A scrap steel packing sorting and identifying device and a sorting and identifying method

Abstract

The application provides a scrap steel packaging sorting and identifying device and a sorting and identifying method, and relates to the technical field of scrap steel packaging. The device comprises, in sequence along a conveying direction, a processing center, a conveying unit, a sorting and identifying unit, a scrap steel packaging unit, a sampling unit and a collecting unit. The processing center controls the sorting and identifying unit and the collecting unit. The conveying unit is used for conveying materials to the sorting and identifying unit. The sorting and identifying unit comprises an identifying system and a sorting device. The identifying system is used for identifying whether the scrap steel on the conveying unit contains a sealed container and sending identifying information to the processing center. The processing center sends a sorting signal to the sorting device. The scrap steel packaging unit is used for receiving the sorted scrap steel and packaging the scrap steel into scrap steel briquettes. The sampling unit is used for taking out the scrap steel briquettes in the scrap steel packaging unit. The collecting unit is used for collecting image information of the scrap steel briquettes and numbering the scrap steel briquettes at the same time, and transmitting the numbers and the collected information to the processing center for storage. The application can improve the quality of the scrap steel packaging briquettes and improve the efficiency of scrap steel processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of scrap steel processing, in particular to a scrap steel packaging method and system integrating sorting and identification, aiming to realize effective sorting of closed containers in scrap steel materials and type identification and information management of scrap steel briquettes through automation and intelligent means. BACKGROUND

[0002] In the process of scrap steel recycling and utilization, closed containers such as oil drums and gas tanks are often mixed in scrap steel materials. These containers may cause safety hazards during packaging and affect the quality of scrap steel. In addition, type identification of scrap steel briquettes is beneficial to improve the hit rate of converter steelmaking terminal, reduce smelting cost, and reduce steel plant inspection and management cost, which is of great significance to improve the efficiency and value of scrap steel recycling. However, the traditional manual sorting and identification method has the problems of low efficiency, high cost, easy to make mistakes, and cannot eliminate fraud.

[0003] At present, the method for detecting scrap steel briquettes mainly includes inputting two-dimensional code input scanning method, density detection method, image recognition processing method, etc. For example, the patent with publication number CN114519504A and publication date May 20, 2022, the invention is named as a scrap steel briquette inspection method. The analysis method of the application is as follows: handheld terminal procurement and APP system development, the supplier uses the equipment to input the briquette information and generates a two-dimensional code, the detector uses the terminal to scan the briquette two-dimensional code to complete the detection, and the inspection of the briquette information is completed. The beneficial effects of the application are as follows: the handheld terminal device can quickly complete the inspection data input, saving, photographing, picture uploading, card swiping and quality abnormality processing, and because the handwritten acceptance documents, account books and other complicated work links are abandoned, the efficiency of scrap steel inspection is greatly improved. However, this method still needs manual input of briquette information using a terminal, which is time-consuming and labor-intensive, not intelligent enough, and increases the production cost of scrap steel briquette suppliers. In view of the above problems, machine vision scrap steel intelligent judgment appears in the prior art, for example, the patent with publication number CN112348791A and publication date February 9, 2021, the invention is named as a scrap steel intelligent judgment method, system, medium and terminal based on machine vision. The application discloses a method for intelligently monitoring scrap steel materials. The analysis method of the application is as follows: the image data of the scrap steel unloading site is labeled to form a data set, a scrap steel detection model is trained, a camera is used to automatically capture scrap steel, and a trained detection model is used to intelligently judge the picture. The beneficial effects of the application are as follows: the machine vision completes the scrap steel judgment, greatly reducing the labor intensity of manual work. However, this method can only realize intelligent identification of scrap steel materials, and cannot realize subsequent sorting operation of scrap steel materials and identification and input operation of briquette types after scrap steel packaging operation.

[0004] In summary, how to realize the scrap steel packaging system with sorting and identification functions is a technical problem to be solved in the prior art. SUMMARY

[0005] 1. Problem to be solved

[0006] The present application provides a scrap steel packaging sorting and identification device, which aims to solve the problems of intelligent sorting and subsequent packaging of existing scrap steel containers, and can improve the quality of scrap steel and improve the efficiency of scrap steel processing. Further, a sorting and identification method for scrap steel containers and scrap steel briquettes is provided.

[0007] 2. Technical solution

[0008] In order to solve the above problems, the technical solution adopted by the present application is as follows:

[0009] As a first aspect of the present application, the present application provides a scrap steel packaging sorting and identifying device, which sequentially includes a processing center, a conveying unit, a sorting and identifying unit, a scrap steel packaging unit, a sampling unit and a collecting unit along the conveying direction, wherein,

[0010] The processing center controls the sorting and identifying unit and the collecting unit, and realizes the automation and intelligentization of the scrap steel packaging process.

[0011] The conveying unit is used to convey the materials to the sorting and identifying unit.

[0012] The sorting and identifying unit includes an identifying system and a sorting device, the identifying system is used to identify whether the scrap steel on the conveying unit contains a sealed container, and sends the identifying information to the processing center; the processing center sends a sorting signal to the sorting device to change the running path of the materials, so as to realize the automatic sorting of the sealed container, that is, the sorting device includes a control system, a first driving column, a second driving column and a shunt table, the first driving column and the second driving column are respectively connected at both ends of the bottom of the shunt table, the control system receives the identifying information sent by the processing center and controls the lifting or lowering of the first driving and the second driving, so as to control the tilting direction of the shunt table and drive the shunt table to change the material path, so as to realize the sorting operation of the scrap steel and the sealed container on the shunt table.

[0013] The scrap steel packaging unit is used to receive the scrap steel on the shunt table and package the scrap steel into scrap steel briquettes.

[0014] The sampling unit is used to take out the scrap steel briquettes in the scrap steel packaging unit.

[0015] The collecting unit is arranged at the output end of the scrap steel packaging unit or cooperates with the sampling unit, is used to collect the image information of the scrap steel briquettes at the output end of the scrap steel packaging unit or taken out by the sampling unit and number them at the same time, and transmit the collected information to the processing center for storage, so as to record the number in the database and facilitate the matching of the briquette information.

[0016] Further, the identifying system can efficiently and accurately identify the sealed container in the materials and transmit the result to the processing center; the specific identifying system is provided with a first image collecting device and a first edge computing terminal, the first image collecting device collects real-time images of the scrap steel materials on the conveying unit and transmits the image information to the first edge computing terminal, the first edge computing terminal judges whether the image information contains a sealed container and transmits the result to the processing center.

[0017] Further, the first edge computing terminal identifies each frame of image data of the real-time collected video by loading the trained YOLOv8 network, and judges whether the materials contain a sealed container.

[0018] If the recognition result is a closed container, the first edge computing terminal sends a signal to the processing center, and the processing center transmits the signal to the control system of the sorting device, controls the first driving column to lower and the second driving column to rise, so that the shunt table is inclined, and the closed container is sorted out.

[0019] If the recognition result is no closed container, the first edge computing terminal does not transmit a signal to the processing center, and the control system controls the first driving column to rise and the second driving column to lower, so that the shunt table is inclined, and the scrap steel is sorted into the scrap steel packaging unit for packaging operation.

[0020] Further, the sorting and identifying unit further comprises a collecting device, which is arranged on one side of the shunt table and close to the first driving column, and is used to receive the closed container sorted out by the shunt table.

[0021] Further, the collecting device is a wheel transfer box, which can transfer the collected closed containers for processing.

[0022] Further, in order to control the speed of the scrap steel entering the scrap steel packaging unit, a conveying unit is further arranged between the second driving column and the scrap steel packaging unit; thus the conveying unit continues to convey the scrap steel without closed containers after sorting to the scrap steel packaging unit, and the scrap steel packaging unit is a scrap steel baler, which compresses and packs the conveyed scrap steel.

[0023] Further, the conveying unit comprises a conveying motor, a conveying belt and a pair of conveying rollers, the conveying motor is connected to at least one conveying roller for controlling the rotation of the conveying roller, and the conveying belt is wound around the side surface of the pair of conveying rollers, and the conveying roller rotates to drive the conveying belt to rotate around it. Further, in order to avoid the scrap steel from falling during transportation, side baffles are arranged on both sides of the conveying direction of the conveying belt, and similarly, side baffles are also arranged on both sides of the length direction of the shunt table to avoid the scrap steel and closed containers from falling during the sorting and inclining process.

[0024] Further, the sampling unit comprises a fixed part, a steering drive, a rotating part, an extension part and a sampling part, the fixed part is vertically arranged and fixed in position, the steering drive is rotatably installed at the top end of the fixed part, and the standard parts that can be used include bearings, flanges and the like, the rotating part is arranged vertically with the fixed part and is fixedly connected to the steering drive, the end of the rotating part away from the steering drive is fixedly connected to the extension part, the extension part is arranged vertically with the rotating part, and the bottom end of the extension part is provided with the sampling part, when the sampling work of scrap steel briquettes is performed, the steering drive rotates at the top of the fixed part, and drives the rotating part to rotate around the steering drive, when the rotating part rotates above the scrap steel packing unit, the extension part retracts downward, the sampling part absorbs the scrap steel briquettes through electromagnetic action, the extension part retracts upward, and the rotating of the steering drive makes the rotating part rotate to be away from the scrap steel packing unit, that is, to be sent to the loading vehicle for continuous transportation. The sampling part is preferably an electromagnetic mechanism, and the extension part is preferably a hydraulic extension, so that the scrap steel briquettes are stably grabbed through the double action of the extension part and the sampling part, and the scrap steel briquettes are transferred and placed through the cooperation of the steering drive and the rotating part.

[0025] Further, the collection unit comprises a second edge computing terminal and at least three image collection devices, i.e., a second image collection device, a third image collection device and a fourth image collection device, which are respectively arranged at different positions of the grabbed scrap steel briquettes, so as to collect image information in real time by shooting the scrap steel briquettes, and transmit the image information to the second edge computing terminal for analyzing and identifying the types of the scrap steel briquettes, recording the information of the briquettes, and transmitting the information to a processing center to generate corresponding information, thereby supporting subsequent data query and management. The processing center numbers the scrap steel packing briquettes, records the identification results, numbers, packing time and packing location information into a database, and completes the generation and management of information.

[0026] As a second aspect of the present application, the present application provides a method for sorting and identifying scrap steel,

[0027] S1, collecting data and making a data set:

[0028] The image data of the individual closed containers and the image data of the mixed scrap steel of the on-site closed containers are collected through a python crawler, and a closed container data set is established;

[0029] The images of different scrap steel briquette types and the images of the scrap steel briquettes in the production site are collected through a python crawler, and a scrap steel briquette type data set is established;

[0030] The function of the python crawler is to download the images of the closed containers, so that the image data is more abundant;

[0031] S2, using labelimg to identify the closed containers in each picture of the closed container data set in step S1;

[0032] S3, using labelimg to recognize the scrap steel briquette category dataset in step S1, and the label information is the category corresponding to the briquette in the picture;

[0033] S4, using the Mixup technology to image enhance the closed container dataset identified in step S2 and the scrap steel briquette category dataset identified in step S3 to improve the diversity of samples and the generalization ability of the recognition model;

[0034] S5, building a Yolov8 deep learning framework on a PC, and performing multi-round training, for example, 60 rounds, on the closed container dataset and the scrap steel briquette category dataset after image enhancement in step S4, to obtain a closed container recognition model and a scrap steel briquette category recognition model, respectively;

[0035] S6, saving the closed container recognition model in step S5 in the SD card storage module of the first edge computing terminal, and saving the scrap steel briquette category recognition model in step S5 in the SD card storage module of the second edge computing terminal. Specifically, the recognition model is trained through a local computer, the optimal model is selected and converted into an ONNX format model, the ONNX format model is transmitted to an edge computing device that has installed an NCNN framework through a local storage medium (SD card), the ONNX format model is converted into a param file and a bin file supported by the NCNN framework, and a model for image processing is constructed.

[0036] Further, in step S4, the Mixup image enhancement generation formula is as follows:

[0037]

[0038] In the formula: is an input vector, is a data label. When training a neural network model, training samples need to be divided into small batches, for example, 1000 samples are divided into 10 batches, that is, 10 batches, (x i , y i ) and (x j , y j ) are two samples and corresponding labels randomly selected in the same batch, and λ is a number randomly sampled from the Beta distribution, λ∈Beta[0,1].

[0039] Further, the identification method of the closed container in step S2 and the identification method of the scrap steel briquette type in step S3 are as follows: using a target detection model Yolov8 as a backbone network, using a BCELoss as a classification loss function, for each class, the predicted probability is p and 1-p, and the BCELoss function calculates the probability of being this class and not being this class: the loss calculated in these two cases will not be 0:

[0040]

[0041] N is the number of samples, y i is the true label of the i-th sample, p i is the probability that the model predicts the i-th sample as positive (the base of the logarithm is e).

[0042] DFLLoss and CIoULoss are used in combination as the boundary box regression loss function. DFL Loss adjusts the weights of targets of different sizes dynamically, making the model more accurate in detecting targets of different sizes; and CIOU Loss improves the IOU Loss by introducing the aspect ratio and angle information, improving the accuracy of boundary box regression.

[0043]

[0044] y i and yi + 1 is the left and right integer value of the floating point value y, S is the output distribution, and DFL can make the network focus more quickly on the values near the target y, increasing their probability.

[0045]

[0046] ρ 2 (b,b gt ) represents the Euclidean distance between the center points of the predicted box and the real box, and c represents the diagonal distance of the smallest closed region that can contain the predicted box and the real box.

[0047] where the formulas for α and v are:

[0048]

[0049] V is a correction factor used to further adjust the loss function, w G ,h G and w p ,w p are the width and height of the target box and the predicted box, respectively.

[0050] 3. Beneficial effects

[0051] Compared with the prior art, the application has the beneficial effects that:

[0052] The application intelligently sorts out the closed container in the material adding process, reduces the safety hidden danger through sorting out the closed container, improves the quality of scrap steel, improves the efficiency of scrap steel treatment, reduces the labor cost and error rate, identifies the types of scrap steel briquettes, is beneficial to improve the hit rate of converter steelmaking endpoint, reduces the smelting cost, facilitates the management of scrap steel briquettes, and reduces the inspection and management cost of the steel plant. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 It is a structural schematic view of the scrap steel packaging sorting and identifying equipment of the application;

[0054] Figure 2 It is a structural schematic view of the transmission unit of the application;

[0055] Figure 3 It is a structural schematic view of the identifying system of the application;

[0056] Figure 4 It is a structural schematic view of the collecting unit of the application;

[0057] Figure 5 It is a flow schematic view of the sorting and identifying unit of the application;

[0058] Figure 6 It is a principle schematic view of the sorting and identifying method of the application;

[0059] In the figure:

[0060] 0, processing center; 100, conveying unit; 110, conveying motor; 120, conveying belt; 130, conveying roller; 140, side baffle; 200, sorting and identifying unit; 210, identifying system; 211, first image collecting device; 212, first edge computing terminal; 220, sorting device; 221, control system; 222, first driving column; 223, second driving column; 224, shunt table; 225, collecting device; 300, scrap steel packaging unit; 400, sampling unit; 410, fixing piece; 420, turning drive; 430, turning piece; 440, telescopic piece; 450, sampling piece; 500, collecting unit; 510, second edge computing terminal; 520, second image collecting device; 530, third image collecting device; 540, fourth image collecting device. DETAILED DESCRIPTION

[0061] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0062] Embodiments

[0063] The image recognition-based scrap steel packaging and sorting device of the present embodiment comprises a processing center 0, a conveying unit 100, a sorting and identifying unit 200, a scrap steel packaging unit 300, a sampling unit 400, and a collecting unit 500. The processing center 0 is used to control the sorting and identifying unit 200 and the collecting unit 500 and store information, realizing the automation and intelligentization of the scrap steel packaging process. The materials are sent to the sorting and identifying unit 200 through the conveying unit 100, and the scrap steel and the sealed container are sorted through the identification of the sorting and identifying unit 200. The scrap steel is sent to the scrap steel packaging unit 300, i.e., the scrap steel baler, for compression and packaging into scrap steel briquettes. The sampling unit 400 grabs and transfers the scrap steel briquettes to the transport vehicle. At the same time, the collecting unit 500 identifies the scrap steel briquettes output from the scrap steel baler or grabbed by the sampling unit 400 and transmits the identification information to the processing center for numbering and storage, facilitating management.

[0064] Specifically:

[0065] The processing center 0 is a high-performance computer or server. The present application selects a high-performance computer, a Windows 11 system, a CPU of Intel i7-14700KF, and a graphics card of MSI 4080S16G. The edge computing device selects Raspberry Pi 4B2018, the operating system is Linux 32bit, and the CPU is armv71. The sorting and identifying unit 200 and the collecting unit 500 are connected through a local area network.

[0066] The conveying unit 100 comprises a conveying motor 110, a conveying belt 120, and a pair of conveying rollers 130. The conveying motor 110 is connected to at least one conveying roller 130 to control the rotation thereof. The conveying belt 120 is wound around the sides of the pair of conveying rollers 130. The conveying rollers 130 rotate to drive the conveying belt to rotate around them. It should be noted that, in order to avoid the scrap steel from falling during transportation, side baffles 140 are arranged on both sides of the conveying direction of the conveying belt, as shown in Figure 2 .

[0067] The sorting and identifying unit 200 comprises an identification system 210 and a sorting device 220, as shown in Figure 3As shown, the recognition system 210 is provided with a first image acquisition device 211 and a first edge computing terminal 212. The first image acquisition device 211 in this embodiment is a camera, which acquires real-time video images of the scrap steel materials on the conveying belt 120 and transmits image information to the first edge computing terminal 212. The first edge computing terminal 212 identifies and processes each frame of image data of the real-time acquired video by loading the trained YOLOv8 network, judges whether the materials contain sealed containers, and transmits the result to the processing center 0. The processing center 0 sends a sorting signal to the sorting device 220. The sorting device 220 includes a control system 221, a first driving column 222, a second driving column 223, and a shunt table 224. The first driving column 222 and the second driving column 223 are respectively connected to the two ends of the bottom of the shunt table 224. The control system 221 receives the identification information sent by the processing center 0 and controls the first driving 222 and the second driving 223 to rise or fall, so as to control the inclination direction of the shunt table 224, drive the shunt table 224 to change the material path, and realize the sorting operation of the scrap steel and the sealed container on the shunt table 224. Similarly, the two sides of the length direction of the shunt table 224 are also provided with side baffles 140 to avoid the falling of scrap steel and sealed containers during the sorting inclination process.

[0068] As shown in Figure 5 If the identification result of the first edge computing terminal 212 is a sealed container, the first edge computing terminal 212 sends a signal to the processing center 0, and the processing center 0 transmits the signal to the control system 221 of the sorting device 220, controls the first driving column 222 to lower and the second driving column 223 to rise, so as to incline the shunt table 224 and sort out the sealed container. It should be noted that a collecting device 225 can be placed, which can be a wheeled transfer box, which can transfer the collected sealed containers for processing.

[0069] If the identification result is no sealed container, the first edge computing terminal 212 does not transmit a signal to the processing center 0, and the control system 221 controls the first driving column 222 to rise and the second driving column 223 to lower, so as to incline the shunt table 224 and sort the scrap steel to the scrap steel packaging unit 300 for packaging operation.

[0070] As shown in Figure 1 In order to control the speed of the scrap steel entering the scrap steel baling press, a conveying unit 100 is further arranged between the second driving column 223 and the scrap steel baling press. The conveying unit 100 continues to convey the scrap steel without sealed containers after sorting to the scrap steel baling press for compression and packaging processing.

[0071] The sampling unit 400 comprises a fixed part 410, a turning drive 420, a rotating part 430, an extension part 440 and a sampling part 450. The fixed part 410 is vertically arranged and fixed in position. The turning drive 420 is rotatably arranged at the top end of the fixed part 410. The rotating part 430 is vertically arranged with the fixed part 410 and is fixedly connected to the turning drive 420. The end of the rotating part 430 away from the turning drive 420 is fixedly connected to the extension part 440. The extension part 440 is vertically arranged with the rotating part 430. The bottom end of the extension part 440 is provided with the sampling part 450. When the sampling unit 400 is used to grab the scrap steel briquettes, the turning drive 420 is rotated at the top of the fixed part 410, and the rotating part 430 is rotated around the turning drive 420. When the rotating part 430 is rotated to above the scrap steel packing unit 300, the extension part 440 is retracted downward, and the sampling part 450 is attracted to the scrap steel briquettes by electromagnetic action. The extension part 440 is retracted upward, and the turning drive 420 is rotated to rotate the rotating part 430 to move the grabbed scrap steel briquettes away from the scrap steel packing unit 300, i.e. to the loading vehicle for further transportation. The sampling part 450 is preferably an electromagnetic mechanism, and the extension part 440 is preferably a hydraulic extension. The scrap steel briquettes are stably grabbed by the double action of the extension part 440 and the sampling part 450, and are transferred and placed by the cooperation of the turning drive 420 and the rotating part 430.

[0072] As shown in Figure 4 The collection unit 500 comprises a second edge computing terminal 510 and at least three cameras, i.e. a second image collection device 520, a third image collection device 530 and a fourth image collection device 540, which are arranged at different positions of the grabbed scrap steel briquettes to capture images of the scrap steel briquettes in real time, transmit the image information to the second edge computing terminal 510 for analysis and identification of the types of the scrap steel briquettes, record the information of the briquettes, and transmit the information to the processing center 0 to generate corresponding information to support subsequent data query and management. The processing center 0 numbers the scrap steel packing briquettes, records the identification results, numbers, packing time and packing location information into the database, and completes the generation and management of the information.

[0073] The working method of the scrap steel packing and sorting and identification equipment based on image recognition in the embodiment is as follows:

[0074] The scrap steel materials are transported to the conveying belt 120. The materials are continuously transported forward. When the materials are transported to the end of the conveying belt 120, the identification system 210 collects real-time video through the camera. The first edge computing terminal 212 uses the YOLOV8 network to identify whether the materials contain sealed containers.

[0075] When the first edge computing terminal 212 identifies that the materials do not contain sealed containers, no signal is sent to the processing center 0, and the sorting device 220 is as shown in Figure 1The first driving column 222 is kept high and the second driving column 223 is kept low, and the material transported to the distribution table 224 will fall onto the conveying belt 120 of the second conveying unit 100 due to gravity, and be transported to the direction of the scrap steel packing machine, and finally fall into the scrap steel packing machine for packing treatment. After the packing is completed, the sampling unit 400 places the sampling member 450 above the scrap steel briquette through the turning drive 420, and then lowers the sampling member 450 onto the scrap steel briquette through the hydraulic action of the telescopic member 440, and then sucks the scrap steel briquette through electromagnetic action. After the sucking is completed, the telescopic member 440 drives the sampling member 450 to move upward, and at this time, the three cameras of the collection unit 500 start to work and shoot the three surfaces of the scrap steel briquette, respectively. The edge computing device 61 identifies and processes the three photos through the pre-trained scrap steel briquette type identification model, the second edge computing terminal 510 transmits the identified scrap steel type, shooting time and shooting location information to the processing center 0. The processing center 0 saves the information in the database.

[0076] When the first driving column 222 is kept low and the second driving column 223 is kept high, the material transported to the distribution table 224 will fall into the collecting device 225 due to gravity.

[0077] When the first driving column 222 is kept low and the second driving column 223 is kept high, the material transported to the distribution table 224 will fall into the collecting device 225 due to gravity.

[0078] The scrap steel packing machine compresses and packs the scrap steel into scrap steel briquettes and delivers them to the output end. The sampling unit 400 rotates the turning drive 420 so that the sampling member 450 is above the scrap steel briquette, the telescopic member 440 moves downward, the sampling member 450 grabs the scrap steel briquette through electromagnetic attraction, the three cameras of the collection unit 500 shoot the scrap steel briquette from three different directions, and the video image information is transmitted to the second edge computing terminal 510 for analysis and identification of the type of scrap steel briquette and recording of the briquette information, and then transmitted to the processing center 0 to generate corresponding information to support subsequent data query and management. After the scrap steel briquette information collection is completed, the telescopic member 440 moves upward, and the turning drive 420 rotates to move the scrap steel briquette out of the scrap steel packing machine. Of course, the process of collecting scrap steel briquette video information by the collection unit 500 can also be directly performed at the output end of the scrap steel packing machine.

[0079] The application provides a method for collecting information of whether a closed container and a scrap steel briquette are contained in scrap steel,

[0080] S1, collecting data and making a data set:

[0081] The image data of the individual closed container and the image data of the mixed scrap steel of the closed container are collected through a python crawler and on-site collection, and a closed container data set is established;

[0082] The image data of different scrap steel briquette types and the image data of the scrap steel briquette in the production site are collected through a python crawler and on-site collection, and a scrap steel briquette type data set is established;

[0083] S2, using labelimg to identify the closed container in each picture of the closed container data set in step S1;

[0084] S3, using labelimg to identify the scrap steel briquette type data set in step S1, and the information marked is the type corresponding to the briquette in the picture;

[0085] S4, using the Mixup technology to perform image enhancement on the closed container data set identified in step S2 and the scrap steel briquette type data set identified in step S3 to improve the diversity of the samples and the generalization ability of the identification model; the Mixup image enhancement generation formula is as follows:

[0086]

[0087] In the formula: is an input vector, is a data label, when training a neural network model, the training samples need to be divided into small batches, for example, 1000 samples are divided into 10 batches, that is, 10 batches, (x i ,y i ) and (x j ,y j ) are two samples and corresponding labels randomly selected in the same batch, and λ is a number randomly sampled from the Beta distribution, λ∈Beta[0,1].

[0088] S5, building a Yolov8 deep learning framework on a PC, and performing multi-round training, for example, 60 rounds, on the closed container data set and the scrap steel briquette type data set after image enhancement in step S4, to obtain a closed container identification model and a scrap steel briquette type identification model, respectively;

[0089] S6, as Figure 6The closed container recognition model in step S5 is saved in the SD card storage module of the first edge computing terminal 212, and the scrap steel briquette type recognition model in step S5 is saved in the SD card storage module of the second edge computing terminal 510. The specific transplantation method is as follows: training the recognition model through a local computer, selecting the optimal model to convert into an ONNX format model, transferring the ONNX format model to an edge computing device that has installed an NCNN framework through a local storage medium (an SD card), converting the ONNX format model into a param file and a bin file supported by the NCNN framework by the NCNN framework, and constructing a model for image processing.

[0090] In particular, the recognition of the closed container in step S2 and the recognition of the scrap steel briquette type in step S3 are as follows: using a target detection model Yolov8 as a backbone network and using BCELoss as a classification loss function, the probability p and 1-p predicted for each class, the BCELoss function calculates the probability of being this class and not being this class, and the loss calculated in these two cases is not 0:

[0091]

[0092] L is the loss function value, N is the sample number, y i is the true label of the i-th sample, p i is the probability that the model predicts the i-th sample as a positive class (the base of the logarithm is e).

[0093] DFLLoss and CIoULoss are used in combination as the boundary box regression loss function. DFL Loss adjusts the weights of targets of different sizes dynamically, so that the model detects targets of different sizes more accurately; and CIOU Loss improves the IOU Loss by introducing the width-height ratio and angle information, thereby improving the accuracy of boundary box regression.

[0094]

[0095] y i and yi + 1 is the left and right integer value of the floating point value y, and S is the output distribution. DFL can make the network focus more quickly on the values near the target y, and increase their probabilities.

[0096]

[0097] IOU reflects the spatial relationship between two sets and is used to describe the degree of overlap between two parts. IOU is the result of dividing the intersection of two regions by the union of the two regions

[0098] p2 (b,b gt ) represents the Euclidean distance of the center points of the prediction box and the real box, and c represents the diagonal distance of the minimum closed region capable of containing the prediction box and the real box.

[0099] Wherein the formula of alpha and v is:

[0100]

[0101] V is a correction factor, used to further adjust the loss function, w G ,h G and w p ,w p are the width and height of the target box and the prediction box respectively.

[0102] After multiple rounds of training, the closed container recognition model has an identification accuracy of 91.43% for closed containers; and after multiple rounds of training, the scrap steel briquette type recognition model has an identification accuracy of 88.76% on the scrap steel briquette data set.

[0103] Through the above steps, the present embodiment realizes efficient sorting, packing and identification management of scrap steel, significantly improves the automation level and traceability of scrap steel processing, and provides strong support for the green development of the steel industry.

[0104] From the common general knowledge, the present application can be realized by other embodiments without departing from the spirit or essential characteristics thereof. Therefore, the above disclosed embodiments, in all aspects, are only illustrative and not restrictive. All changes within the scope of the present application or within the equivalent scope of the present application are included in the present application.

Claims

1. A scrap bale sorting and identification apparatus, characterized by: The processing center (0), the conveying unit (100), the sorting and identifying unit (200), the scrap steel packaging unit (300), the sampling unit (400) and the collection unit (500) are sequentially arranged along the conveying direction, wherein The processing center (0) controls the sorting and identifying unit (200) and the collection unit (500); The conveying unit (100) is used for conveying materials to the sorting and identifying unit (200); The sorting and identifying unit (200) comprises an identifying system (210) and a sorting device (220), the identifying system (210) is used for identifying whether the scrap steel on the conveying unit (100) contains a sealed container, and sending the identifying information to the processing center (0); the processing center (0) sends a sorting signal to the sorting device (220) to change the running path of the materials, and automatically sorts the sealed container, wherein the sorting device (220) comprises a control system (221), a first driving column (222), a second driving column (223) and a shunt table (224), the first driving column (222) and the second driving column (223) are respectively connected at two ends of the bottom of the shunt table (224), the control system (221) receives the identifying information sent by the processing center (0) and controls the lifting or lowering of the first driving column (222) and the second driving column (223) to control the tilting direction of the shunt table (224), drives the shunt table (224) to change the material path, and realizes the sorting operation of the scrap steel and the sealed container on the shunt table (224); The scrap steel packaging unit (300) is used for receiving the scrap steel on the shunt table (224) and packaging the scrap steel into scrap steel briquettes; The sampling unit (400) is used for taking out the scrap steel briquettes in the scrap steel packaging unit (300); The collection unit (500) is arranged at the output end of the scrap steel packaging unit (300) or is used in cooperation with the sampling unit (400), is used for collecting image information of the scrap steel briquettes at the output end of the scrap steel packaging unit (300) or taken out by the sampling unit (400) and numbering, and transmitting the numbering and the collection information to the processing center (0) for storage.

2. A scrap bale sorting and identification apparatus according to claim 1, characterised in that: The identifying system (210) is provided with a first image acquisition device (211) and a first edge computing terminal (212), the first image acquisition device (211) acquires real-time images of the scrap steel materials on the conveying unit (100) and transmits the image information to the first edge computing terminal (212), and the first edge computing terminal (212) judges whether the image information contains a sealed container and transmits the result to the processing center (0).

3. A scrap bale sorting and identification apparatus according to claim 2, characterised in that: The first edge computing terminal (212) identifies each frame of image data of the real-time acquisition video by loading the trained YOLOv8 network, judges whether the material contains a sealed container.

4. A scrap bale sorting and identification apparatus as claimed in claim 1, characterized in that: The sorting and identifying unit (200) further comprises a collecting device (225), the collecting device (225) is arranged on one side of the shunt table (224) and close to the first driving column (222), and is used for receiving the sealed container sorted out by the shunt table (224).

5. A scrap bale sorting and identification apparatus as claimed in claim 1, wherein: A conveying unit (100) is further arranged between the second driving column (223) and the scrap steel packing unit (300); the conveying unit (100) continuously conveys the sorted scrap steel without the sealed container to the scrap steel packing unit (300), and the scrap steel packing unit (300) is a scrap steel packing machine for compressing and packing the conveyed scrap steel.

6. A scrap bale sorting and identification apparatus according to claim 5, wherein: The conveying unit (100) comprises a conveying motor (110), a conveying belt (120) and a pair of conveying rollers (130), the conveying motor (110) is connected with at least one conveying roller (130) for controlling the rotation of the conveying roller (130), and the conveying belt (120) is wound around the side of the pair of conveying rollers (130), and the conveying roller (130) rotates to drive the conveying belt to rotate.

7. A scrap bale sorting and identification apparatus as claimed in claim 1, wherein: The sampling unit (400) comprises a fixing member (410), a turning drive (420), a rotating member (430), an extension member (440) and a sampling member (450), the fixing member (410) is vertically arranged in a fixed position, the turning drive (420) is rotatably arranged at the top end of the fixing member (410), the rotating member (430) is vertically arranged on the fixing member (410) and is fixedly connected with the turning drive (420), the end of the rotating member (430) away from the turning drive (420) is fixedly connected with the extension member (440), the extension member (440) is vertically arranged on the rotating member (430), and the bottom end of the extension member (440) is provided with the sampling member (450), when the scrap steel briquettes are grabbed, the turning drive (420) rotates at the top of the fixing member (410) and drives the rotating member (430) to rotate around the turning drive (420), when the rotating member (430) rotates above the scrap steel packing unit (300), the extension member (440) is retracted downward, the sampling member (450) absorbs the scrap steel briquettes through electromagnetic action, the extension member (440) is retracted upward, the turning drive (420) rotates to make the rotating member (430) rotate to the scrap steel briquettes grabbed away from the scrap steel packing unit (300), i.e. to the loading vehicle for continuous transportation; wherein the sampling member (450) is an electromagnetic mechanism, the extension member (440) is a hydraulic extension, the scrap steel briquettes are stably grabbed through the double action of the extension member (440) and the sampling member (450), and the scrap steel briquettes are transferred and placed through the cooperation of the turning drive (420) and the rotating member (430).

8. A scrap bale sorting and identification apparatus as claimed in claim 5, wherein: The collection unit (500) includes a second edge computing terminal (510) and at least three image collection devices, i.e., a second image collection device (520), a third image collection device (530), and a fourth image collection device (540), which are respectively arranged at different positions of the grabbed scrap steel briquettes, collect image information by shooting the scrap steel briquettes in real time, and transmit the image information to the second edge computing terminal (510) to analyze and identify the types of the scrap steel briquettes, record the briquette information, transmit the information to a processing center (0) to generate corresponding information, and support subsequent data query and management; the processing center (0) numbers the scrap steel packing briquettes, records the identification result, the number, the packing time, and the packing location information into a database, and completes the generation and management of information.

Images