Automatic sorting equipment for recyclable household garbage

Abstract

The application discloses a kind of automatic sorting equipment of recyclable domestic waste, it is related to solid waste recycling field, including the feeding module of sequentially arranged according to sorting order, monocular sorting magnetic metal module, air separation module, screening module, binocular sorting object module, monocular sorting magnetic metal module is used to screen the magnetic metal in garbage;Air separation module is used to screen paper, plastic film in garbage;Screening module is used to screen garbage of different volumes;The binocular sorting object module is screened again to the garbage in garbage.This automatic sorting equipment utilizes monocular and binocular vision technology, accurately identifies four kinds of articles such as plastic, glass, metal and paper in recyclable domestic waste, greatly improves recovery rate, and the sorting process is high in efficiency and high in accuracy, reduces manual intervention and labor cost.

Description

Technical Field

[0001] This invention relates to the field of solid waste recycling, and more specifically to an automatic sorting device for recyclable household waste. Background Technology

[0002] With population growth and accelerated urbanization, the amount of recyclable waste is increasing daily. Common recyclable waste includes newspapers, magazines, cardboard boxes, beverage bottles, food plastic containers, cosmetic bottles, glass bottles, glass containers, glass cups, tin cans, aluminum cans, shopping bags, cardboard boxes, paper packaging bags, scrap iron utensils, sheet metal, business cards, printed materials, and old textiles. These can be categorized into five types based on material: plastic, glass, metal, paper, and textiles. These items have high recycling value, and separate collection and sorting of recyclable waste facilitates subsequent processing and recycling, improving efficiency. Therefore, automated sorting and classification of recyclable items is crucial. Against this backdrop, leveraging advancements in computer vision, machine learning, and robotics, we can achieve automated sorting and classification of these five main categories of recyclable items: plastic, glass, metal, paper, and textiles. This automation technology can increase recycling rates, reduce the need for landfill and incineration, and decrease dependence on limited natural resources. However, textiles already have existing disposal sites for processing within the recyclable household waste category. Therefore, we will focus on the sorting and processing of four categories of items: plastic, glass, metal, and paper, in order to drive innovation and progress in the field of recyclable municipal solid waste management.

[0003] Currently, most recyclable household waste sorting and processing is done manually, which is not only inefficient but also requires a large amount of manpower. Some recyclable household waste sorting equipment has emerged, which mostly uses magnetic separation, air separation, and screening methods. Magnetic separation equipment can separate metallic and non-metallic waste, air separation equipment can separate lightweight waste such as paper scraps and plastic film, and screening equipment can separate waste of different sizes. The use of these devices can improve the sorting efficiency of small pieces of waste while reducing the labor intensity and noise pollution of manual sorting. However, there is still room for improvement in the sorting accuracy and efficiency of these devices. Therefore, a more efficient sorting technology combining machine vision is needed to complete the sorting of recyclable household waste. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an automatic sorting device for recyclable household waste. The sorting device has a high degree of automation, high sorting efficiency and sorting accuracy, and reduces the sorting cost of recycling plants.

[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0006] An automatic sorting device for recyclable household waste includes a feeding module, a single-eye magnetic metal sorting module, an air separation module, a screening module, and a dual-eye object sorting module arranged in sequence according to the sorting order. The single-eye magnetic metal sorting module is used to screen magnetic metals in the waste; the air separation module is used to screen paper and plastic film in the waste; the screening module is used to screen waste of different volumes; and the dual-eye object sorting module performs a second screening of the waste.

[0007] In one specific implementation, the feeding module includes a feeding frame, a feeding port, a perforated roller, a blowing device I, and a conveyor belt I. The feeding port is located on the feeding side of the feeding frame, the conveyor belt I is located inside the feeding frame along the feeding direction, the perforated roller is installed inside the feeding frame and close to the feeding side, the perforated roller is installed above the conveyor belt I and below the feeding port, and the blowing device I is installed inside the feeding frame and located on the axial side of the perforated roller.

[0008] As one specific implementation, the monocular sorting magnetic metal module includes a camera fixing frame I, a conveyor belt II, a monocular camera, a telescopic rod, a conveying pipe, a metal magnetic attraction module, and a magnetic metal sorting frame. The conveyor belt II is arranged along the feeding direction. The monocular camera is mounted above the feeding side of the conveyor belt II through the camera fixing frame I. The axis of the conveying pipe is arranged perpendicular to the conveying direction of the conveyor belt II. The magnetic metal sorting frame is located on the side of the conveyor belt II. The metal magnetic attraction module is connected to the conveying pipe through the telescopic rod and can slide along the length of the conveying pipe.

[0009] As a specific implementation method, the process of the monocular camera identifying metal in the monocular sorting magnetic metal module is as follows:

[0010] Step 1: Take pictures of garbage, use an automatic annotation algorithm to preprocess the data, extract image features to automatically group similar images, and then predict the bounding box position and category of objects in the pictures to build a dataset;

[0011] Step 2: Build an improved YOLOv5-trashnet model, specifically:

[0012] 1) Replace the PANet module in the YOLOv5 network with the BIFPN module;

[0013] 2) EIOU_LOSS is used as the regression loss function for the object detection task, and its formula is:

[0014] ;

[0015] 3) Add a CA attention mechanism to the model;

[0016] 4) Replace the original ordinary convolutional layers with the Ghost module;

[0017] 5) Set the hyperparameter batch_size to 16 and the learning rate strategy to warm-up;

[0018] Step 3: Use the dataset obtained in Step 1 to train the improved YOLOV5-trashnet model established in Step 2;

[0019] Step 4: Input the garbage images obtained by the monocular camera into the YOLOV5-trashnet model trained in Step 3 for metal recognition.

[0020] In one specific implementation, the air separation module includes an air separation frame, a blowing device II, and a paper-plastic sorting frame. The air separation frame is installed on the discharge end side of the conveyor belt II. There are three blowing devices II, which are distributed on both sides of the air separation frame along the conveying direction perpendicular to the conveyor belt II. One blowing device II is provided on one side and has an opening above it; two blowing devices are provided on the other side and are distributed in the vertical direction. The blowing device II located below is symmetrically arranged with the blowing device II on the other side, and the blowing device II located above is symmetrically arranged with respect to the opening. The air separation frame has an opening on the outside of the blowing device II, which is located on the paper-plastic sorting frame.

[0021] In one specific implementation, the binocular sorting module includes a parallel delta robotic arm mounted above conveyor belt III and a binocular camera mounted above the feed direction side of conveyor belt III. The parallel delta robotic arm is located behind the binocular camera, and a suction nozzle is provided on the output end of the parallel delta robotic arm.

[0022] As a specific implementation method, the process of binocular camera identifying garbage in the binocular object sorting module is as follows:

[0023] Step 1: Take garbage images using a dual-lens camera, preprocess the data using an automatic annotation algorithm, automatically group similar images by extracting image features, and then predict the bounding box positions and categories of objects in the images to build a dataset;

[0024] Step 2: Build an improved NanoDet-Plus_trashnet model, specifically:

[0025] 1) Set the basic units and downsampling basic units in the model to 5*5 convolution kernels;

[0026] 2) Add SE attention mechanism to the model;

[0027] 3) Use Bayesian optimization to adjust the hyperparameters;

[0028] 4) Binocular camera calibration, stereo correction and stereo matching: The binocular algorithm is added to the NanoDet-Plus_trashnet model to obtain the range measurement capability;

[0029] Step 3: Use the dataset obtained in Step 1 to train the improved NanoDet-Plus_trashnet model established in Step 2;

[0030] Step 4: Input the garbage images obtained by the stereo camera into the NanoDet-Plus_trashnet model trained in Step 3 for garbage identification.

[0031] As one specific implementation, the parallel delta robotic arm has three output shafts, which are arranged in a triangular pyramidal structure.

[0032] In one specific implementation, the screening module includes a fixed frame consisting of four columns and a screening screen mounted on the fixed frame by springs.

[0033] Compared with the prior art, the advantages of the present invention are:

[0034] 1. Improved Recycling Rate: By using monocular and binocular vision technology, the system can accurately identify four categories of recyclable household waste: plastic, glass, metal, and paper. This effectively improves the recycling rate, ensuring that these recyclable items are reused and reducing resource waste.

[0035] 2. Automated and efficient processing: By using a monocular positioning system to locate and pick up magnetic metal objects, the system can quickly identify and separate metal items. Then, using air separation technology, lightweight objects such as paper and plastic can be effectively separated from mixed waste. Subsequently, according to the size of the objects, large and small suction nozzles are used to pick up the objects separately, achieving precise sorting of items of different sizes.

[0036] 3. The entire process is highly automated, which greatly improves sorting efficiency and accuracy, and reduces manual intervention and labor costs. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1This is an overall appearance view of the automatic sorting equipment for recyclable household waste described in this invention;

[0039] Figure 2 This is a schematic diagram of the feeding module structure in the automatic sorting equipment for recyclable household waste described in this invention;

[0040] Figure 3 This is an exploded view of the feeding module;

[0041] Figure 4 This is a schematic diagram of the structure of the automatic sorting device for recyclable household waste described in this invention, which is a single-eye magnetic metal sorting module.

[0042] Figure 5 This is an exploded view of a monocular magnetic metal sorting module;

[0043] Figure 6 This is a schematic diagram of the air separation module structure in the automatic sorting equipment for recyclable household waste described in this invention;

[0044] Figure 7 This is an exploded view of the wind sorting module;

[0045] Figure 8 This is a schematic diagram of the screening module structure in the automatic sorting equipment for recyclable household waste described in this invention;

[0046] Figure 9 This is an exploded view of the screening module;

[0047] Figure 10 This is a schematic diagram of the binocular large object sorting module in the automatic sorting equipment for recyclable household waste described in this invention;

[0048] Figure 11 This is an exploded view of the binocular sorting module for large objects;

[0049] Figure 12 This is a schematic diagram of the binocular small object sorting module in the automatic sorting equipment for recyclable household waste described in this invention;

[0050] Figure 13 This is an exploded view of the binocular small object sorting module;

[0051] Figure 14 This is a flowchart illustrating the implementation of the automatic sorting equipment for recyclable household waste described in this invention.

[0052] Figure 15 This is the framework for the improved YOLOV5-trashnet model in this invention;

[0053] Figure 16 This is the framework for the improved NanoDet-Plus_trashnet model in this invention;

[0054] The module includes: 100, feeding module; 200, single-eye magnetic metal sorting module; 300, air separation module; 400, screening module; 500, binocular large object sorting module; and 600, binocular small object sorting module.

[0055] 110. Feeding frame; 120. Motor I; 130. Hollowed-out roller; 140. Feed inlet; 150. Blowing device I; 160. Conveyor belt I; 210. Camera mounting frame I; 220. Conveyor belt II; 230. Monocular camera; 240. Telescopic rod; 250. Connecting block; 260. Conveying pipe; 270. Metal magnetic module; 280. Magnetic metal sorting frame; 310. Air separator frame; 320. Blowing device II; 330. Paper-plastic sorting frame; 410. Screening mesh; 420. Spring; 4 30. Motor II; 440. Mounting frame; 510. Robotic arm mounting frame I; 520. Parallel delta robotic arm I; 530. Camera mounting frame II; 540. Binocular camera I; 550. Conveyor belt III; 560. Sorting frame; 570. Large object suction nozzle; 580. Divider; 610. Camera mounting frame III; 620. Binocular camera II; 630. Robotic arm mounting frame II; 640. Parallel delta robotic arm II; 650. Small object suction nozzle; 660. Non-magnetic metal sorting frame. Detailed Implementation

[0056] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] An automated sorting device for recyclable household waste, see [link to relevant documentation]. Figure 1 As shown, along the waste sorting sequence, the module includes a feeding module 100, a single-eye magnetic metal sorting module 200, an air separation module 300, a screening module 400, and a binocular object sorting module. The binocular object sorting module further includes a binocular large object sorting module 500 and a binocular small object sorting module 600.

[0058] See Figure 2 , 3As shown, the feeding module 100 includes a feeding frame 110, a feeding port 140, a hollow roller 130, a blowing device I 150, a conveyor belt I 160, and a motor I 120. The feeding port 140 is located on the feeding side of the feeding frame 110. The conveyor belt I 160 is located inside the feeding frame 110 along the feeding direction. The hollow roller 130 is installed inside the feeding frame 110 and close to the feeding side. The hollow roller 130 is installed above the conveyor belt and below the feeding port 140 and is driven to rotate by the motor I 120. The blowing device I 150 is installed inside the feeding frame 110 and located on the axial side of the hollow roller 130. The garbage is poured into the feed inlet 140 and enters the hollow roller 130. The motor I 120 drives the hollow roller 130 to rotate. Lightweight objects scattered on the hollow roller 130 are blown up by the blower I 150, causing the garbage to scatter onto the conveyor belt I 160.

[0059] See Figure 4 , 5 As shown, the monocular sorting magnetic metal module includes a camera mounting frame I 210, a conveyor belt II 220, a monocular camera 230, a telescopic rod 240, a conveying pipe 260, a metal magnetic module 270, and a magnetic metal sorting frame 280. The conveyor belt II 220 is positioned along the feeding direction. The monocular camera 230 is mounted above the feeding side of the conveyor belt II 220 via the camera mounting frame I 210. The axis of the conveying pipe 260 is perpendicular to the conveying direction of the conveyor belt II 220. The magnetic metal sorting frame 280 is located on the side of the conveyor belt II 220. The metal magnetic module 270 is connected to the conveying pipe 260 via the telescopic rod 240 and can slide along the length of the conveying pipe 260.

[0060] Specifically, the telescopic rod 240 is slidably mounted on the conveyor pipe 260 via a connecting block 250 with mounting holes. Conveyor belt I 160 and conveyor belt II 220 are connected. The garbage scattered on conveyor belt I 160 is conveyed to conveyor belt II 220. The monocular camera 230 determines the location of the metal in the garbage. When the metal is at the same level as the metal magnetic module 270, the conveyor belt II 220 stops running. The telescopic rod 240 and the metal magnetic module 270 are used to pick up the metal. Then, the connecting block 250 drives the telescopic rod 240 to slide on the conveyor pipe 260 to above the magnetic metal sorting frame 280. Then, the power is cut off, and the metal falls into the frame. Then, the conveyor belt I 160 continues to run, and the metal magnetic module 270 returns to above the conveyor belt I 160 to continue picking up metal. The above work is repeated.

[0061] Here, the process of the monocular camera 230 identifying metal in the monocular magnetic metal sorting module is as follows:

[0062] Step 1: Take pictures of garbage and preprocess the data using automatic annotation algorithms, such as spectral clustering-based annotation methods. By extracting image features, similar images are automatically grouped, and the bounding box positions and categories of objects in the images are predicted. This enables automatic annotation of large amounts of data, reduces the workload of manual annotation, and establishes a dataset.

[0063] Step 2: Build an improved YOLOv5-trashnet model, see [link / reference]. Figure 15 As shown, specifically:

[0064] 1) Replace the PANet module in the YOLOv5 network with the BIFPN module. The BiFPN module uses two complementary pyramid-shaped feature maps to achieve multi-scale feature propagation and fusion. Specifically, one pyramid upsamples features from a low-resolution feature map to a high-resolution one, and the other pyramid downsamples features from a high-resolution one to a low-resolution one. Finally, the output features are obtained through element-wise weighted fusion.

[0065] 2) EIOU_LOSS is used as the regression loss function for the object detection task, and its formula is:

[0066] ,

[0067] By using EIoU_Loss to weightedly combine the classification cross-entropy loss and regression loss, the classification and regression performance are better balanced.

[0068] 3) Add a CA attention mechanism to the model. The CA attention mechanism utilizes both channel and spatial attention to collaboratively enhance features, effectively improving the performance of the object detection model.

[0069] 4) Replacing the original ordinary convolutional layer with the Ghost module can make the model lighter. The Ghost module decomposes the convolution operation into different groups of convolutional kernels through group parameterization technology. The parameters are shared between the groups, which greatly reduces the model size. At the same time, the independent operation space between the groups preserves the integrity of feature extraction. In this way, the object detection model can achieve depth and width expansion while maintaining accuracy, and more efficiently balance performance and computational cost.

[0070] 5) Set the hyperparameter batch_size to 16 and the learning rate strategy to warm-up;

[0071] Step 3: Use the dataset obtained in Step 1 to train the improved YOLOV5-trashnet model established in Step 2;

[0072] Step 4: Input the garbage images obtained from the monocular camera into the YOLOV5-trashnet model trained in Step 3 for metal recognition. See also Figure 6 , 7 As shown, the air separation module includes an air separation frame 310, a blowing device II 320, and a paper-plastic sorting frame 330. The air separation frame 310 is installed on the discharge end side of the conveyor belt II 220. There are three blowing devices II, which are distributed on both sides of the air separation frame 310 along the conveying direction perpendicular to the conveyor belt II 220. One blowing device II is provided on one side and has an opening above it; two blowing devices are provided on the other side and are distributed vertically. The blowing device II located below is symmetrically arranged with the blowing device II on the other side, and the blowing device II located above is symmetrically arranged with respect to the opening. The air separation frame 310 has an opening on the outside of the blowing device II, which is located on the paper-plastic sorting frame 330. When the waste passes through the air separation module, the two lower blowing devices II blow up the paper and plastic film, and the upper blowing device II blows the blown paper and plastic film to the opposite opening, from where they fall into the paper and plastic sorting box 330 below.

[0073] See Figure 8 , 9 As shown, the screening module includes a screening screen 410, a spring 420, a motor II 430, and a fixing frame 440. After the magnetic metal, paper, and plastic film have been sorted by the monocular magnetic metal sorting module and the air separation module, the waste passes through the screening module. The rotation of the motor II 430 drives the spring 420 to cause the screening screen 410 to vibrate. Under the action of gravity, the smaller waste falls through the gaps in the screening screen 410, while the larger waste remains on the screening screen 410, thereby screening waste of different sizes.

[0074] After passing through the screening module, the waste enters the binocular sorting module. This module includes a parallel delta robotic arm mounted above conveyor belt III550 and a binocular camera mounted above the feed direction side of conveyor belt III550. The parallel delta robotic arm is located behind the binocular camera, and a suction nozzle is installed at its output end. For objects of different sizes, this binocular sorting module is divided into a large object sorting module and a small object sorting module.

[0075] See Figure 10 , 11As shown, the binocular sorting module for large objects includes a robotic arm fixing frame I 510, a parallel delta robotic arm I 520, a camera mounting bracket II 530, a binocular camera I 540, a large object suction nozzle 570, and a separator plate 580 I. The binocular camera I 540 is mounted above the feed direction side of the conveyor belt III 550 via the camera mounting bracket II 530. The parallel delta robotic arm I 520 is mounted above the conveyor belt III 550 and located behind the binocular camera I 540 via the robotic arm fixing frame I 510.

[0076] The parallel delta robotic arm I 520 is driven by three motors, which in turn move the large object suction nozzle 570 via three sets of output shafts. These three output shafts form a triangular pyramid structure, with the large object suction nozzle positioned at the front end of each shaft. After the binocular camera I 540 captures and identifies the object, obtaining its position and depth information, the parallel delta robotic arm I 520 uses the output shafts to deliver the large object suction nozzle 570 to the designated location to pick up the target object. The object is then placed into three sorting boxes 560, one for paper, one for plastic, and one for glass. Metal objects slide directly into the binocular camera's small object sorting module. The parallel delta robotic arm I 520 ensures more precise and stable grasping. A separator plate 580 is used to separate the multiple sorting boxes.

[0077] The binocular small object sorting module includes a robotic arm mounting frame II 630, a parallel delta robotic arm II 640, a camera mount III 610, a binocular camera II 620, a small object suction nozzle 650, and a separator plate 580. Similarly, the binocular camera II 620 is mounted above the conveyor belt III 550 via the camera mount III 610, and the parallel delta robotic arm II 640 is mounted above the conveyor belt III 550 via the robotic arm mounting frame II 630. The parallel delta robotic arm II 640 is also driven by three motors, which are driven by three sets of output shafts. The small object suction nozzle 650 moves, and the binocular camera II 620 takes pictures and identifies the objects, obtaining position and depth information. The parallel delta robotic arm II 640 then uses its output shaft to send the small object suction nozzle 650 to the designated position to pick up the target object and place it into the designated sorting frame 560. This sorting frame includes three frames located on both sides of the conveyor belt III 550, used for paper, plastic, and glass respectively. A non-magnetic metal sorting frame 660 is also provided at the end of the conveyor belt III. A separator plate 580 is used to separate the multiple sorting frames on the side of the conveyor belt III. After the binocular camera sorts the large object module and the binocular camera sorts the small object module, the remaining large and small metal objects slide together into the non-magnetic metal sorting frame 660 at the end of the conveyor belt III.

[0078] Here, in the binocular object sorting module, the binocular camera takes pictures of the garbage and inputs the pictures into the trained NanoDet-Plus_trashnet model for recognition. The specific process is as follows:

[0079] Step 1: Take garbage images using a dual-lens camera, and preprocess the data using automatic annotation algorithms, such as spectral clustering-based annotation methods. By extracting image features, similar images are automatically grouped, and the bounding box positions and categories of objects in the images are predicted. This enables automatic annotation of large amounts of data, reduces the workload of manual annotation, and establishes a dataset.

[0080] Step 2: Build the improved NanoDet-Plus_trashnet model, see [link / reference] Figure 16 As shown, specifically:

[0081] 1) Set the basic units and downsampling basic units in the model to 5*5 convolution kernels;

[0082] 2) Add SE attention mechanism to the model. SE attention mechanism has small changes in parameter quantity. Adding SE attention mechanism can improve the expressive power and performance of neural network by adaptively adjusting the weight of channel feature map.

[0083] 3) The hyperparameters are tuned using Bayesian optimization. It optimizes the objective function by constructing a surrogate function, which is a Gaussian process that predicts the value of the objective function based on the prior distribution and the known value of the objective function. The maximum value of the surrogate function is the suggested value of the next hyperparameter. This process is iterated until the predetermined number of iterations is reached or the objective function converges.

[0084] 4) Binocular camera calibration, stereo correction and stereo matching: The binocular algorithm is added to the NanoDet-Plus_trashnet model to obtain the range measurement capability;

[0085] Step 3: Use the dataset obtained in Step 1 to train the improved NanoDet-Plus_trashnet model established in Step 2;

[0086] Step 4: Input the garbage images obtained by the stereo camera into the NanoDet-Plus_trashnet model trained in Step 3 for garbage identification.

[0087] See Figure 14 As shown, the automatic sorting equipment for recyclable household waste using the present invention operates as follows:

[0088] 1. Use the PLC to start all the conveyor belts to move, and start the parallel delta robotic arm to dump the garbage into the feed inlet;

[0089] 2. Then, the monocular camera in the monocular sorting magnetic metal module visually determines whether the waste contains metal. If yes, the magnetic metal is attracted by the metal magnetic attraction module; otherwise, it enters the air separation module.

[0090] 3. Paper and plastic film are separated by the air separation module;

[0091] 4. Then it enters the screening module, where small-volume objects are screened out;

[0092] 5. Enter the binocular sorting module. In the binocular sorting module for large objects, use the large object suction nozzle to pick up large-volume items and place them into the corresponding boxes. In the binocular sorting module for small objects, use the small object suction nozzle to place small-volume waste into the corresponding boxes. Place the remaining non-magnetic metal into the non-magnetic metal sorting box.

[0093] 6. End.

[0094] The entire process is highly automated, with high sorting efficiency and accuracy, which greatly saves labor costs.

[0095] It should be emphasized that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. An automatic sorting device for recyclable household waste, characterized in that, The system includes a feeding module, a single-eye magnetic metal sorting module, an air separation module, a screening module, and a binocular object sorting module arranged in the sorting order. The single-eye magnetic metal sorting module is used to screen magnetic metals in the waste; the air separation module is used to screen paper and plastic film in the waste; the screening module is used to screen waste of different sizes; the binocular object sorting module performs a second screening of the waste. The binocular object sorting module includes a parallel delta robotic arm installed above conveyor belt III and a binocular camera installed above the feeding side of conveyor belt III. The parallel delta robotic arm is located behind the binocular camera. The output end of the parallel delta robotic arm is equipped with a suction nozzle. For objects of different sizes, the binocular object sorting module is divided into a binocular large object sorting module and a binocular small object sorting module. The process of binocular camera identifying waste in the binocular object sorting module is as follows: Step 1: Take garbage images using a dual-lens camera, preprocess the data using an automatic annotation algorithm, automatically group similar images by extracting image features, and then predict the bounding box positions and categories of objects in the images to build a dataset; Step 2: Build an improved NanoDet-Plus_trashnet model, specifically: 1) Set the basic units and downsampling basic units in the model to 5*5 convolution kernels; 2) Add SE attention mechanism to the model; 3) Use Bayesian optimization to adjust the hyperparameters; 4) Binocular camera calibration, stereo correction and stereo matching: The binocular algorithm is added to the NanoDet-Plus_trashnet model to obtain the range measurement capability; Step 3: Train the improved NanoDet-Plus_trashnet model established in Step 2 using the dataset obtained in Step 1; Step 4: Input the garbage images obtained by the stereo camera into the NanoDet-Plus_trashnet model trained in Step 3 for garbage identification.

2. The automatic sorting equipment for recyclable household waste according to claim 1, characterized in that, The feeding module includes a feeding frame, a feeding port, a perforated roller, a blowing device I, and a conveyor belt I. The feeding port is located on the feeding side of the feeding frame. The conveyor belt I is located inside the feeding frame along the feeding direction. The perforated roller is installed inside the feeding frame and close to the feeding side. The perforated roller is installed above the conveyor belt I and below the feeding port. The blowing device I is installed inside the feeding frame and located on the axial side of the perforated roller.

3. The automatic sorting equipment for recyclable household waste according to claim 1, characterized in that, The monocular sorting magnetic metal module includes a camera fixing frame I, a conveyor belt II, a monocular camera, a telescopic rod, a conveying pipe, a metal magnetic attraction module, and a magnetic metal sorting frame. The conveyor belt II is arranged along the feeding direction. The monocular camera is mounted above the feeding side of the conveyor belt II through the camera fixing frame I. The axis of the conveying pipe is perpendicular to the conveying direction of the conveyor belt II. The magnetic metal sorting frame is located on the side of the conveyor belt II. The metal magnetic attraction module is connected to the conveying pipe through the telescopic rod and can slide along the length of the conveying pipe.

4. An automatic sorting device for recyclable household waste according to claim 1, characterized in that, The air separation module includes an air separation frame, a blowing device II, and a paper-plastic sorting frame. The air separation frame is installed on the discharge end side of the conveyor belt II. There are three blowing devices II, which are distributed on both sides of the air separation frame along the conveying direction perpendicular to the conveyor belt II. One blowing device II is provided on one side and has an opening above it; two blowing devices are provided on the other side and are distributed vertically. The blowing device II located below is symmetrically arranged with the blowing device II on the other side, and the blowing device II located above is symmetrically arranged with respect to the opening. The air separation frame has an opening on the outside of the blowing device II, which is located on the paper-plastic sorting frame.

5. An automatic sorting device for recyclable household waste according to claim 1, characterized in that, The parallel delta robotic arm has three output shafts, which are arranged in a triangular pyramidal structure.

6. An automatic sorting device for recyclable household waste according to claim 1, characterized in that, The screening module includes a fixed frame consisting of four columns and a screening screen mounted on the fixed frame by springs.

Images