A waste sorting and processing device and method

Abstract

This invention relates to a waste sorting and processing device and method. The invention pertains to the field of waste sorting technology. The device includes: a top cover with a waste inlet; a container located below the top cover, supporting the top cover; sorting bins located at the bottom of the container; a sorting module for receiving waste entering through the inlet and sequentially disposing of the waste into the corresponding sorting bins according to its type; and a control module electrically connected to the sorting module, which identifies the type of waste and controls the sorting module to sequentially dispose of the waste into the corresponding sorting bins according to its type. This application enables the sequential sorting of multiple types of waste input at once, significantly improving the accuracy and efficiency of waste sorting.

Description

Technical Field

[0001] This invention relates to the field of waste treatment technology, and in particular to a waste sorting and treatment device and method. Background Technology

[0002] Uncollected garbage not only affects the cleanliness and aesthetics of the environment, but also easily breeds bacteria and produces harmful gases, endangering human health. Therefore, developing an automated, efficient, and accurate garbage sorting technology has become particularly important.

[0003] Currently, some waste sorting methods require manual input of waste into sorting devices, resulting in low sorting efficiency. Other methods allow for the input of multiple types of waste simultaneously and use traditional visual methods, such as machine vision technology, to identify the image information of the waste. However, traditional visual methods are limited by the variety of datasets and the size of the model, making them less effective at waste sorting in real-world scenarios. They fail to classify waste efficiently and accurately, and if multiple types of waste are identified and sorted simultaneously, errors are likely to occur, reducing the accuracy of waste sorting. Summary of the Invention

[0004] Therefore, the technical problem to be solved by this invention is to overcome the low efficiency and low accuracy of waste sorting detection in the prior art. To solve the above technical problems, this invention provides a waste sorting and processing device and method, which can sequentially identify and classify multiple pieces of waste that are put in at one time, achieving efficient and accurate sorting and processing of large quantities of waste.

[0005] To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0006] In a first aspect, the present invention provides a waste sorting and processing device, comprising: a top cover having a waste inlet; a housing disposed below the top cover, the housing supporting the top cover; a sorting waste bin disposed at the bottom of the housing; a sorting module disposed within the housing and between the top cover and the sorting waste bin; the sorting module receiving waste to be sorted from the inlet and sequentially disposing of the waste into the corresponding sorting waste bin according to its type; and a control module electrically connected to the sorting module, the control module identifying the type of waste and controlling the sorting module to sequentially dispose of the waste into the corresponding sorting waste bin according to its type.

[0007] In one embodiment of the present invention, the sorting module includes: a conveying component disposed below the feeding port; the conveying component is used to receive the waste and convey the waste sequentially according to the order of disposal; a sorting component disposed between the conveying component and the sorting waste bins; the sorting component is used to receive the waste conveyed by the conveying component sequentially and dispose of the waste into the corresponding sorting waste bins according to the type of waste.

[0008] In one embodiment of the present invention, the conveying assembly includes: a first conveying assembly, a second conveying assembly, and a second matrix fiber optic sensor; the first conveying assembly is disposed below the feeding port and is used to receive waste entering from the feeding port and sequentially convey the waste to the second conveying assembly; the second conveying assembly is disposed below the first conveying assembly and between the first conveying assembly and the sorting assembly, and is used to sequentially receive the waste conveyed by the first conveying assembly and sequentially convey the waste to the sorting assembly; the second matrix fiber optic sensor is disposed inside the housing and between the first conveying assembly and the second conveying assembly; the second matrix fiber optic sensor is used to detect whether waste is passing between the first conveying assembly and the second conveying assembly; wherein, when only the second matrix fiber optic sensor detects the waste, the first conveying assembly stops conveying.

[0009] In one embodiment of the present invention, the first conveying component includes: a primary conveyor belt located below the feeding port and horizontally disposed within the container; a first preset distance is provided between the end of the primary conveyor belt away from the feeding port and the inner wall of the container, the first preset distance being used for the passage of the waste; a first matrix fiber optic sensor disposed at the end of the primary conveyor belt away from the feeding port; the first matrix fiber optic sensor is used to detect whether the waste is conveyed to the end of the primary conveyor belt away from the feeding port; wherein, when the first matrix fiber optic sensor does not detect the waste, the conveying direction of the primary conveyor belt is from the end near the feeding port to the end away from the feeding port; when only the first matrix fiber optic sensor detects the waste, the primary conveyor belt reduces its conveying speed; when the first matrix fiber optic sensor and the second matrix fiber optic sensor simultaneously detect the waste, the conveying direction of the primary conveyor belt changes to from the end away from the feeding port to the end near the feeding port, until the first matrix fiber optic sensor fails to detect the waste.

[0010] In one embodiment of the present invention, the second conveying component includes: a secondary conveyor belt located below the primary conveyor belt, horizontally disposed within the box, and disposed at the end of the primary conveyor belt away from the feeding port; a second preset distance is provided between the end of the secondary conveyor belt near the primary conveyor belt and the inner wall of the box, the second preset distance being used for the passage of the waste; the secondary conveyor belt is used to receive the waste conveyed by the primary conveyor belt and sequentially convey the waste to the sorting component; a third matrix fiber optic sensor located at the end of the secondary conveyor belt near the primary conveyor belt; the third matrix fiber optic sensor is used to detect whether the waste has been conveyed to the end of the secondary conveyor belt near the primary conveyor belt; wherein, when the third matrix fiber optic sensor does not detect the waste, the conveying direction of the secondary conveyor belt is from the end away from the primary conveyor belt to the end near the primary conveyor belt; when the third matrix fiber optic sensor detects the waste, the secondary conveyor belt reduces its conveying speed until the third matrix fiber optic sensor fails to detect the waste.

[0011] In one embodiment of the present invention, the sorting assembly includes: a tray located below the secondary conveyor belt and disposed at one end of the secondary conveyor belt near the primary conveyor belt; the tray is used to receive the waste conveyed by the secondary conveyor belt; a servo motor located between the tray and the sorting waste bin, and rotatably connected to the tray; the servo motor is used to drive the tray to dispose of the waste into the sorting waste bin; a fourth matrix fiber optic sensor located inside the housing and disposed between the tray and the secondary conveyor belt; the fourth matrix fiber optic sensor is used to detect whether waste is passing between the secondary conveyor belt and the tray; wherein, when the fourth matrix fiber optic sensor detects whether waste is passing between the secondary conveyor belt and the tray, the fourth matrix fiber optic sensor detects whether waste is passing between the secondary conveyor belt and the tray. When the matrix fiber optic sensor detects the waste, the secondary conveyor belt stops conveying; when the third and fourth matrix fiber optic sensors simultaneously detect the waste, the secondary conveyor belt changes its conveying direction from one end closer to the primary conveyor belt to the end farther away from the primary conveyor belt, until the third matrix fiber optic sensor fails to detect the waste; a camera is installed inside the container; the camera is used to capture images of the waste in the tray; wherein, the control module is used to identify the type of waste based on the image, and control the sorting component to place the waste into the corresponding classified waste bin according to the type of waste.

[0012] In one embodiment of the present invention, a full-load detection sensor is provided inside the sorting trash can; the full-load detection sensor is used to detect whether the volume ratio of the trash in the sorting trash can to the volume of the sorting trash can exceeds a preset ratio; wherein, when the full-load detection sensor detects that the volume ratio of the trash in the sorting trash can to the volume of the sorting trash can exceeds the preset ratio, the full-load detection sensor sends a full-load signal to the control module, and the control module issues a full-load warning; the sorting trash can is also provided with a trash compression device, which is used to compress the trash in the sorting trash can; the trash compression device further includes a puncture rebound device, which is used to puncture the trash.

[0013] Secondly, the present invention provides a waste sorting and processing method, applied to the waste sorting and processing device described in the first aspect above. The method includes: acquiring image data of waste to be sorted; preprocessing the image data to obtain processed image data; establishing a waste detection model; inputting the processed image data into the waste detection model to obtain the waste type; controlling the sorting component to place the waste into the corresponding sorting waste bin according to the waste type; and issuing a full-load warning if a full-load signal is obtained when the volume ratio of the waste in the sorting waste bin to the volume of the sorting waste bin exceeds a preset ratio.

[0014] In one embodiment of the present invention, the preprocessing includes: scaling, brightness normalization, removal of redundant parts, and edge filling.

[0015] In one embodiment of the present invention, the specific steps for establishing a spam detection model include: establishing an initial YOLOv5-serial model; the YOLOv5 model includes: a Backbone module, a Neck module, a Head module, and a serial communication module; wherein the serial communication module is used to receive the output data of the Head module and transmit the output data via communication; obtaining a target dataset, optimizing the target dataset using data augmentation techniques to obtain an optimized dataset; wherein the data augmentation techniques include: random image rotation, random scaling, random color jitter, perspective transformation, and random cropping; training the initial YOLOv5-serial model by minimizing a loss function based on the optimized dataset, so that the initial YOLOv5-serial model converges, and using the converged initial YOLOv5-serial model as the spam detection model.

[0016] Compared with the prior art, the above-described technical solution of the present invention has the following advantages:

[0017] The waste sorting and processing device and method described in this invention sorts multiple different types of waste into the system sequentially using a sorting module. Through the cooperation of the sorting module and the control module, the system accurately identifies and classifies multiple quantities of waste sequentially, avoiding the possibility of misclassification when multiple types of waste are present. This improves the accuracy and efficiency of waste sorting. Furthermore, the waste detection model trained using a self-collected target dataset exhibits extremely high recognition accuracy, further enhancing the efficiency and accuracy of waste sorting. Attached Figure Description

[0018] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein...

[0019] Figure 1 This is a schematic diagram of the overall structure of a waste sorting and processing device provided in a preferred embodiment of the present invention;

[0020] Figure 2 for Figure 1 A schematic diagram of the conveying component of the waste sorting and processing device;

[0021] Figure 3 for Figure 1 A schematic diagram of the sorting components of the waste sorting and processing device;

[0022] Figure 4 for Figure 1 A schematic diagram of the structure of the waste sorting bins in the waste sorting and processing device;

[0023] Figure 5 for Figure 1 A schematic diagram of the waste compression device in the aforementioned waste sorting and processing apparatus;

[0024] Figure 6 for Figure 1 A schematic diagram of the puncture rebound device of the aforementioned waste sorting and processing device;

[0025] Figure 7 for Figure 1 A schematic diagram of the high-torque telescopic motor of the aforementioned waste sorting and processing device;

[0026] Figure 8 for Figure 1 A side view of the waste sorting and processing device;

[0027] Figure 9 for Figure 1 A schematic diagram of the microcontroller control module of the aforementioned waste sorting and processing device;

[0028] Figure 10 A flowchart of a waste sorting and treatment method provided in another preferred embodiment of the present invention;

[0029] Figure 11 for Figure 10 The flowchart for establishing the detection model of the aforementioned waste sorting method.

[0030] Explanation of reference numerals in the accompanying drawings: 1. Top cover; 11. Feeding port; 12. Visualization module; 2. Box body; 3. Conveying assembly; 311. Primary conveyor belt; 312. First matrix fiber optic sensor; 313. Horizontal support; 32. Second matrix fiber optic sensor; 331. Secondary conveyor belt; 332. Third matrix fiber optic sensor; 333. Fixed support; 4. Sorting assembly; 41. Pallet; 42. Servo motor; 43. Fourth matrix fiber optic sensor; 44. Support frame; 45. Camera; 5. Classified waste bin; 51. Full load detection sensor; 52. Waste compression device; 521. Puncture rebound device; 53. High torque telescopic motor. Specific Implementation

[0031] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0032] Reference Figures 1 to 8 As shown, in a first aspect, a preferred embodiment of the present invention provides a waste sorting and processing device, hereinafter referred to as "this device". This device includes a top cover 1, a housing 2, sorting waste bins 5, and a sorting module. The top cover 1 is disposed on the top of the housing 2, and the sorting waste bins 5 are disposed on the bottom of the housing 2. A waste inlet 11 is provided on the top cover 1. The sorting module is disposed inside the housing 2, located between the top cover 1 and the sorting waste bins 5. The sorting module receives the waste to be sorted entering through the inlet 11 and disposes of the waste into the corresponding sorting waste bins 5 according to its type. To automate the entire process, this application includes a control module electrically connected to the sorting module. The control module identifies the type of waste and controls the sorting module to dispose of the waste into the corresponding sorting waste bins 5 according to its type. In this embodiment, refer to... Figure 9 As shown, the control module adopts a single-chip microcomputer control module, which ensures the efficiency and accuracy of the entire waste sorting process. In accordance with the current waste sorting system in China and the "Classification Marks for Domestic Waste" issued by the Ministry of Housing and Urban-Rural Development in 2019, this embodiment sets the number of waste bins 5 to 4, namely, a waste bin for recyclables, a waste bin for kitchen waste, a waste bin for other waste, and a waste bin for hazardous waste.

[0033] To make this device more practical in actual use, in one specific embodiment, the device is equipped with a visualization module 12, which is mounted on the top cover 1. The visualization module 12 includes a 5.8G radar sensing module and a display screen. The 5.8G radar sensing module uses the Doppler effect to achieve real-time monitoring of pedestrians. The output terminal of the 5.8G radar sensing module is electrically connected to the microcontroller control module. When someone approaches, the 5.8G radar sensing module sends a high-level signal to the microcontroller control module, which in turn controls the display screen to start playing a video. The display screen then enters the garbage sorting interface, at which point the device enters the garbage sorting working state. When no one is passing by, the display screen plays garbage sorting promotional videos in a loop, and the device is in a standby state when it is not in operation.

[0034] The sorting module includes a conveying component 3 and a sorting component 4. The conveying component 3 is located below the feeding port 11. The conveying component is used to receive the waste to be sorted entering from the feeding port 11 and to convey the waste to the sorting component 4 in sequence. The sorting component 4 is located between the conveying component 3 and the sorting waste bins 5. The sorting component 4 is used to receive the waste conveyed by the conveying component in sequence and to put the waste into the sorting waste bins 5 in sequence according to the type of waste, thereby realizing the sequential and accurate sorting of multiple types of waste.

[0035] Specifically, the conveying component 3 includes a first conveying component, a second conveying component, and a second matrix fiber optic sensor 32. The first conveying component is located below the feeding port 11. The waste to be sorted enters the first conveying component from the feeding port 11, and the first conveying component sequentially conveys the waste to the second conveying component. The second conveying component is located below the first conveying component. The second conveying component sequentially receives the waste conveyed by the first conveying component and then sequentially conveys it to the sorting component 4. The second matrix fiber optic sensor 32 is located between the first and second conveying components. In this embodiment, the second matrix fiber optic sensor 32 is located on the inner wall of the housing 2. The second matrix fiber optic sensor 32 is used to detect whether there is waste passing between the first and second conveying components. If there is waste passing through, it indicates that there is waste being conveyed from the first conveying component to the second conveying component. At this time, in order to avoid multiple pieces of waste falling from the first conveying component to the second conveying component at the same time, when only the second matrix fiber optic sensor 32 detects waste, the first conveying component stops working and no longer conveys.

[0036] The first conveying component includes a primary conveyor and a first matrix fiber optic sensor 312. The primary conveyor belt 31 is horizontally positioned below the feeding port 11. In this embodiment, a horizontal support 313 is provided inside the housing 2, and the primary conveyor belt is positioned on the horizontal support 313. Specifically, to ensure that the waste falls from the primary conveyor belt 31 to the second conveying component, the primary conveyor belt 31 is positioned at one end away from the feeding port 11, i.e., closer to the second conveying component, and a certain distance is maintained between it and the inner wall of the housing 2. This distance allows the waste to fall smoothly from the primary conveyor belt 31 to the second conveying component. To detect the position of the waste conveyed in the primary conveyor belt 31, this application includes a primary conveyor... A first matrix fiber optic sensor 312 is installed at the end of the primary conveyor belt 31 furthest from the feeding port 11. The first matrix fiber optic sensor 312 is used to detect whether waste has been conveyed to the end of the primary conveyor belt 31 furthest from the feeding port 11. When the first matrix fiber optic sensor 312 does not detect waste, the primary conveyor belt 31 operates normally, with the conveying direction from the end near the feeding port 11 to the end furthest from the feeding port 11. In this embodiment, the conveying speed of the primary conveyor belt 31 during normal operation is 100 rpm. When the first matrix fiber optic sensor 312 detects waste, it transmits the detection signal to the control module, which then controls the conveyor belt to... The primary conveyor belt 31 slows down, but does not change its conveying direction. In this embodiment, the conveying speed of the primary conveyor belt 31 is reduced to 25 rpm. When the first matrix fiber optic sensor 312 detects a signal, it indicates that the garbage has approached the edge of the primary conveyor belt 31. The slowing down at this time is to avoid multiple pieces of garbage falling onto the second conveyor assembly at the same time, which would affect the accuracy of subsequent garbage sorting. The garbage on the primary conveyor belt 31 at the end away from the feeding port 11 continues to be conveyed by the primary conveyor belt 31 until it falls onto the second conveyor assembly. At this time, the second matrix fiber optic sensor 32 located between the first and second conveyor assemblies will detect the presence of garbage. The second matrix fiber optic sensor 32 transmits the detection signal to the control module, which then controls the primary conveyor belt 31 to stop operating. If the first matrix fiber optic sensor 312 can still detect garbage within its detection range, the control module will simultaneously receive the detection signals from both the first and second matrix fiber optic sensors 312. The control module will then control the primary conveyor belt 31 to change its conveying direction to the opposite direction, i.e., from the end furthest from the feed inlet 11 to the end closest to the feed inlet 11, until the first matrix fiber optic sensor 312 no longer detects garbage, thus preventing garbage from continuously falling into the second conveying assembly.

[0037] The second conveying component includes a secondary conveyor belt 331 and a third matrix fiber optic sensor 332. The secondary conveyor belt 331 is horizontally positioned below the primary conveyor belt 31. In this embodiment, a fixed support 333 is provided inside the housing 2 below the horizontal support 313, and the secondary conveyor belt 331 is fixed within the fixed support 333. Similarly, to allow waste to fall from the secondary conveyor belt 331 to the sorting component 4, a certain distance is provided between the end of the secondary conveyor belt 331 near the primary conveyor belt 31 and the inner wall of the housing 2. This distance allows waste to fall smoothly from the secondary conveyor belt 331 to the sorting component 4. To detect the position of waste conveyed in the secondary conveyor belt 331, this application provides a third matrix fiber optic sensor 332 at the end of the secondary conveyor belt 331 near the primary conveyor belt 31. The third matrix fiber optic sensor 332 is used for... The system detects whether waste has been conveyed to the end of the secondary conveyor belt 331 that is close to the primary conveyor belt 31. When waste falls from the primary conveyor belt 31 to the secondary conveyor belt 331 but is not detected by the third matrix fiber optic sensor, the conveying direction of the secondary conveyor belt 331 is from the direction away from the primary conveyor belt 31 to the direction close to the primary conveyor belt 31. In this embodiment, the conveying speed of the secondary conveyor belt is 60 rpm. When waste is conveyed to the detection range of the third matrix fiber optic sensor 332, it indicates that waste is about to fall from the secondary conveyor belt 331 to the sorting component 4. To prevent multiple pieces of waste from falling at the same time, after the control module obtains the detection signal of the third matrix fiber optic sensor 332, it controls the secondary conveyor belt 331 to reduce its speed. In this embodiment, the speed is reduced to 25 rpm until waste falls to the sorting component 4.

[0038] The sorting component 4 includes a tray 41, a servo motor 42, a fourth matrix fiber optic sensor 43, and a camera 45. The tray 41 is rotatably connected to the servo motor 42. The combination of the tray 41 and the servo motor 42 is positioned between the sorting bin 5 and the secondary conveyor belt 331, and is located at one end of the secondary conveyor belt 331 near the primary conveyor belt 31. In this embodiment, the servo motor 42 is mounted on the inner wall of the bottom of the housing 2 via a support frame 44. The tray 41 is used to receive garbage falling from the secondary conveyor belt 331. The camera 45 is used to capture images of the garbage on the tray 41 and transmit the garbage images to the control module. The control module acquires the garbage image data and identifies the type of garbage based on the garbage image data. In this embodiment, to obtain the most accurate image data, the camera 45 is positioned directly above the tray 41. The servo motor 42 is used to drive the tray 41 to dispose of the garbage according to its type. In the classified trash can 5; the fourth matrix fiber optic sensor 43 is set between the secondary conveyor belt 331 and the tray 41. In this embodiment, the fourth matrix fiber optic sensor is set on the inner wall of the box 2; the fourth matrix fiber optic sensor 43 is used to detect whether trash has fallen from the secondary conveyor belt 331 onto the tray 41. When the fourth matrix fiber optic sensor detects that trash has fallen, it transmits the detection signal to the control module, and the control module controls the secondary conveyor belt 331 to stop operating immediately; if the control module simultaneously obtains the detection signals of the third matrix fiber optic sensor 332 and the fourth matrix fiber optic sensor 43, it indicates that there is still a possibility that trash may fall from the secondary conveyor belt 331 onto the tray 41. At this time, the control module controls the secondary conveyor belt 331 to change the conveying direction, and the conveying direction is changed to conveying in a direction away from the primary conveyor belt 31, until the third matrix fiber optic sensor 332 no longer detects trash.

[0039] This application includes a waste compression device 52 driven by a high-torque telescopic motor 53 in the recyclable waste bin. In this embodiment, the high-torque telescopic motor 53 is located at the bottom of one of the sorting waste bins 5, and the waste compression device 52 is used to compress recyclable waste. The waste compression device 52 also includes a puncture and rebound device 521, which is used to puncture sealed bottles and cans to facilitate air release and compression. This application also includes a full-load detection sensor 51 in the sorting waste bin 5. The full-load detection sensor 51 is used to detect whether the volume ratio of waste in the sorting waste bin 5 to the volume of the sorting waste bin 5 exceeds a preset ratio. In this embodiment, a full-load detection sensor 51 is installed at 3 / 4 of the height of each waste bin. When the stacked height of the waste exceeds 3 / 4 of the height of the sorting waste bin 5, the distance measured by the full-load detection sensor 51 changes, and a full-load signal is sent to the microcontroller control module to control the display screen to issue a full-load reminder and display full-load information.

[0040] In actual operation, the entire workflow is as follows: The 5.8G radar sensing module uses the Doppler effect to achieve real-time detection of pedestrians and determine whether garbage has been disposed of. When no one approaches, the display screen on the top cover 1 will play garbage sorting promotional videos in a loop. At this time, the device is in standby mode. When someone approaches, the 5.8G radar sensing module will send a high-level signal to the microcontroller control module, which will then control the display screen to stop playing the video and enter the working state. In this embodiment, it is assumed that three pieces of garbage, A, B, and C, are disposed of at the same time. In operation, the primary conveyor belt 31 rotates at a constant speed of 100 rpm, while the secondary conveyor belt 331 remains stationary. Waste falls from the feeding port 11 onto the primary conveyor belt 31 and is driven by it to move towards the end of the secondary conveyor belt 331 until it blocks the first matrix fiber optic sensor 312. Upon detecting the waste, the first matrix fiber optic sensor 312 transmits a signal to the control module. Upon receiving the signal, the control module slows the speed of the primary conveyor belt 31 to 25 rpm, causing the waste to move slowly. When waste falls from the primary conveyor belt 31 onto the secondary conveyor belt... When conveyor belt 331 moves, it will be detected by the second matrix fiber optic sensor 32, which is located between the primary and secondary conveyor belts 31 and 331. At this time, the primary conveyor belt 31 stops operating. If the first matrix fiber optic sensor 312 still detects that there is still garbage within its detection range, the control module will receive the detection signals from both the first and second matrix fiber optic sensors 312. The control module will then control the primary conveyor belt 31 to move towards the end closer to the feeding port 11 until the first matrix fiber optic sensor 312 no longer detects any objects, preventing further garbage from falling. When trash falls onto the secondary conveyor belt 331, even if multiple pieces of trash fall simultaneously due to objective factors, they will be dispersed by the impact and transported one by one to the tray 41 for detection by the secondary conveyor belt 331, without affecting the workflow. The secondary conveyor belt 331 runs at a speed of 60 rpm towards the primary conveyor belt 31 until trash obstructs the third matrix fiber optic sensor 332. The third matrix fiber optic sensor 332 transmits the detection signal to the control module, which then controls the secondary conveyor belt 331 to move slowly at a speed of 25 rpm until trash falls into the tray 41. After the fourth matrix fiber optic sensor 43 detects the falling trash signal, it transmits the detection signal to the control module, which then controls the secondary conveyor belt 331 to stop immediately. If the third matrix fiber optic sensor 332 detects that there is still an object within its detection range, the control module will receive the detection signals from both the third matrix fiber optic sensor 332 and the fourth matrix fiber optic sensor 43. The control module then controls the secondary conveyor belt 331 to move away from the primary conveyor belt 31 until the third matrix fiber optic sensor no longer detects a signal. At this point, the device has completed the task of sorting out a single piece of waste from multiple pieces of waste that were put in at the same time.Afterwards, conveyor component 3 stands by, waiting for the trash to be disposed of into the corresponding sorting bin 5. Once the trash is disposed of into the bin 5, the secondary conveyor belt 331 moves towards the primary conveyor belt 31 at a speed of 60 rpm for a certain period of time. This time is the time it takes for the secondary conveyor belt 331 to complete one revolution at 60 rpm. This operation is to verify the number of trash items that fell on the primary conveyor belt 31 in the previous cycle. If only one piece of trash fell, it will be sorted, and the secondary conveyor belt 331 will be empty. The primary conveyor belt 31 should continue moving to sort the remaining trash. If x (x>1) pieces of trash fell at once, it can be considered that the secondary conveyor belt 331 only sorted one piece of trash, leaving x-1 pieces of trash remaining. The secondary conveyor belt 331 needs to continue moving. The conveying process is verified by the third matrix fiber optic sensor 332 detecting trash during one revolution of the secondary conveyor belt 331. If the third matrix fiber optic sensor 332 does not detect trash, it indicates that only one piece of trash fell on the primary conveyor belt 31 in the previous cycle, and the secondary conveyor belt 331 is empty. The secondary conveyor belt 331 then stops, and the device returns to its initial state. Otherwise, it indicates that multiple pieces of trash fell on the primary conveyor belt 31 in the previous cycle, and trash still exists on the secondary conveyor belt 331. The secondary conveyor belt 331 will continue to execute the above logic in a loop until, in a certain loop, the third matrix fiber optic sensor 332 does not detect any trash signal, and the program returns to its initial state. After trash falls into the tray 41, the camera 45 captures an image of the trash. The control module acquires the image data and identifies the type of trash based on the image data. It then controls the servo motor 42 to place the trash on the tray 41 into the corresponding sorting bin 5, achieving sequential sorting of the trash. The entire process is efficient and accurate.

[0041] refer to Figure 10 , Figure 11 As shown, in a second aspect, another preferred embodiment of the present invention provides a waste sorting and processing method, applied to the sorting and processing apparatus described in the first aspect, comprising:

[0042] S1, acquire image data of the waste to be sorted, preprocess the image data, and obtain the processed image data;

[0043] The preprocessing methods mainly include scaling, brightness normalization, removing redundant parts, and edge filling. Scaling primarily scales the input image data to 640x640 pixels to fit the network structure's input requirements. Since the actual machine captures only the steering wheel and its surrounding area, the surrounding area is removed and filled with a solid color to prevent false recognition results from the surrounding area. In practical applications, different types of waste may cause inconsistent exposure of the camera 45, resulting in varying brightness in the input photos. Brightness normalization unifies the brightness of the input image data to a pre-set value. These preprocessing methods effectively improve the robustness of the model after the image data is input.

[0044] S2, Establish a waste detection model; specifically including:

[0045] S21, Establish the initial YOLOv5-serial model; the YOLOv5 model includes: Backbone module, Neck module, Head module and serial communication module; the serial communication module is used to receive the output data of the Head module and transmit the output data through communication.

[0046] YOLOv5 (You Only Look Once version 5) is the latest version of the YOLO (You Only Look Once) object detection algorithm. YOLOv5 is an object detection model used to identify and locate objects in videos or images.

[0047] The Backbone module uses the CSP-Darknet53 (Feature Extraction Network) architecture, which is based on the Darknet53 (Convolutional Neural Network) and applies the Cross Stage Partial (CSP) strategy to reduce the number of model parameters and enhance feature extraction efficiency. The CSP strategy effectively reduces redundant gradient information and improves the model's real-time detection speed by segmenting the base-level feature maps and merging them in the cross-stage hierarchy. The Backbone module includes convolutional layers (Conv), batch normalization (BN), and the SiLU activation function. Convolutional layers (Conv) are core components in deep learning for image processing, especially in computer vision tasks. They extract features from the input image through convolutional operations, which involve a small, learnable filter (or convolutional kernel) sliding across the input image, calculating the dot product between the filter and a local region of the image. Each convolutional kernel is responsible for extracting a specific feature from the image, such as edges, color, or texture. By stacking multiple convolutional layers, the network can learn a feature hierarchy from simple to complex. Batch Normalization (BN) is a technique used to accelerate the training of deep networks while also helping to reduce sensitivity to initial weights and prevent overfitting. It normalizes the layer inputs to a mean of 0 and a standard deviation of 1, thus stabilizing the network's learning process. BN layers are typically placed after convolutional layers and before activation functions. It also introduces two learnable parameters, a scaling factor and a shift, to maintain the network's expressive power. This treatment can reduce the so-called internal covariate shift problem, which occurs when the distribution of network layer inputs changes during training.

[0048] The Neck module employs SPPF (Spatial Pyramid Pooling Fast) and a modified PANet (Path Aggregation Network). SPPF significantly increases the receptive field without reducing network speed by aggregating input information and outputting a fixed-length output. The improved PANet introduces BottleNeckCSP (Bottle Crossing Phase SP) into its structure, further optimizing the information flow and contributing to accurate pixel-level localization. The Neck module includes Feature Pyramid Networks (FPN) and a Path Aggregation Network (PAN). The Head module is used for object localization and classification in a YOLOv5-based object detection system.

[0049] The Head module consists of multiple sub-modules that work together to generate accurate object detection results. The Head module includes three convolutional layers for predicting bounding box locations, scores, and object categories. Furthermore, this application uses SiLU (Sigmoid Linear Unit) and the Sigmoid function for activation functions. SiLU is applied to the convolutional operations in the hidden layers, while Sigmoid is used in the output layer. The SiLU activation function (Sigmoid Linear Unit) is a non-linear activation function that allows deep learning networks to learn complex feature representations. The SiLU function is expressed as:

[0050] ;

[0051] in, It is the Sigmoid function, represented as: , The SiLU function is the input The output of .

[0052] The main advantage of the SiLU function is its smooth, non-monotonic nature, which allows negative values ​​to pass through. This contrasts with the traditional ReLU (Rectified Linear Unit) activation function, which outputs zero for all negative values. This characteristic allows the SiLU activation function to provide better performance than ReLU in some cases because it allows the network to capture information from negative input values, thereby improving learning dynamics and the overall performance of the model.

[0053] The serial module is designed to output the output of the Head module via a hardware serial port. This module receives detection result data from the Head module and converts it into a specific format data frame for transmission via a standard serial communication interface. This application makes a series of improvements based on YOLOv5, including replacing the Focus layer to reduce parameters and computational load, and eliminating mesh sensitivity issues, thereby improving bounding box detection capabilities. The serial module is designed to send information to the serial port of the hardware device running the model, thus sending classification results to the main controller. First, the serial module obtains the recognition information obtained by the Head module from the model's output layer, including object identifiers, bounding box coordinates, class labels, confidence scores, etc. Then, the module encodes this information into a decimal data frame, converts it to hexadecimal, and outputs it to the serial port. The module's data frame structure is as follows:

[0054] 1. Start bit: A data frame begins with a specific start bit to identify the start of the data frame. In this case, it is 1.

[0055] 2. Data payload:

[0056] a) Object Identifier: A unique identifier for each detected object. It is a 16-bit standard UUID.

[0057] b) Bounding box coordinates: Contains the bounding box coordinates for each detected object. Here, these are the x-coordinate, y-coordinate, width, and height. All of the above data are 8-bit floating-point numbers.

[0058] c) Category Label: The category label of the detected object. This is a 4-digit integer.

[0059] d) Confidence score: The confidence score of the detection result. It is an 8-bit floating-point number.

[0060] 3. Checksum: A checksum or verification bit used to verify the integrity and correctness of the data frame during transmission. This data frame format uses CRC16 checksum.

[0061] 4. End bit: A newline character that marks the end of the data frame.

[0062] Regarding the loss function, this application employs the Binary Cross-Entropy (BCE) and Complete Cross-Union (CIoU) loss calculation methods, respectively, to calculate the loss for detected category, object identity, and location. This method effectively improves the model's accuracy and efficiency in identifying garbage. The Binary Cross-Entropy (BCE) loss is a loss function used for binary classification tasks. In object detection, it is used to calculate the difference between the predicted object category or object presence and the true label. Given a predicted value p (the probability of being predicted as positive) and a true label y (0 or 1), the BCE loss calculation formula is:

[0063]

[0064] The Complete Cross-Union (CIoU) loss is a bounding box regression loss function for object location estimation in object detection. It considers not only the overlap area between the predicted and ground truth boxes, but also the distance between their center points and their aspect ratio, providing more accurate gradient information. The formula for calculating CIoU loss is:

[0065] ;

[0066] Where IoU is the intersection-union ratio between the predicted box and the ground truth box; ρ(b,b^gt) is the Euclidean distance between the center b of the predicted box and the center b^gt of the ground truth box; c is the diagonal length of the smallest closed region containing the predicted box and the ground truth box; v measures the consistency of aspect ratio and is used to penalize shape deviation; α is a tradeoff coefficient used to balance the overlap area and shape difference.

[0067] S22, Obtain the target dataset, and optimize the target dataset using data augmentation techniques to obtain the optimized dataset; among which, the data augmentation techniques include: random image rotation, random scaling, random color jitter, perspective transformation, and random cropping;

[0068] The target dataset used in this application contains four major categories and 24 subcategories of common household waste, and data collection and location labeling were performed on these categories. The target dataset includes 5126 training images and 570 test images. The target dataset is shown in Table 1.

[0069] Table 1

[0070]

[0071] This application applies several data augmentation methods to the training and test sets of images. These methods include Rotation, Scaling, Color Jittering, Perspective Transformation, and Random Cropping. These methods consider the characteristics of garbage image data in waste sorting and the practical scenarios in actual implementation, effectively improving the recognition rate of the model trained on this dataset. Specifically, since garbage may be facing different directions during sorting, Rotation is applied to the dataset to improve reliability. When applying this method, a random angle is selected for the input image, and then the new position of each pixel after rotation is calculated based on the selected angle. Any blank areas that may appear after rotation are filled with pure white. Since the size of garbage in the image may vary during actual sorting, Scaling is used to improve the model's generalization ability. When applying this method, a scaling factor is first randomly generated between [0.8, 1.2], and then the image is scaled using the selected scaling factor. Bilinear interpolation is used during the scaling process. Finally, the image is adjusted back to 640*640 pixels. Since camera 45 may generate noise during actual classification, potential stains may cause reflections or blurry images, and changes in lighting conditions may lead to inconsistencies in image contrast, saturation, and brightness, this method is needed to improve the model's robustness. The application of this method consists of several independent image processing steps: 1. Adding simulated black-and-white and color noise to the image. The noise is implemented using a Gaussian noise generator; 2. Randomly selecting several stain patterns, including circular spots (simulating water droplets or dust), irregular stripes (simulating fingerprints), and relatively thick straight lines (simulating scratches), generating the coordinate range to be processed. Then, applying the stain effect to the image to be processed uses methods such as random non-uniform Gaussian blur, random brightness increases and decreases, and overlaying a gray-black irregular patch with semi-transparency to simulate stain effects as much as possible. Since the actual camera 45 installation angle may not be 90 degrees perpendicular to the garbage being classified, appropriate perspective transformation of the dataset can significantly improve the model's adaptability to different deployment environments. This step first generates a random trapezoid as the quadrilateral after perspective transformation. Then, it uses the perspective transformation function (getPerspectiveTransform) from OpenCV to obtain the perspective transformation matrix. Finally, it uses the perspective transformation function (warpPerspective) to apply the transformation to the image. Due to the actual characteristics of garbage, most garbage will have defects, so it is necessary to add random cropping blocks to the image to increase the model's ability to recognize incomplete targets.This step randomly generates several cropped square coordinate regions of 10-50 pixels in the image, and then fills these regions with a random solid color to mask the target features. The data augmentation method described effectively improves the quality of the dataset and greatly enhances the model's ability to cope with various situations.

[0072] S23. Based on the optimized dataset, train the initial YOLOv5-serial model by minimizing the loss function, so that the initial YOLOv5-serial model converges, and use the converged initial YOLOv5-serial model as the garbage detection model.

[0073] S3, input the processed image data into the garbage detection model to obtain the garbage type;

[0074] S4, according to the type of waste, the sorting component 4 will put the waste into the corresponding classified waste bin 5;

[0075] S5. If a full load signal is obtained when the volume ratio of the garbage in the classified garbage bin 5 to that in the classified garbage bin 5 exceeds a preset ratio, a full load warning will be issued.

[0076] In this embodiment, the parameters of each device are shown in Table 2, and the parameters involved in the design are shown in Table 3.

[0077] Table 2

[0078]

[0079] Table 3

[0080]

[0081] In this implementation, the accuracy rate of waste sorting is close to 100%, which greatly increases the accuracy of waste sorting.

[0082] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0084] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0085] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0086] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A waste sorting processing device, characterized in that, The utility model relates to a garbage classification device, comprising: a top cover with a garbage feeding port; a box body arranged below the top cover, the box body being used to support the top cover; a classified garbage can arranged at the bottom of the box body; a classification module arranged in the box body and between the top cover and the classified garbage can, the classification module being used to receive garbage to be classified from the feeding port and sequentially feed the garbage to the corresponding classified garbage can according to the type of the garbage; a control module electrically connected to the classification module, the control module being used to identify the type of the garbage and control the classification module to sequentially feed the garbage to the corresponding classified garbage can according to the type of the garbage; the classification module comprises: a conveying assembly arranged below the feeding port, the conveying assembly being used to receive the garbage and sequentially convey the garbage in the feeding order; a sorting assembly arranged between the conveying assembly and the classified garbage can, the sorting assembly being used to sequentially receive the garbage conveyed by the conveying assembly and feed the garbage to the corresponding classified garbage can according to the type of the garbage; the conveying assembly comprises a first conveying assembly, a second conveying assembly and a second matrix optical fiber sensor; the first conveying assembly is arranged below the feeding port, the first conveying assembly being used to receive the garbage from the feeding port and sequentially convey the garbage to the second conveying assembly; the second conveying assembly is arranged below the first conveying assembly and between the first conveying assembly and the sorting assembly, the second conveying assembly being used to sequentially receive the garbage conveyed by the first conveying assembly and sequentially convey the garbage to the sorting assembly; the second matrix optical fiber sensor is arranged in the box body and between the first conveying assembly and the second conveying assembly, the second matrix optical fiber sensor being used to detect whether the garbage passes between the first conveying assembly and the second conveying assembly; when only the second matrix optical fiber sensor detects the garbage, the first conveying assembly stops conveying; the first conveying assembly comprises: a first conveying belt arranged below the feeding port and horizontally in the box body, a first preset distance being arranged between the end of the first conveying belt away from the feeding port and the inner wall of the box body, the first preset distance being used to pass the garbage; a first matrix optical fiber sensor arranged at the end of the first conveying belt away from the feeding port, the first matrix optical fiber sensor being used to detect whether the garbage is conveyed to the end of the first conveying belt away from the feeding port; When the first matrix optical fiber sensor does not detect the garbage, the conveying direction of the first conveying belt is from one end close to the feeding port to one end away from the feeding port; when only the first matrix optical fiber sensor detects the garbage, the first conveying belt reduces the conveying speed; when the first matrix optical fiber sensor and the second matrix optical fiber sensor simultaneously detect the garbage, the conveying direction of the first conveying belt is changed to be from one end away from the feeding port to one end close to the feeding port until the first matrix optical fiber sensor fails to detect the garbage.

2. The garbage classification processing device according to claim 1, characterized in that: The second conveying assembly comprises: a second conveying belt located below the first conveying belt, horizontally arranged in the box, and arranged at one end of the first conveying belt away from the feeding port; a second preset distance is arranged between one end of the second conveying belt close to the first conveying belt and the inner wall of the box, and the second preset distance is used for passing the garbage; the second conveying belt is used for receiving the garbage conveyed by the first conveying belt and sequentially conveying the garbage to the sorting assembly; a third matrix optical fiber sensor located at one end of the second conveying belt close to the first conveying belt; the third matrix optical fiber sensor is used for detecting whether the garbage is conveyed to one end of the second conveying belt close to the first conveying belt; When the third matrix optical fiber sensor does not detect the garbage, the conveying direction of the second conveying belt is from one end away from the first conveying belt to one end close to the first conveying belt; when the third matrix optical fiber sensor detects the garbage, the second conveying belt reduces the conveying speed until the third matrix optical fiber sensor fails to detect the garbage.

3. The garbage classification processing device according to claim 2, characterized in that: The sorting assembly comprises: a tray located below the second conveying belt and arranged at one end of the second conveying belt close to the first conveying belt; the tray is used for receiving the garbage conveyed by the second conveying belt; a rudder located between the tray and the classified garbage can and rotationally connected with the tray; the rudder is used for driving the tray to put the garbage into the classified garbage can; a fourth matrix optical fiber sensor located in the box and arranged between the tray and the second conveying belt; the fourth matrix optical fiber sensor is used for detecting whether the garbage passes between the second conveying belt and the tray; when the fourth matrix optical fiber sensor detects the garbage, the second conveying belt stops conveying; when the third matrix optical fiber sensor and the fourth matrix optical fiber sensor simultaneously detect the garbage, the second conveying belt changes the conveying direction to be from one end close to the first conveying belt to one end away from the first conveying belt until the third matrix optical fiber sensor fails to detect the garbage; A camera is arranged in the box, and the camera is configured to capture an image of the garbage in the tray. The control module is configured to identify the type of garbage according to the image, and control the sorting assembly to drop the garbage into the corresponding classified garbage can according to the type of garbage.

4. The garbage classification processing device according to claim 1, characterized in that: A full load detection sensor is arranged in the classified garbage can, and the full load detection sensor is configured to detect whether the volume ratio of the garbage in the classified garbage can to the classified garbage can exceeds a preset ratio. When the full load detection sensor detects that the volume ratio of the garbage in the classified garbage can to the classified garbage can exceeds the preset ratio, the full load detection sensor sends a full load signal to the control module, and the control module performs a full load warning. The classified garbage can further comprises a garbage compression device configured to compress the garbage in the classified garbage can, and the garbage compression device further comprises a piercing and rebounding device configured to pierce the garbage. 5.A waste sorting processing method applied to the sorting processing device according to any one of claims 1 to 4, characterized in that, The method comprises: acquiring image data of garbage to be classified, pre-processing the image data to obtain processed image data; establishing a garbage detection model; inputting the processed image data into the garbage detection model to obtain a garbage type; controlling a sorting assembly to drop the garbage into a corresponding classified garbage can according to the garbage type; if a full load signal is acquired when the volume ratio of the garbage in the classified garbage can to the classified garbage can exceeds a preset ratio, a full load warning is issued.

6. The garbage classification processing method according to claim 5, characterized in that: the pre-processing comprises scaling, brightness normalization, removing redundant parts, and edge padding.

7. The garbage classification processing method according to claim 6, characterized in that, The specific steps for establishing the garbage detection model comprise: establishing an initial YOLOv5-serial model, wherein the YOLOv5 model comprises a Backbone module, a Neck module, a Head module, and a serial communication module, and the serial communication module is configured to receive output data of the Head module and perform communication transmission of the output data; acquiring a target data set, and optimizing the target data set by using a data enhancement technique to obtain an optimized data set, wherein the data enhancement technique comprises random image rotation, random scaling, random color jitter, perspective transformation, and random cropping; training the initial YOLOv5-serial model by minimizing a loss function according to the optimized data set, so that the initial YOLOv5-serial model converges, and the converged initial YOLOv5-serial model is used as a garbage detection model.

Images