Rotary-cut plate grading and sorting control system based on visual identification

Abstract

The invention relates to the technical field of plate sorting, in particular to a rotary-cut plate grading and sorting control system based on visual identification, which comprises an execution module, an image acquisition module, an image processing module and a control module, the execution module comprises a hydraulic loading and unloading unit, a variable-frequency feeding and receiving unit, a crank plate disassembling unit, a conveying servo roller, a variable-frequency conveying belt, a servo sorting electric cylinder and a sensor unit. The image acquisition module is used for acquiring an image of the plate flattened by the conveying servo roller; the image processing module is used for acquiring data of the image acquisition module so as to analyze the quality of the flattened plate; the control module is used for data processing and action control of the execution module, the image obtaining module and the image processing module, and the control module controls sorting of the servo sorting electric cylinder according to the quality of the plates through data of the image processing module. The rotary-cut plates are subjected to defect identification and classification, and the plate screening operation efficiency can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of plate sorting, and in particular to a rotary cut plate grading and sorting control system based on visual recognition. Background Art

[0002] Wood processing is the processing of wood as raw material, mainly by mechanical or chemical methods, and its products still retain the basic characteristics of wood. In the forest industry, wood processing industry and forest product chemical processing are both follow-up industries of forest felling and transportation, and are important sectors for the comprehensive utilization of wood resources. Wood has the advantages of light weight, high strength-to-weight ratio, good elasticity, impact resistance, rich and beautiful texture and color, and easy processing. It has been listed as an important raw material since ancient times.

[0003] Due to objective reasons such as the small scale and scattered distribution of peeling plants in the wood industry, the quality of veneers purchased by board manufacturers is different, and it is easy to have a high product scrap rate during subsequent processing, which affects production efficiency. In board processing, manual screening is generally used, and the degree of automation is extremely low, but there are still a series of problems such as high cost, low efficiency, and unstable screening quality. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a rotary cut sheet material grading and sorting control system based on visual recognition, which can effectively improve the efficiency of sheet material screening operations by identifying and grading defects in the rotary cut sheet material.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A rotary cut sheet material grading and picking control system based on visual recognition, comprising an execution module, an image acquisition module, an image processing module and a control module.

[0007] The execution module includes a hydraulic loading and unloading unit, a frequency conversion feeding and receiving unit, the crank plate removal unit, a conveying servo roller, a frequency conversion conveying belt, a servo sorting electric cylinder, and a sensor unit.

[0008] The hydraulic loading and unloading unit is used for loading unprocessed plates and unloading sorted plates in the variable frequency supply and receiving unit; the variable frequency supply and receiving unit is used for collecting plates in the hydraulic loading and unloading unit and collecting sorted plates in the servo sorting electric cylinder; the crank plate disassembly unit is used for disassembling plates in the variable frequency supply and receiving unit; the conveying servo roller is used for flattening plates in the crank plate disassembly unit; the variable frequency conveying belt is used for conveying flattened plates; the servo sorting electric cylinder is used for sorting plates on the variable frequency conveying belt to the variable frequency supply and receiving unit; the sensor unit is used to monitor the plates in each unit through sensors to obtain status data of equipment and plates in each unit;

[0009] The image acquisition module is used to acquire the image of the plate after being flattened by the conveying servo roller;

[0010] The image processing module is used for acquiring data from the image acquisition module to analyze the quality of the flattened plate;

[0011] The control module is used for data processing and action control of the execution module, the image acquisition module, and the image processing module, and the control module controls the sorting of the servo sorting electric cylinder according to the quality of the plate through the data of the image processing module.

[0012] Furthermore, after receiving the plate, the conveying servo roller obtains the position information of the plate through the encoder and the sensor. After the plate reaches the preset position, the control module sends a capture signal to the image acquisition module so that the image acquisition module captures the image of the plate through the camera.

[0013] Furthermore, the image processing module includes a labeling submodule, a model training submodule, a preprocessing submodule and an analysis submodule.

[0014] The marking submodule is used for marking defects of the plate;

[0015] The model training submodule is used for data acquisition of the marking submodule, and the model training submodule performs model training according to the marked plate defects to obtain a recognition model;

[0016] The preprocessing submodule is used to acquire the plate image taken by the camera, and the preprocessing submodule preprocesses the plate image to obtain a preprocessed image;

[0017] The analysis submodule is used for data acquisition of the preprocessing submodule, and the analysis submodule analyzes the preprocessing image through the recognition model to obtain the recognition result of the plate defect.

[0018] Furthermore, the marking submodule calculates and obtains acquisition parameters according to the resolution of the camera and the marking size required for the plate.

[0019] Furthermore, the model training submodule imports the data of the labeling submodule into the training model, and the model training submodule performs model training by setting the training parameters of the number of iterations, basic learning rate, patch size, model capability, angle capability, and whether to prune to obtain a recognition model.

[0020] Furthermore, the preprocessing submodule segments the board image, and the preprocessing submodule collects two boards through an image source, and processes two objects in one image source into two independent images to be processed based on wooden boards to obtain the preprocessed image.

[0021] Furthermore, after the analysis submodule sets the information of the calculation threshold, the region of interest, and the number of calculated defects for the recognition model, the pre-processed image is input into the recognition model.

[0022] Furthermore, the recognition model extracts defect information of each marked defect type, area size, quantity, angle, position, and confidence level from the preprocessed image, and the analysis submodule performs weight analysis based on the defect information to obtain plate grade information, and the analysis submodule sends the plate grade information to the control module, so that the servo sorting electric cylinder receives the plate grade information from the control module to sort the plates according to different grades.

[0023] Furthermore, the image processing module further includes a data formatting submodule, which is used for acquiring data of the analyzing submodule, and the data formatting submodule integrates the defect information into a data packet that can be recognized by the executing module.

[0024] Furthermore, the data formatting submodule sends the data packet to the control module, and the control module sends the data packet to the execution module through the TCP / IP protocol to input a control signal to the execution module.

[0025] The beneficial effects of the present invention are:

[0026] 1. The execution module unloads, collects, disassembles, flattens, conveys and sorts the plates in sequence. At the same time, the image acquisition module acquires the image of the flattened plate. The image processing module identifies the plate, and the control module performs weight analysis based on the defect information to obtain the grade of the plate. The servo sorting cylinder sorts the plate according to the plate grade signal of the control module. The control module transmits the control signal to the execution module through the field bus. The execution module is connected to the mechanical execution structure and drives each mechanical structure. The sensor unit detects the movement result of the mechanical execution structure in each execution module and feeds the result back to the drive unit, thereby realizing automatic defect recognition and grading of the peeled plate, which can effectively improve the efficiency of plate screening operations.

[0027] 2. After the conveying servo roller flattens the plate, the camera can obtain higher quality images. At the same time, the camera does not need to stop the conveyor line during the acquisition process, which further improves the sorting efficiency, and can acquire images of the front and back of the plate in a more limited space. The image is affine transformed by the preprocessing submodule to improve the defect recognition accuracy of the analysis submodule. The defect types and plate quality grades are quantified by the analysis submodule standards, so that the plates are accurately graded and sorted, overcoming the drawback of being too subjective in manually identifying and sorting plates based on experience. At the same time, the present invention can be applied to relatively large plates. There is no need to transport and move the plates during the identification and sorting process, which effectively improves the efficiency of graded sorting. In addition, there is no need to frequently turn over the plates during the graded sorting process to avoid damage to the plates. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a structural block diagram of a preferred embodiment of the present invention.

[0029] Figure 2 It is a structural block diagram of an image processing module according to a preferred embodiment of the present invention.

[0030] Figure 3 It is a recognition flow chart of a preferred embodiment of the present invention.

[0031] In the figure, 1-execution module, 11-hydraulic loading and unloading unit, 12-frequency feeding and receiving unit, 13-the crank plate removal unit, 14-conveying servo roller, 15-frequency conveying belt, 16-servo sorting electric cylinder, 17-sensor unit, 2-image acquisition module, 3-image processing module, 31-labeling submodule, 32-model training submodule, 33-preprocessing submodule, 34-analysis submodule, 35-data formatting submodule, 4-control module. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] It should be noted that when a component is referred to as being "fixed to" another component, it may be directly on the other component or there may also be a component centered. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may also be a component centered. When a component is considered to be "set on" another component, it may be directly set on the other component or there may also be a component centered. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0035] Please also see Figures 1 to 3 A preferred embodiment of the present invention is a rotary-cut sheet material grading and sorting control system based on visual recognition, which includes an execution module 1, an image acquisition module 2, an image processing module 3 and a control module 4.

[0036] The execution module 1 includes a hydraulic loading and unloading unit 11, a frequency conversion feeding and receiving unit 12, a crank plate removal unit 13, a conveying servo roller 14, a frequency conversion conveying belt 15, a servo sorting electric cylinder 16, and a sensor unit 17.

[0037] The hydraulic loading and unloading unit 11 is used for loading unprocessed plates and unloading sorted plates in the variable frequency supply and receiving unit 12. The hydraulic loading and unloading unit 11 includes an electrically controlled hydraulic valve.

[0038] The variable frequency material supply and receiving unit 12 is used for receiving the plates in the hydraulic loading and unloading unit 11 and the plates that have been sorted in the servo sorting electric cylinder 16, and the variable frequency material supply and receiving unit 12 includes a variable frequency material supply unit and a variable frequency material receiving unit. The variable frequency material supply unit is used for receiving the plates in the hydraulic loading and unloading unit 11 to feed the crank plate removal unit 13; the variable frequency material receiving unit is used for receiving the plates that have been sorted by the servo sorting electric cylinder 16. The variable frequency material supply and receiving unit 12 includes a frequency converter and a three-phase motor.

[0039] The crank plate disassembly unit 13 is used for disassembling the plates in the variable frequency feeding and receiving unit 12. The crank plate disassembly unit 13 comprises a vertical crank mechanism, a horizontal crank mechanism, a bus servo driver, a servo motor, an absolute encoder, a vacuum generator, and a vacuum airflow control cylinder.

[0040] The conveying servo roller 14 is used for flattening the plate in the crank plate removal unit 13. The conveying servo roller 14 includes a conveying roller, a bus servo driver, a servo motor, an absolute value frequency converter, and a laser thickness measurement module.

[0041] The variable frequency conveyor belt 15 is used for conveying the flattened plate. The variable frequency conveyor belt 15 comprises a conveyor belt, a frequency converter, and a three-phase motor.

[0042] The servo sorting electric cylinder 16 is used to sort the plates on the variable frequency conveying belt 15 to the variable frequency feeding and receiving unit 12. The servo sorting electric cylinder 16 includes a screw electric cylinder, a bus servo driver, a servo motor, and an absolute encoder.

[0043] The sensor unit 17 is used to monitor the plates in each unit through sensors to obtain the status data of the equipment and plates in each unit. The sensor unit 17 collects the operating status of the on-site equipment and the plate conveying status and transmits them back to the control module 4.

[0044] The upper servo motors are connected to the driving rollers, electric cylinders and cranks of the corresponding mechanical structures through couplings; the three-phase motors are connected to the conveyor chains and conveyor belts of the corresponding mechanical structures through sprockets.

[0045] The image acquisition module 2 is used to acquire the image of the plate after being flattened by the conveying servo roller 14 .

[0046] The image processing module 3 is used for acquiring data from the image acquisition module 2 to analyze the quality of the flattened plate.

[0047] The control module 4 is used for data processing and motion control of the execution module 1, the image acquisition module 2, and the image processing module 3, and the control module 4 controls the sorting of the servo sorting cylinder 16 according to the quality of the plate through the data of the image processing module 3.

[0048] In this embodiment, a host computer is provided, which is an industrial computer installed with WINDOWS system, the control module 4 is a PLC controller, and the human-computer interaction terminal is a touch screen; the touch screen display communicates with the control module 4 through the codesys v3 (ethernet) protocol, and the host computer communicates with the control module 4 using the TCP / IP free protocol.

[0049] The control module 4 includes a frequency converter VDF1-VDF10, a servo driver SER1-SER13, a hydraulic solenoid valve KA1-KA20, and a pneumatic solenoid valve KA21-KA25. KA1-KA20 controls the on-off of the hydraulic oil circuit, KA21-KA25 controls the vacuum generator, and the vacuum airflow controls the cylinder. The frequency converter VDF1-VDF10 controls the operation of the variable frequency feeding and receiving unit 12 conveyor chain and the variable frequency conveyor belt, and the servo driver SER1-SER13 controls the operation of the crank, electric cylinder, and servo roller.

[0050] Among them, the PLC controller is a HCQ series controller, the frequency converter is a GT200 series, the servo driver is an X3EB series, and the servo motor is an X2 series.

[0051] In this embodiment, the execution module 1 unloads, collects, disassembles, flattens, conveys and sorts the plates in sequence, while the image acquisition module acquires the image of the flattened plate, the image processing module identifies the plate, and the control module performs weight analysis based on the defect information to obtain the grade of the plate. The servo sorting cylinder 16 sorts the plate according to the plate grade signal of the control module 4. The control module transmits the control signal to the execution module through the field bus. The execution module is connected to the mechanical execution structure and drives each mechanical structure. The sensor unit detects the movement result of the mechanical execution structure in each execution module and feeds the result back to the drive unit, thereby realizing automatic defect recognition and grading of the peeled plate, which can effectively improve the efficiency of plate screening operations. This embodiment adopts an on-site communication method to avoid complicated on-site wiring. Through the feedback of the sensor unit, the idle stroke of the mechanical movement is greatly reduced, energy consumption is reduced, and operation efficiency is improved.

[0052] In this embodiment, after receiving the plate, the conveying servo roller 14 obtains the position information of the plate through the encoder and the sensor. After the plate reaches the preset position, the control module 4 sends a capture signal to the image acquisition module 2 so that the image acquisition module 2 captures the image of the plate through the camera.

[0053] In this embodiment, the control logic of the control module 4 is:

[0054] Confirm that the material cart is in place and that the plates on the material cart are normally present. The solenoid valve Y1 of the hydraulic loading and unloading unit 11 at the feeding station is actuated to control the hydraulic cylinder to rise and load the plates, thereby raising the plates to the feeding and handover position.

[0055] After the plate rises to the feeding handover position, VDF1 of the variable frequency feeding unit starts to drive the receiving chain to take over the plate on the hydraulic cylinder to the variable frequency feeding unit and lift it to the plate removal position.

[0056] After the plate rises to the plate removal position, SER1 of the crank plate removal unit 13 uses KA21 to control the vacuum generator to create vacuum while moving downward. After successfully sucking the plate, it rises vertically and feeds it horizontally to the conveying servo roller 14.

[0057] After receiving the plate, the conveying servo roller 14 confirms the actual position of the plate through the encoder and sensor and continues to convey it. When it reaches the specified position, it starts to send a start acquisition signal to the camera to control it to start acquiring images. After the acquisition is completed, the image processing module 3 uses a deep learning model to analyze the type, quantity, area, and position of the defects and evaluates the quality level based on the comprehensive weight.

[0058] After processing the image and assessing the quality level, the image processing module 3 sends the result to the control module 4. The control module 4 controls the SER2 of the conveying servo roller 14 to convey the plate to the designated sorting station via the VDF2 of the variable frequency conveying belt 15.

[0059] After the VDF2 of the variable frequency conveyor belt 15 conveys the plate to the designated sorting station, the SER3 of the servo sorting electric cylinder 16 moves to push the plate to the variable frequency receiving unit and memorizes the received quantity.

[0060] When the material receiving count of the variable frequency receiving unit reaches the set value, the servo sorting electric cylinder SER3 will be temporarily blocked and sorting will not be allowed. The variable frequency receiving unit will transport the plate downward to the unloading and handover position of the hydraulic loading and unloading unit 11. The KA3 action of the hydraulic loading and unloading unit 11 will transfer the plate from the variable frequency receiving unit to the bottom of the internal buffer area of ​​the hydraulic loading and unloading unit 11, and at the same time unlock the SER3 of the servo sorting electric cylinder 16.

[0061] After the plate arrives below the internal buffer area of ​​the hydraulic loading and unloading unit 11, the fork driven by the hydraulic cylinder transfers the plate to the material truck, and a complete sorting process is completed.

[0062] The image processing module 3 includes a labeling submodule 31 , a model training submodule 32 , a preprocessing submodule 33 , an analysis submodule 34 , and a data formatting submodule 35 .

[0063] The marking submodule 31 is used for marking defects of the plate.

[0064] The marking submodule 31 calculates and obtains the acquisition parameters according to the camera resolution and the required marking size of the plate. The camera in this embodiment is a linear array camera with a resolution of 8k (8192 pixels per line). The length of the plate in this embodiment is marked as 1280mm. The acquisition parameters obtained by calculation are set to 8192 acquisition row pixels, 16000 acquisition column pixels, and 12500HZ acquisition frequency.

[0065] The model training submodule 32 is used to obtain data from the annotation submodule 31, and the model training submodule 32 performs model training according to the annotated plate defects to obtain a recognition model. The model training submodule 32 can train a deep learning model and convert it into an ONNX model.

[0066] After the labeling submodule 31 imports the training set image source, it labels different types of defects with rectangles or polygons and saves the labeling information. The labeling data is used to form the types and locations of wood defects, and the training set, verification set, and test set are divided according to the ratio of 7:2:1.

[0067] The model training submodule 32 imports the data of the labeling submodule 31 into the training model, and the model training submodule 32 performs model training by setting the training parameters of the number of iterations, basic learning rate, patch size, model capability, angle capability, and whether to prune to obtain a recognition model.

[0068] The preprocessing submodule 33 is used to acquire the plate image taken by the camera, and the preprocessing submodule 33 preprocesses the plate image to obtain a preprocessed image.

[0069] The preprocessing submodule 33 segments the board image, and the preprocessing submodule 33 collects two boards through an image source, and processes the two objects in one image source into two independent images to be processed with wooden boards as units to obtain a preprocessed image, and the two independent images to be processed have

[0070] The analysis submodule 34 is used for acquiring data from the preprocessing submodule 33 , and the analysis submodule 34 analyzes the preprocessing image through a recognition model to obtain a recognition result of the plate defect.

[0071] After the analysis submodule 34 sets the information of the calculation threshold, the region of interest, and the number of calculated defects for the recognition model, the pre-processed image is input into the recognition model.

[0072] The recognition model obtains defect information such as the type, area, quantity, angle, position, and confidence level of each marked defect by extraction in the preprocessed image. The analysis submodule 34 performs weight analysis based on the defect information to obtain information on the plate grade, and the analysis submodule 34 sends the plate grade information to the control module 4, so that the servo sorting electric cylinder 16 receives the plate grade information from the control module 4 to sort the plates according to different grades.

[0073] The data formatting submodule 35 is used for acquiring data from the analyzing submodule 34 , and the data formatting submodule 35 integrates the defect information into a data packet that can be recognized by the executing module 1 .

[0074] The data formatting submodule 35 sends the data packet to the control module 4 , and the control module 4 sends the data packet to the execution module 1 through the TCP / IP protocol to input a control signal to the execution module 1 .

[0075] The analysis submodule 34 of this embodiment can analyze and digitize the recognition results of the model solution, integrate the recognition results of various defects through the data formatting submodule 35 and format them into an agreed communication format to facilitate notification of the sorting device, and the third-party communication module is used to connect to the sorting device and issue sorting notifications.

[0076] In this embodiment, after the conveying servo roller 14 flattens the plate, the camera can obtain higher quality images. At the same time, the camera does not need to stop the conveyor line during the acquisition process, which further improves the sorting efficiency, and can collect images of the front and back of the plate in a more limited space. The image is affine transformed by the preprocessing submodule 33 to improve the defect recognition accuracy of the analysis submodule 34. The analysis submodule 34 standard quantifies the defect types and plate quality grades, so that the plates are accurately graded and sorted, overcoming the disadvantage of being too subjective in manually identifying and sorting plates based on experience. At the same time, the present invention can be applied to relatively large plates. There is no need to transport and move the plates during the identification and sorting process, which effectively improves the efficiency of graded sorting. In addition, there is no need to frequently turn over the plates during the graded sorting process to avoid damage to the plates.

[0077] like Figure 3 As shown, the recognition process of the image processing module 3 in this embodiment is:

[0078] S1. Acquire images containing wood defects on site as a data set, wherein the wood defects include dead knots, live knots, bark, cracks, notches, and dry scars.

[0079] S2. Label the data to form the types and locations of wood defects, and divide them into training set, validation set, and test set in a ratio of 7:2:1;

[0080] S3. Import annotation information for model training: Use a graphics card to train the model. In the embodiment, one RTX3060, one RTX4070, and one RTX A4000 are used for model training respectively and the model parameters with the highest recognition accuracy, that is, the optimal model parameters, are saved.

[0081] S4. Segment the image according to the actual application scenario, and use one image source to simultaneously capture two wooden boards. In this process, the two objects in one image source are processed into two independent images to be processed with wooden boards as units to obtain a preprocessed image.

[0082] 5. Substitute the preprocessed image into the recognition model for solution and output the basic attribute information of the solution. It is necessary to set the model file path, solution threshold, area of ​​interest, number of solution defects and other information.

[0083] S6. Concretely analyze the basic attribute information calculated in the recognition model, and extract the defect information of each marked defect type, area size, quantity, angle, location, and confidence level.

[0084] S7. Integrate the information such as the type, area, quantity, angle, location, confidence level, etc. of each defect extracted in step S6 into a data packet that can be understood by the sorting device in a specific format;

[0085] S8. Send the data packets integrated in step S7 to the execution module 1 via the TCP / IP free protocol for hierarchical sorting.

Claims

1. A control system for grading and picking peeled sheets based on visual recognition, characterized in that: It comprises an execution module (1), an image acquisition module (2), an image processing module (3) and a control module (4), The execution module (1) comprises a hydraulic loading and unloading unit (11), a frequency conversion material supply and receiving unit (12), the crank plate removal unit (13), a conveying servo roller (14), a frequency conversion conveying belt (15), a servo sorting electric cylinder (16), and a sensor unit (17). The hydraulic loading and unloading unit (11) is used for loading unprocessed plates and unloading sorted plates in the variable frequency supply and receiving unit (12); the variable frequency supply and receiving unit (12) is used for collecting plates in the hydraulic loading and unloading unit (11) and collecting sorted plates in the servo sorting electric cylinder (16); the crank plate disassembly unit (13) is used for disassembling plates in the variable frequency supply and receiving unit (12); the conveying servo roller (14) is used for flattening plates in the crank plate disassembly unit (13); the variable frequency conveying belt (15) is used for conveying flattened plates; the servo sorting electric cylinder (16) is used for sorting plates on the variable frequency conveying belt (15) to the variable frequency supply and receiving unit (12); the sensor unit (17) is used for monitoring the plates in each unit through sensors to obtain status data of equipment and plates in each unit; The image acquisition module (2) is used to acquire the image of the plate after being flattened by the conveying servo roller (14); The image processing module (3) is used for acquiring data from the image acquisition module (2) to analyze the quality of the flattened plate; The control module (4) is used for data processing and action control of the execution module (1), the image acquisition module (2) and the image processing module (3), and the control module (4) controls the sorting of the servo sorting electric cylinder (16) according to the quality of the plate material through the data of the image processing module (3).

2. According to claim 1, a control system for grading and sorting peeled sheets based on visual recognition, characterized in that: After receiving the plate, the conveying servo roller (14) obtains the position information of the plate through an encoder and a sensor. After the plate reaches a preset position, the control module (4) sends a capture signal to the image capture module (2) so that the image capture module (2) captures an image of the plate through a camera.

3. According to claim 2, a control system for grading and sorting peeled sheets based on visual recognition, characterized in that: The image processing module (3) comprises a labeling submodule (31), a model training submodule (32), a preprocessing submodule (33) and an analysis submodule (34). The marking submodule (31) is used for marking defects of the plate; The model training submodule (32) is used for acquiring data from the marking submodule (31), and the model training submodule (32) performs model training according to the marked plate defects to obtain a recognition model; The preprocessing submodule (33) is used to acquire the plate image taken by the camera, and the preprocessing submodule (33) preprocesses the plate image to obtain a preprocessed image; The analysis submodule (34) is used for acquiring data from the preprocessing submodule (33), and the analysis submodule (34) analyzes the preprocessing image through the recognition model to obtain a recognition result of a plate defect.

4. According to claim 3, a visual recognition-based rotary cut sheet grading and sorting control system is characterized in that: The marking submodule (31) calculates and obtains acquisition parameters according to the resolution of the camera and the marking dimensions required for the plate.

5. According to claim 3, a visual recognition-based rotary cut sheet grading and sorting control system is characterized in that: The model training submodule (32) imports the data of the labeling submodule (31) into the training model, and the model training submodule (32) performs model training by setting the training parameters of the number of iterations, basic learning rate, patch size, model capability, angle capability, and whether to prune, so as to obtain a recognition model.

6. According to claim 3, a control system for grading and sorting peeled sheets based on visual recognition, characterized in that: The preprocessing submodule (33) segments the board image, and the preprocessing submodule (33) collects two boards through an image source, and processes two objects in one image source into two independent images to be processed with wooden boards as units, so as to obtain the preprocessed image.

7. According to claim 3, a control system for grading and sorting peeled sheets based on visual recognition, characterized in that: After the analysis submodule (34) sets the information of the calculation threshold, the region of interest, and the number of calculated defects for the recognition model, the pre-processed image is input into the recognition model.

8. According to claim 7, a control system for grading and sorting peeled sheets based on visual recognition, characterized in that: The recognition model extracts defect information of each marked defect type, area size, quantity, angle, position, and confidence level from the preprocessed image, and the analysis submodule (34) performs weight analysis based on the defect information to obtain information on the plate grade. The analysis submodule (34) sends the plate grade information to the control module (4) so ​​that the servo sorting electric cylinder (16) receives the plate grade information from the control module (4) to sort the plates according to different grades.

9. The control system for grading and sorting peeled sheet materials based on visual recognition according to claim 8, characterized in that: The image processing module (3) further comprises a data formatting submodule (35), wherein the data formatting submodule (35) is used for acquiring data of the analysis submodule (34), and the data formatting submodule (35) integrates the defect information into a data packet that can be recognized by the execution module (1).

10. The control system for grading and sorting peeled sheet materials based on visual recognition according to claim 9, characterized in that: The data formatting submodule (35) sends the data packet to the control module (4), and the control module sends the data packet to the execution module (1) through the TCP / IP protocol to input a control signal to the execution module (1).

Images