and give the corresponding grade.

[0011] Preferably, in step S1, in addition to the camera scanning information, manual control is also incorporated, and the specified passage path is indicated by the manually input classification.

[0012] Preferably, in step S2, the visual recognition system is constructed using the YOLOv5 target detection algorithm: an image contour index dataset is constructed, material photos are imported into the mosaic dataset for comparison, and the materials on the conveyor belt are classified into three levels—compliant, suspected, and non-compliant—based on the comparison values ​​and quickly classified and transported. Non-compliant materials are directly weighed and screened out, while suspected materials enter the secondary recognition channel.

[0013] Preferably, in step S3, the suspected materials screened in S2 are conveyed and identified. A camera is used to take a picture, which is then uploaded to the corresponding suspected items for manual identification. The manual identification data is uploaded to the system, and a robotic arm is used to sort the materials on the conveyor belt again to determine the impurity removal data.

[0014] Preferably, in step S4, the total impurity removal data is calculated, compared with the reference data, the system grade is automatically given, and the data is saved in the corresponding file.

[0015] Preferably, the visual recognition system includes an image preprocessing module, which performs sample extraction functions such as importing and extracting images, random sampling, and on-image sampling, as well as temporary database operation functions such as printing temporary layers and deleting temporary layers.

[0016] "Import Extraction" is a function used to extract images from quality inspection images according to an Excel list. Users need to prepare an Excel data table containing two fields: serial number and sample name. The sample name column is filled with the image number to be used for quality inspection extraction.

[0017] The "Random Sampling" function allows users to randomly select a fixed number of images for inspection. Users can enter a fixed value, which cannot exceed the total number of images. This function will evenly select departmental samples from the image folder for random inspection.

[0018] "Image Extraction" is a function that allows users to extract images through interactive operations. Users can manually select the extraction range on the monitor or randomly select it on the computer, and then extract the image data within the range after confirming the selection.

[0019] Preferably, the visual recognition system further includes an image viewing module, used to quickly perform batch image comparisons based on a database, and to sort the comparison data using a robotic arm.

[0020] The comparison data is expressed as a percentage and is divided into three levels according to the data size, from low to high: 0-35%, 36-65%, and 66-100%. Materials with a recognition accuracy of 66-100% pass directly through screening, materials with a recognition accuracy of 36-65% are swept into another conveyor line, and materials with a recognition accuracy of 0-35% are screened out for impurity removal.

[0021] The suspected materials that were sorted out were photographed again and uploaded for manual processing. The images were then attached to the suspected materials and uploaded for manual processing. The manual judgment data was uploaded, and the conveyed materials were sorted again based on the judgment.

[0022] This invention discloses an intelligent detection system for scrap steel processing and sorting, which has the following beneficial effects:

[0023] 1. This intelligent detection system for scrap steel processing and sorting uses an AI vision system to quickly sort the collected scrap steel, reducing labor costs. The visual recognition system, built using the YOLOv5 object detection algorithm, compares the sorted materials with the database. For materials with low recognition accuracy, it adds attached images and provides manual assistance for identification, avoiding the problem of insufficient accuracy of the training set in existing systems and thus improving the system's discrimination accuracy.

[0024] 2. This intelligent detection system for scrap steel processing and sorting replaces the traditional manual sorting and grading method with a sorting method that combines AI recognition with manual processing, thereby reducing costs. In addition, the AI ​​recognition system is equipped with a grading algorithm, which uploads materials with low recognition accuracy and makes judgments with human assistance. This not only improves the sorting accuracy of the system, but also allows for continuous training through human feedback, thereby enhancing the system's sorting capabilities. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of the system workflow of the present invention;

[0027] Figure 2 Import the extracted Excel list image for this invention;

[0028] Figure 3 This is a screenshot of the extraction tool page for this invention;

[0029] Figure 4 This invention randomly selects pages;

[0030] Figure 5 This is a screenshot of the image comparison sample output page of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] This invention discloses an intelligent detection system for scrap steel processing and sorting;

[0033] According to the appendix Figure 1 As shown, the following identification and sorting steps are included:

[0034] S1. Scan the vehicles, classify them according to the criteria, and input the information and allocate the unloading area;

[0035] S2. Multi-angle image capture of the unloading area is used to remove impurities through AI visual recognition.

[0036] S3. Images with low recognition accuracy are uploaded manually, while simultaneously being sorted by the corresponding equipment;

[0037] S4. Perform secondary identification on sorted materials with low identification accuracy, integrate and finally judge impurities, and give the corresponding grade.

[0038] In step S1, in addition to the information scanned by the camera, manual control is also incorporated, and the specified passage path is indicated by the classification input by the user.

[0039] In step S2, the visual recognition system is constructed using the YOLOv5 target detection algorithm: an image contour index dataset is built, material photos are imported into the mosaic dataset for comparison, and the materials on the conveyor belt are divided into three levels: conforming, suspected, and non-conforming, based on the comparison values ​​and quickly classified and transported. Non-conforming materials are directly weighed and screened out, while suspected materials enter the secondary recognition channel.

[0040] In step S3, the suspected materials screened in S2 are conveyed and identified. A camera takes a picture, which is then uploaded to the corresponding suspected items for manual identification. The manually identified data is then uploaded to the system. A robotic arm sorts the materials on the conveyor belt again to determine the impurity removal data.

[0041] In step S4, the total cleaned data is calculated, compared with the reference data, the system grade is automatically given, and the data is saved in the corresponding file.

[0042] The visual recognition system includes an image preprocessing module, which performs sample extraction functions such as importing and extracting images, random sampling, and on-image sampling, as well as temporary database operation functions such as printing temporary layers and deleting temporary layers.

[0043] "Import Extraction" is a function used to extract images from quality inspection images according to an Excel list. Users need to prepare an Excel data table containing two fields: serial number and sample name. The sample name column is filled with the image number to be used for quality inspection extraction.

[0044] See attached document Figure 2 and attached Figure 3 After opening the "Import Extraction" tool, a tool for importing sample names and extracting images will pop up, requiring you to fill in three parameters.

[0045] ① Enter the image folder: Enter the root directory of the folder containing the images to be inspected.

[0046] ② Import sample Excel: Enter the absolute path of the xlsx file containing the serial number and image name.

[0047] ③ Output sample folder: Specify an empty folder to store the extracted samples.

[0048] The "Random Sampling" function allows users to randomly select a fixed number of images for inspection. Users can enter a fixed value, which cannot exceed the total number of images. This function will evenly select departmental samples from the image folder for random inspection.

[0049] See attached document Figure 4 As shown, after opening the "Random Selection" tool, you need to fill in the corresponding parameters:

[0050] ① Enter the image folder: Enter the root directory of the folder containing the images to be inspected.

[0051] ② Number of samples to be drawn: Select a number, or you can manually enter an integer.

[0052] ③ Output sample folder: Specify an empty folder to store the extracted samples.

[0053] "Image Extraction" is a function that allows users to extract images through interactive operations. Users can manually select the extraction range on the monitor or randomly select it on the computer, and then extract the image data within the range after confirming the selection.

[0054] See attached document Figure 5 The following parameters are required for the extraction toolbar on the image:

[0055] ① Sample extraction layer: Select a layer from the current layer list using the drop-down menu. Generally, select the outline layer of the image. If the layer is not in the drop-down menu, refresh the layer by clicking the "Refresh" button on the right.

[0056] ② Sample extraction field: Select a field in the sample layer through the drop-down menu. This field must correspond to the name of the image.

[0057] ③ Number of samples to be drawn: Specify an integer value.

[0058] ④ Select Image Folder: Specify the root directory folder for storing images.

[0059] ⑤ Sample Image Folder: Specify an empty folder to store the extracted image samples.

[0060] ⑥ Select on the map: Used to facilitate user interaction and selection. Clicking "Select on the map" allows you to select multiple contour surfaces on the map. The sequence number and image name of the selected contour surface feature will be listed in the list box below.

[0061] ⑦ Random Selection: This function is used to randomly select a certain number of samples for sample extraction by computer.

[0062] ⑧ Delete Selection: Individual image samples listed in the list box can be selected for deletion.

[0063] 9. Location on the map: Select a sample in the list box to quickly locate its position on the map.

[0064] ⑩ Export Samples: After confirming the export of samples, this function allows you to export selected samples in batches to a specified image folder.

[0065] The visual recognition system also includes an image viewing module, which is used to quickly perform batch image comparisons based on a database and sort the comparison data using a robotic arm.

[0066] The comparison data is expressed as a percentage and is divided into three levels according to the data size, from low to high: 0-35%, 36-65%, and 66-100%. Materials with a recognition accuracy of 66-100% pass directly through screening, materials with a recognition accuracy of 36-65% are swept into another conveyor line, and materials with a recognition accuracy of 0-35% are screened out for impurity removal.

[0067] The suspected materials that were sorted out were photographed again and uploaded for manual processing. The images were then attached to the suspected materials and uploaded for manual processing. The manual judgment data was uploaded, and the conveyed materials were sorted again based on the judgment.

[0068] This intelligent detection system for scrap steel processing and sorting uses an AI vision system to quickly sort the collected scrap steel, reducing labor costs. The visual recognition system, built using the YOLOv5 object detection algorithm, compares the sorted materials with a database, adds diagrams and provides manual assistance for materials with low recognition accuracy, thus avoiding the problem of insufficient accuracy in the training set of existing systems and improving the system's recognition accuracy accordingly.

[0069] The sorting method that combines AI recognition with manual processing replaces the traditional manual sorting and grading, reducing costs. In addition, the AI ​​recognition system is equipped with a grading algorithm, and materials with low recognition accuracy are uploaded for manual judgment. This not only improves the sorting accuracy of the system, but also allows for continuous training through human feedback, thereby enhancing the system's sorting capabilities.

[0070] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A smart detection system for scrap steel processing and sorting, characterized in that, The following identification and sorting steps are included: S1. Scan the vehicles, classify them according to the criteria, and input the information and allocate the unloading area; S2. Multi-angle image capture of the unloading area is used to remove impurities through AI visual recognition. S3. Images with low recognition accuracy are uploaded manually, while simultaneously being sorted by the corresponding equipment; S4. Perform secondary identification on sorted materials with low identification accuracy, integrate and finally judge impurities, and give the corresponding grade.

2. The intelligent detection system for scrap steel processing and sorting as described in claim 1, characterized in that: In step S1, in addition to the information scanned by the camera, manual control is also incorporated, and the specified passage path is indicated by the classification input by the user.

3. The intelligent detection system for scrap steel processing and sorting as described in claim 1, characterized in that: In step S2, the visual recognition system is constructed using the YOLOv5 target detection algorithm: an image contour index dataset is built, material photos are imported into the mosaic dataset for comparison, and the materials on the conveyor belt are divided into three levels: conforming, suspected, and non-conforming, based on the comparison values ​​and quickly classified and transported. Non-conforming materials are directly weighed and screened out, while suspected materials enter the secondary recognition channel.

4. The intelligent detection system for scrap steel processing and sorting as described in claim 1, characterized in that: In step S3, the suspected materials screened in S2 are conveyed and identified. A camera takes a picture, which is then uploaded to the corresponding suspected items for manual identification. The manually identified data is then uploaded to the system. A robotic arm sorts the materials on the conveyor belt again to determine the impurity removal data.

5. The intelligent detection system for scrap steel processing and sorting as described in claim 1, characterized in that: In step S4, the total cleaned data is calculated, compared with the reference data, the system grade is automatically given, and the data is saved in the corresponding file.

6. The intelligent detection system for scrap steel processing and sorting as described in claim 1, characterized in that: The visual recognition system includes an image preprocessing module, which performs sample extraction functions such as image import extraction, random extraction, and image-based extraction, as well as temporary database operation functions such as printing temporary layers and deleting temporary layers.

7. The intelligent detection system for scrap steel processing and sorting as described in claim 1, characterized in that: The visual recognition system also includes an image viewing module, which is used to quickly perform batch image comparison based on the database and sort the comparison data using a robotic arm; Suspected materials that have been sorted out are photographed again and uploaded for manual processing. The materials are then sorted again using the manual data.

Images