Tea garden intelligent precise pesticide spraying system and pesticide spraying method based on machine vision

Abstract

The invention provides an intelligent precise pesticide spraying system and method for a tea garden based on machine vision, and belongs to the technical field of tea garden machinery, and the method comprises the steps: S1, constructing a tea disease detection model; s2, collecting tea garden field pictures; and S3, transmitting the tea garden field pictures acquired in the step S2 to an NVIDIA Jetson Orin Nano master control system, detecting through the detection model obtained in the step S1, judging whether diseases and insect pests occur or not, and if so, calculating the positions where the diseases and insect pests occur and then performing pesticide spraying treatment. Precise identification of tea garden diseases and insect pests and precise pesticide spraying are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of tea garden machinery, and particularly relates to an intelligent and precise spraying system and a spraying method for a tea garden based on machine vision. Background Art

[0002] At present, tea leaves are easily affected by various diseases during the growth process, resulting in a decline in the quality and yield of tea leaves. Accurately detecting tea leaf diseases is an important technical prerequisite for tea farmers to carry out early prevention and reduce losses. However, traditional methods for identifying tea leaf diseases require manual operation and are easily interfered by complex backgrounds, resulting in low recognition accuracy and missed or misdetected phenomena. Currently, common spraying methods in tea gardens mainly include manual targeted spraying and uniform large-area spraying by agricultural machinery and drones. These methods have problems such as time-consuming, labor-intensive, high labor costs, waste of pesticides, and environmental pollution, which are not conducive to the sustainable development of agricultural ecology.

[0003] In addition, the ecological tea garden tillage mode of intercropping green manure crops emerging in hilly and mountainous areas makes the tea gardens in hilly and mountainous areas with serious soil compaction have problems such as complex roots and easy entanglement of tools with grass during the ditch-opening process. The tea ridge spacing is only about 60 cm, making most existing middle-tillage fertilizing machines unable to meet the operation requirements of ecological tea gardens. The existing tea garden tillers in hilly and mountainous ecological tea gardens cannot reach the required ditch-opening depth, and manual fertilization and soil covering are still required after ditch-opening. Due to the relatively large soil moisture and higher adhesiveness in southern regions, the ditch shapes opened by tillers are uneven, and the uniformity of fertilization cannot be guaranteed, which reduces the efficiency of middle-tillage operations in tea gardens. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an intelligent and precise spraying system and a spraying method for a tea garden based on machine vision, which can achieve precise identification of pests and diseases and precise spraying of pesticides during mechanical operations in the tea garden.

[0005] To solve the above technical problem, the technical solution adopted by the present invention is: An intelligent and precise spraying system for a tea garden based on machine vision, comprising an electric engine 1, a frame 2, traveling wheels 3, a gearbox 4, a handrail device 11, and a precise spraying device 12; The electric engine 1 is installed behind the frame 2, the frame 2 is welded to the gearbox 4, and the handrail device 11 is installed above the gearbox 4; The described precise spraying device 12 includes an identification camera 12-1, a camera bracket 12-2, a medicine box bracket 12-3, a medicine box 12-4, an electric pump 12-5, a solenoid valve 12-6, a spraying bracket 12-7, a spraying hose 12-8, a spray head 12-9, an outer horizontal sliding guide rail 12-10, an inner horizontal sliding guide rail 12-11, a longitudinal sliding guide rail 12-12, and an NVIDIA Jetson Orin Nano main control system 12-13. The camera bracket 12-2 is welded to the front end of the frame 2, and the identification camera 12-1 is installed at both ends of the camera bracket 12-2 for identifying the situation of tea pests and diseases. The liquid medicine is transported to the solenoid valve 12-6 through the electric pump 12-5. Both ends of the spraying hose 12-8 are respectively connected to the solenoid valve 12-6 and the spray head 12-9. One outer horizontal sliding guide rail 12-10, one inner horizontal sliding guide rail 12-11, and two longitudinal sliding guide rails 12-12 are installed on each side of the spraying bracket 12-7. The end of the extension section of the outer horizontal sliding guide rail 12-10 is connected to a longitudinal sliding guide rail 12-12, and the sliding block of the inner horizontal sliding guide rail 12-11 is connected to a longitudinal sliding guide rail 12-12. The NVIDIA Jetson Orin Nano main control system 12-13 is used to receive and analyze the information of the tea disease area detected by the identification camera 12-1, and control the spray head 12-9 to spray medicine on the disease area.

[0006] Furthermore, it also includes an inclined ditching device 6 and a belt transmission device 10; The power of the electric engine 1 is transmitted to the gearbox 4 through the belt transmission device 10. The gearbox 4 is provided with an output shaft one and an output shaft two. The output shaft one of the gearbox 4 is connected to the traveling wheel 3 to drive the traveling wheel 3 to rotate; The inclined ditching device 6 includes a chain transmission box 6-1, a transmission shaft 6-2, a cross coupling 6-3, a rotary tillage device frame 6-4, a rotary tillage transmission mechanism 6-5, and a rotary tillage mechanism 6-6. The bottom plate installed below the chain transmission box 6-1 is fixedly installed on the front end of the frame 2 through bolts. Each of the left and right sides of the chain transmission box 6-1 is connected and driven to a rotary tillage transmission mechanism 6-5 through a cross coupling 6-3. One rotary tillage mechanism 6-6 is connected to each of the two rotary tillage transmission mechanisms 6-5. The output shaft two of the gearbox 4 transmits power to the chain transmission box 6-1 through chain transmission, and drives the cross coupling 6-3, the rotary tillage transmission mechanism 6-5, and the rotary tillage mechanism 6-6 to operate; The axial direction of the transmission shaft 6-2 forms an angle of 70° with the axis of the rotary tillage transmission mechanism 6-5. The axis of the rotary tillage transmission mechanism 6-5 forms an angle greater than 70° and less than 90° with the forward direction. The rotary tillage device frame 6-4 covers the outside of the cross coupling 6-3. The rotary tillage device frame 6-4 is fixedly connected to the chain transmission box 6-1 and the rotary tillage transmission mechanism 6-5 by bolts.

[0007] Further, the rotary tillage mechanism 6-6 includes a ditching and soil throwing knife 6-61, a mounting cutter disc 6-62, a rotary tillage cutter shaft 6-63, and a soil throwing iron sheet 6-64. The rotary tillage mechanism 6-6 is installed in a left-right opposed manner. When viewed from the front, the angle formed by the disk surfaces where the mounting cutter discs 6-62 on the left and right sides are located is an acute angle greater than 20° and less than 40°. There are 5 arc-shaped ditching and soil throwing knives 6-61 circumferentially distributed on each mounting cutter disc 6-62. A circular soil throwing iron sheet 6-64 is welded on the back of the ditching and soil throwing knife 6-61. A rotary tillage cutter shaft 6-63 is coaxially arranged and fixedly connected to a mounting cutter disc 6-62. A rotary tillage cutter shaft 6-63 is fixed on the output shaft of a rotary tillage transmission mechanism 6-5.

[0008] The above device can achieve precise spraying in the tea garden through the precise spraying device 12. A spraying method for an intelligent precise spraying system in a tea garden based on machine vision. The method is as follows: Step S1: Construct a tea leaf disease detection model, specifically: Step S11: Collect a tea leaf disease data set containing different environments and types. Step S12: Label the tea leaf disease data set collected in step S11 and divide it into a training set, a validation set, and a test set according to a certain proportion. Step S13: Perform data enhancement and expansion on the training set in step S12. Step S14: Construct a tea leaf disease detection model based on GDE-YOLOv8n: Introduce a global attention mechanism (GAM) at the end of the Neck of the original YOLOv8n model, use a diverse branch block (DBB) in combination with the C2f module in the Neck network of the original YOLOv8n model, and replace the original CIoU loss function with an EIoU loss function. Step S15: Train and experiment on the model constructed in step S14 to obtain a detection model. Step S16: Deploy the detection model obtained in step S15 to run on the NVIDIA Jetson Orin Nano main control system. Step S2: Collect pictures of the tea garden site. Step S3: Transmit the on-site pictures of the tea garden collected in Step S2 to the NVIDIA Jetson Orin Nano main control system, and perform detection through the detection model obtained in Step S1 to determine whether there are pests and diseases. If so, calculate the location where the pests and diseases occur and then perform spraying treatment.

[0009] Further, in Step S11, collect a tea leaf disease dataset containing different environments and species, specifically: collect pictures of tea leaf pests and diseases in different regions, different types of tea leaf pests and diseases, and different environmental conditions.

[0010] Further, in Step S12, label the dataset and divide it into a training set, a validation set, and a test set according to a ratio, specifically: use the LabelImg picture annotation tool to annotate the disease images collected in Step S11 in TXT format to form a label file, and finally divide the labeled dataset into a training set, a validation set, and a test set according to a ratio of 7:2:1.

[0011] Further, in Step S13, perform data augmentation operations on the training set divided and labeled in Step S12, specifically: perform transformations on the training set in Step S12, including brightness conversion, noise addition, and random angle rotation, and finally expand the original dataset by four times.

[0012] Further, Step S15: Train and experiment on the model constructed in Step S14 to obtain a detection model, specifically: use the Backbone backbone network to extract features from the input image to obtain input features at three scales; in the Neck neck network, use the Feature Pyramid Network (FPN) and the Path Aggregation Network (PAN) to fuse the input features at three scales to obtain feature output maps at three scales; the Detection Head performs final regression and classification operations on the feature output maps at three scales, predicts the bounding box regression values and the confidence levels of the existence of targets for each anchor box, and adjusts the anchor boxes using the regression prediction values to generate the final prediction boxes.

[0013] The beneficial effects of the present invention are as follows: It can achieve the accurate identification of pests and diseases and the accurate spraying of pesticides during the mechanical operation of the tea garden. Description of the Drawings

[0014] Figure 1 It is a schematic structural diagram of the intelligent precision spraying system for tea gardens based on machine vision according to an embodiment of the present invention; Figure 2 It is a schematic structural diagram of the intelligent precision spraying system for tea gardens based on machine vision according to an embodiment of the present invention; Figure 3 It is a schematic structural diagram of the inclined trenching device; Figure 4It is a schematic diagram of the rotary tillage mechanism; Figure 5 It is a schematic diagram of the structure of the double-disc soil covering device; Figure 6 It is a schematic diagram of the structure of the precision spraying device; Figure 7 It is a schematic diagram of the structure of the precision spraying device; Figure 8 It is a schematic diagram of the device working in the field; Figure 9 It is a structure diagram of the GDE-YOLOv8n model; Figure 10 It is Figure 9 an enlarged view of the backbone part of Figure 11 It is Figure 9 an enlarged view of the neck network (Neck) and detection head (Detection Head) parts of Figure 12 It is a structure diagram of the GAM model; Figure 13 It is a structure diagram of the DBB model; Figure 14 It is a schematic diagram of the principle of the EIoU loss function; Figure 15 It is a flow chart of the spraying method of the intelligent precision spraying system for tea gardens based on machine vision; Figure 16 It is a sub-flow chart of step S1 in the spraying method.

[0015] Label description: 1. Electric engine; 2. Frame; 3. Traveling wheel; 4. Gearbox; 5. Double-disc soil covering device; 6. Inclined ditching device; 7. Depth-limiting device; 8. Soil retaining mechanism; 9. Fertilizer discharging device; 10. Belt transmission device; 11. Armrest device; 12. Precision spraying device; 13. Positioning block; 14. Fertilizer discharging hose; 15. Front support; 16. Battery; 17. Soil retaining plate; 18. Tea tree; 5-1. Vertical rod; 5-2. Notched disc cutter; 5-3. V-shaped double-disc shaft; 5-4. Bearing seat; 5-5. Protective cover; 5-6. Soil covering plate; 5-7. Soil covering plate connecting piece; 5-8. Scraping plate; 5-9. Cover mounting plate; 5-10. Fertilizer guiding pipe; 6-1. Chain transmission box; 6-2. Transmission shaft; 6-3. Cross coupling; 6-4. Rotary tillage device frame; 6-5. Rotary tillage transmission mechanism; 6-6 Rotary tillage mechanism; 6-61. Ditching and soil throwing knife; 6-62. Mounting cutter disc; 6-63. Rotary tillage knife shaft; 6-64. Soil throwing iron sheet; 7-1. Depth-limiting wheel adjusting rod; 7-2. Depth-limiting wheel; 9-1. Fertilizer box; 9-2. External fluted roller fertilizer applicator; 9-3. Bracket of fertilizer application device 12-1. Identification camera; 12-2. Camera bracket; 12-3. Bracket of medicine box; 12-4. Medicine box; 12-5. Electric pump; 12-6. Solenoid valve; 12-7. Bracket of spraying medicine; 12-8. Spraying medicine hose; 12-9. Nozzle; 12-10. External horizontal sliding guide rail; 12-11. Internal horizontal sliding guide rail; 12-12. Longitudinal sliding guide rail; 12-13. NVIDIA Jetson Orin Nano main control system 18-1. Distal end of tea tree; 18-2. Proximal end of tea tree Detailed implementation mode

[0016] To describe the technical content, achieved purpose and effects of the present invention in detail, the following is described in conjunction with the implementation modes and with reference to the drawings

[0017] Please refer to Figures 1 to 8 , the embodiment provided by the present invention is as follows An intelligent and precise tea garden spraying system based on machine vision, comprising an electric engine 1, a frame 2, traveling wheels 3, a gearbox 4, a double-disc soil covering device 5, an inclined ditch-opening device 6, a depth-limiting device 7, a soil retaining mechanism 8, a fertilizer application device 9, a belt drive device 10, a handrail device 11, and a precise spraying device 12. The electric engine 1 is installed behind the frame 2, the frame 2 is welded to the gearbox 4, the handrail device 11 is installed above the gearbox 4, the inclined ditch-opening device 6 is installed in front of the frame 2, the fertilizer application device 9 is installed above the front end of the frame, the double-disc soil covering device 5 is installed below the fertilizer application device 9, the depth-limiting device 7 is installed in front of the inclined ditch-opening device 6 through a front bracket 15, and the soil retaining mechanism 8 is symmetrically installed on both sides of the front bracket 15

[0018] In the embodiment of the present invention, the electric engine 1 transmits power to the gearbox 4 through the belt drive device 10. The gearbox 4 is provided with an output shaft one and an output shaft two. The output shaft one of the gearbox 4 is connected to the traveling wheels 3 to drive the traveling wheels 3 to rotate

[0019] In the embodiment of the present invention, the inclined ditching device 6 includes a chain transmission box 6-1, a transmission shaft 6-2, a cross coupling 6-3, a rotary tillage device frame 6-4, a rotary tillage transmission mechanism 6-5, and a rotary tillage mechanism 6-6. A bottom plate is installed below the chain transmission box 6-1 and is fixedly installed on the front end of the frame 2 through bolts. Each of the left and right sides of the chain transmission box 6-1 is connected and driven to a rotary tillage transmission mechanism 6-5 through a cross coupling 6-3. One rotary tillage mechanism 6-6 is connected to each of the two rotary tillage transmission mechanisms 6-5. The output shaft II of the gearbox 4 transmits power to the chain transmission box 6-1 through chain drive, and drives the cross coupling 6-3, the rotary tillage transmission mechanism 6-5, and the rotary tillage mechanism 6-6 to operate. The axial direction of the transmission shaft 6-2 forms an angle of 70° with the axis of the rotary tillage transmission mechanism 6-5, and the axis of the rotary tillage transmission mechanism 6-5 forms an angle greater than 70° and less than 90° with the forward direction. The rotary tillage device frame 6-4 covers the outside of the cross coupling 6-3, and the rotary tillage device frame 6-4 is fixedly connected to the chain transmission box 6-1 and the rotary tillage transmission mechanism 6-5 through bolts.

[0020] In the example of the present invention, the rotary tillage mechanism 6-6 includes a ditching and soil throwing knife 6-61, a mounting cutter disc 6-62, a rotary tillage cutter shaft 6-63, and a soil throwing iron sheet 6-64. The rotary tillage mechanisms 6-6 are installed in a left-right opposed manner. In a front view, as Figure 3 shown, the included angle α formed by the disk surfaces where the left and right mounting cutter discs 6-62 are located is an acute angle greater than 20° and less than 40°. Five arc-shaped ditching and soil throwing knives 6-61 are circumferentially distributed on each mounting cutter disc 6-62. A circular soil throwing iron sheet 6-64 is welded on the back of the ditching and soil throwing knife 6-61. A rotary tillage cutter shaft 6-63 is coaxially arranged and fixedly connected to a mounting cutter disc 6-62. A rotary tillage cutter shaft 6-63 is fixed on the output shaft of a rotary tillage transmission mechanism 6-5. It can play a role in soil throwing while breaking the soil. The left and right mounting cutter discs 6-62 are staggered at a certain angle, so that the ditching and soil throwing knife 6-61 enters the soil first each time when cutting the soil. While the rotary tillage mechanism 6-6 completes the primary ditching, the soil at the bottom of the ditch is thrown out. By means of the inclined and staggered rotary tillage method, the cutting contact area between the ditching knife and the soil is reduced, and the resistance during the ditching process is lowered.

[0021] In the embodiment of the present invention, the fertilizer discharge device 9 includes a fertilizer discharge box 9-1, an outer groove wheel fertilizer discharger 9-2, a battery 16, and a fertilizer discharge device bracket 9-3. The outer groove wheel fertilizer discharger 9-2 is fixed to the bottom of the fertilizer discharge box 9-1 by bolts, the bottom plate of the fertilizer discharge box 9-1 is fixed to the fertilizer discharge device bracket 9-3 by bolts, and the bottom plate of the fertilizer discharge bracket 9-3 is fixed to the frame 2 by bolts. The battery 16 is used to control the rotation speed of the groove wheel of the outer groove wheel fertilizer discharger 9-2 to adjust the fertilizer discharge amount. The fertilizer discharge port of the outer groove wheel fertilizer discharger 9-2 is connected to the fertilizer guide pipe 5-10 of the double disc covering device 5 through a fertilizer discharge hose 14. The double disc covering device 5 applies fertilizer to the groove while performing secondary trenching. The groove wheel rotation speed and the size of the fertilizer discharge outlet of the fertilizer discharge box 9-1 are controlled by the battery 16 to control the fertilizer discharge amount.

[0022] In the embodiment of the present invention, the double disc covering device 5 includes a vertical pole 5-1, a notched disc knife 5-2, a V-shaped double disc shaft 5-3, and a bearing seat 5-4. A positioning block 13 is provided on the frame 2. A slot for inserting the pole is provided on the positioning block 13. The pole 5-1 is provided with evenly distributed positioning holes in the vertical direction. The pole 5-1 inserted into the slot is locked and positioned by passing through the positioning holes thereon and bolts and nuts on the positioning block 13. The height of the double disc covering device 5 can be adjusted by locking and positioning through positioning holes of different heights on the pole 5-1. A V-shaped double disc shaft 5-3 is provided at the bottom of the pole 5-1. Two notched disc knives 5-2 arranged in a V shape are installed on the V-shaped double disc shaft 5-3 through a bearing seat 5-4.

[0023] In the embodiment of the present invention, the double disc covering device 5 also includes a shield 5-5, a covering plate 5-6, a covering plate connecting piece 5-7, a scraper plate 5-8 and a shield mounting plate 5-9. A shield mounting plate 5-9 is welded on the vertical pole 3. The shield 5-5 is fixed to the cover mounting plate 5-9 on the vertical pole 5-1 by bolts. A fertilizer guide pipe 5-10 is provided on the shield 5-5, and the fertilizer eventually passes through the fertilizer guide pipe 5-10 and falls into the opened ditch. The covering plate connecting piece 5-7 is welded to the vertical pole 5-1, and a connecting hole is provided on the connecting piece 5-7. The covering plate 5-6 and the covering plate connecting piece 5-7 are fixed by pins. The scraper plate 5-8 is located on the outside of the two notched disc cutters and is fixed on the shield 5-5.

[0024] In the embodiment of the present invention, the depth limiting device 7 includes a front bracket 15, a depth limiting wheel adjusting rod 7-1, and a depth limiting wheel 7-2. A fixed tube is welded on the front bracket 15, and the depth limiting wheel adjusting rod 7-1 is inserted into the fixed tube. The fixed tube and the depth limiting wheel adjusting rod 7-1 are matched by a positioning pin to achieve the adjustment of the trenching depth.

[0025] It further includes a retaining plate 17. Hinge holes are provided on both sides of the front bracket 15, and the retaining plate 17 is connected to the front bracket 15 through a hinge.

[0026] In the embodiment of the present invention, the precise spraying device 12 includes an identification camera 12-1, a camera bracket 12-2, a medicine box bracket 12-3, a medicine box 12-4, an electric pump 12-5, a solenoid valve 12-6, a spraying bracket 12-7, a spraying hose 12-8, a nozzle 12-9, an outer lateral sliding guide rail 12-10, an inner lateral sliding guide rail 12-11, a longitudinal sliding guide rail 12-12, and an NVIDIA Jetson Orin Nano main control system 12-13. The camera bracket 12-2 is welded to the front end of the frame 2, and the identification camera 12-1 is installed at both ends of the camera bracket 12-2 for identifying the tea plant diseases and pests. The liquid medicine is conveyed to the solenoid valve 12-6 by the electric pump 12-5. Both ends of the spraying hose 12-8 are respectively connected to the solenoid valve 12-6 and the nozzle 12-9. One outer lateral sliding guide rail 12-10, one inner lateral sliding guide rail 12-11, and two longitudinal sliding guide rails 12-12 are installed on each side of the spraying bracket 12-7. The end of the extension section of the outer lateral sliding guide rail 12-10 is connected to a longitudinal sliding guide rail 12-12, which is responsible for spraying the distal end 18-1 of the tea tree. The sliding block of the inner lateral sliding guide rail 12-11 is connected to a longitudinal sliding guide rail 12-12, which is responsible for spraying the proximal end 18-2 of the tea tree, so that the nozzle 12-9 can be aligned with the disease area of the tea ridge to achieve point spraying. The NVIDIA Jetson Orin Nano main control system 12-13 is used to receive and analyze the information of the tea disease area detected by the identification camera 12-1, and then control the corresponding nozzle to spray the disease area precisely, which can reduce the amount of pesticide used and reduce environmental pollution.

[0027] In an embodiment of the present invention, a method for using a tool inclined type ditch opening, fertilizing and soil covering device is as follows: (1) Before work, first adjust the depth of ditch opening through the depth limiting device 7 according to the agronomic requirements and actual situation of fertilization in the ecological tea garden; (2) When opening the ditch, use the cutting blades of the ditch opening and soil throwing knives 6-61 of the two rotary tillage mechanisms 6-6 symmetrically and obliquely installed in the front to cut the soil, and the soil throwing iron sheets 6-64 throw the soil at the bottom of the ditch, completing the preliminary ditch opening of the soil between the tea ridges; (3) When the double-disc soil covering device 5 is working, use the outer circumferential cutting edge of the notch disc knives 5-2 to cut the soil. Since the double discs are installed in a V shape and arranged with a narrow front and a wide rear, when the whole machine moves forward, the soil is pushed laterally and displaced, realizing secondary ditch opening and completing the leveling of the ditch shape; (4) When fertilizing, control the fertilization amount through the external groove wheel fertilizer distributor 9-2. Both ends of the fertilizer discharge hose 14 are respectively connected to the lower fertilizer discharge port of the external groove wheel fertilizer distributor 9-2 and the fixed fertilizer guiding pipe 5-10 on the double-disc soil covering device 5, and the fertilizer is dropped into the shaped bottom of the ditch; (5) The soil covering plate 5-6 is connected to the double-disc soil covering device 5 through the connecting piece welded on the soil covering plate 5-6, and the soil on both sides is pushed into the bottom of the ditch to realize soil covering; (6) When the device is walking in the tea garden, the recognition camera 12-1 collects the tea leaf disease information in front in real time. When the recognition camera 12-1 detects a disease, immediately analyze and process the central coordinate data of the disease area through the NVIDIA Jetson Orin Nano main control system 12-13, and then control the solenoid valve 12-6 of the corresponding nozzle 12-9 to open, and accurately spray the disease area, reducing the amount of pesticide used, playing a role in protecting the environment and being beneficial to the sustainable development of agricultural ecology.

[0028] Please refer to Figures 9 to 16 , a spraying method for an intelligent precision spraying system for tea gardens based on machine vision, and the method is as follows: Step S1: Construct a tea leaf disease detection model, specifically: Step S11: Collect a tea leaf disease data set containing different environments and types; Step S12: Label the tea leaf disease data set collected in Step S11, and divide it into a training set, a validation set and a test set according to a ratio; Step S13: Perform data enhancement and expansion on the training set in Step S12; Step S14: Construct a tea leaf disease detection model based on GDE-YOLOv8n; Step S15: Train and experiment on the model constructed in Step S14 to obtain a detection model; Step S16: Deploy the detection model obtained in Step S15 to run on the NVIDIA Jetson Orin Nano main control system; Step S2: Collect on-site pictures of the tea garden; Step S3: Transmit the on-site pictures of the tea garden collected in Step S2 to the NVIDIA Jetson Orin Nano main control system, and perform detection through the detection model obtained in Step S1 to determine whether there are pests and diseases. If so, calculate the location where the pests and diseases occur and then perform spraying treatment.

[0029] This method realizes precise spraying of pests and diseases in the tea garden, reduces the amount of pesticides used, plays a role in protecting the environment, and is conducive to the sustainable development of agricultural ecology.

[0030] In Step S11, the collected disease dataset comes from the tea garden, and 1,811 pictures with different types of tea diseases and different environmental conditions are obtained.

[0031] In Step S12, use the LabelImg picture annotation tool to annotate the disease images collected in Step S11 in TXT format to form a label file. Finally, divide the annotated dataset into a training set, a validation set, and a test set according to a ratio of 7:2:1. Among them, the training set is used for learning model parameters and optimizing the model through algorithms such as backpropagation; the validation set is used to evaluate model performance, adjust hyperparameters (such as learning rate, number of network layers), or select different model architectures during the training process to prevent the model from overfitting to the training data; the test set serves as an objective evaluation standard for the final performance of the model, simulating the generalization ability of the model to unknown data in a real scenario, and is only used when evaluating the final model after training is completed.

[0032] In Step S13, perform data augmentation operations on the training set divided and annotated in Step S12, including brightness conversion, noise addition, and random angle rotation. Finally, expand the original dataset four times.

[0033] In Step S14, as Figure 9 shown, the GDE-YOLOv8n tea disease detection model includes a backbone network, a neck network, and a detection head connected in sequence. The backbone network is used to extract feature information in the picture for use by the subsequent network; the neck network is used to better utilize the feature information extracted by the backbone network and plays a role in feature fusion; the role of the detection head is to use the features extracted by the backbone network and the neck network to make predictions on the input. Compared with the original YOLOv8n model, the GDE-YOLOv8n tea disease detection model introduces a global attention mechanism (GAM) at the end of the Neck ( Figure 11 layer 23 in Figure 11In the 13th, 16th, 19th, and 22nd layers), replace the original CIoU loss function with the EIoU loss function to improve the detection accuracy and speed of the model.

[0034] 1) Introduce a global attention mechanism (GAM) at the end of the Neck of the original YOLOv8n model: Traditional attention mechanisms ignore the importance of retaining information in both the channel and spatial dimensions. GAM strengthens the connection between channels and space by combining the advantages of the channel attention mechanism (Channel Attention, abbreviated as CA) and the spatial attention mechanism (Spatial Attention, abbreviated as SA), reduces the loss of target information in complex environments, and amplifies cross-dimensional feature information. As Figure 12 shown, CA first uses a 3D arrangement to retain the three-dimensional information in the input feature map; then uses a two-layer Multilayer Perceptron (MLP) with a compression ratio of r to amplify the cross-dimensional channel-space dependence; finally, applies a Sigmoid activation function to map the variables in the input feature map to between [0,1], generating a channel attention feature map. The Sigmoid activation function is calculated by formula (1): (1) where x represents the variable in the input feature map.

[0035] SA takes the feature map generated by multiplying the channel attention feature map by the residual of the GAM original input feature map as the input to develop adaptive features. To focus on spatial information, SA uses two convolutions for spatial information fusion. In addition, since the max pooling operation reduces the information utilization rate, it is removed to further retain the feature map. Given the input feature map, the intermediate state and output are defined by formulas (2) and (3) respectively.

[0036] (2) (3) where represents the input feature map, represents the output feature map, is the intermediate transition feature, and are the channel attention feature map and the spatial attention feature map respectively, represents element-wise multiplication.

[0037] 2) Combine the diverse branch block (DBB) with the C2f module in the Neck network: When improving the model performance, the computational cost and inference time often increase. However, DBB achieves a perfect balance between the two. AsFigure 13 As shown, DBB adopts a complex microstructure during training and is equivalent to a single convolution during inference, which improves the model performance while maintaining light weight. Generally, a convolution kernel is composed of the output channels , the input channels and the kernel size . It is essentially a fourth-order tensor and an optional bias . It takes channel feature maps as the input and outputs channel feature maps , where and are determined by , padding, and stride configurations. Use to represent the convolution operator and represent the bias as . The convolution form is shown in Equation (4): (4) The value at the th output channel is given by Equation (5): (5) where represents the convolution kernel with the input channel and the kernel size at the th output channel. is the th channel of corresponding to the sliding window at the position on . This correspondence is determined by padding and stride. The linear properties of convolution, including homogeneity and additivity, can be easily deduced from Equation (5), and the specific formulas are shown in Equations (6) and (7).

[0038] In Equation (6), represents channel feature maps, ; represents a feature map matrix of any size, and all elements in the matrix are real numbers, , where ; represents the convolution kernel composed of the output channels , the input channels and the kernel size , . In Equation (7), Denote the convolutional kernel composed of the output channels , the input channels and the kernel size ; ; Denote the convolutional kernel composed of the output channels , the input channels and the kernel size ; .

[0039] Note that additivity is only satisfied when two convolutions have the same configuration (e.g., number of channels, kernel size, padding, stride, etc.).

[0040] Based on the above two basic properties, six transformations are summarized: batch normalization (BN), branch addition, depth connection, multi-scale operation, average pooling, and convolutional sequence. The DBB structure in this model uses , and to enhance the original convolutional layer, greatly enriching the feature space by fusing branches of multiple scales and complexities, and improving the performance of the model without additional inference time cost.

[0041] 3) Replace the original CIoU loss function with the EIoU loss function: YOLOv8n uses CIoU as the BBox regression loss, which adds the center point distance and aspect ratio on the basis of IoU to provide a more comprehensive metric. However, the second penalty term of CIoU only constrains the aspect ratio similarity and ignores the independent differences in width and height, resulting in a penalty of 0 when matching the aspect ratio and limiting the optimization effect of the model. While EIOU separates the influence factors of the aspect ratios of the predicted box and the ground truth box on the basis of the penalty term of CIoU, and calculates the width and height of the predicted box and the ground truth box separately. This loss function consists of three parts: IoU loss, distance loss, and width-height loss. The first two parts follow the methods in CIoU, but the width-height loss directly minimizes the differences in the width and height of the predicted box and the ground truth box, enabling the model to have a faster convergence speed and higher regression accuracy, providing a more accurate fit. As Figure 14 shown, the EIoU loss function can be calculated by formula (8): In formula (8) , , represent the IoU loss, distance loss, and width-height loss respectively. is the Euclidean distance between the center points of the ground truth box and the predicted box, and represent the width and height differences between the ground truth box and the predicted box respectively. and respectively represent the width and height of the minimum bounding rectangle of the predicted bounding box and the ground truth bounding box.

[0042] In step S15, the process of training the constructed GDE-YOLOv8n model using the training set to obtain the detection model is as follows: The input image is subjected to feature extraction using the Backbone backbone network to obtain input features at three scales; in the Neck neck network, the feature pyramid (FPN) and path aggregation network (PAN) are used to fuse the input features at three scales to obtain feature output maps at three scales; the Detection Head detection head performs final regression and classification operations on the feature output maps at three scales, predicts the bounding box regression values and the confidence of the presence of the target for each anchor box, and adjusts the anchor box using the regression prediction values to generate the final prediction box.

[0043] The method for extracting features from the input data using the Backbone backbone network is as follows: Refer to Figure 10 As shown, the input image first passes through the entry convolutional layers of the first and second layers to extract low-level feature information (edges, textures), and then these feature information are passed to the C2f module (layers 3 to 5) to fuse features at different levels, and output large-size feature information X5. The C2f module splits the input features into two parts, one part directly short-circuits (Shortcut) to transfer shallow information; the other part extracts deep features through multiple Bottleneck residual blocks; finally, the shallow and deep features are fused through Concat to enhance the gradient flow. Then, the same extraction operation of the Conv convolutional layer (layer 6) + C2f module (layer 7) combination is adopted to fuse more detailed features and output medium-size feature information X4. Finally, through the deeper feature extraction operation of the same Conv convolutional layer (layer 8) + C2f module (layer 9), combined with the serial pooling operation of the SPPF layer (layer 10) to fuse multi-scale context information, small-size feature information X3 is output. The SPPF module sequentially performs multiple MaxPool layers on the input feature map to generate multi-scale features, and splices (Concat) the pooling results at different scales with the original features to improve the robustness of the model to object sizes. The Backbone network outputs small-size feature information X3, medium-size feature information X4, and large-size feature information X5.

[0044] The method for fusing the input features at three scales in the Neck neck network is as follows: Refer to Figure 11As shown, the Neck network is composed of FPN and PAN. FPN transmits deep feature semantics from top to bottom, and PAN transmits deep target localization from bottom to top. During the bottom-up transmission of PAN, the small-size feature information X3 is input into the first Upsample (layer 11) upsampling module, and the resolution of the feature map is improved at low cost through a simple interpolation operation. It cooperates with the Concat module (layer 12) to splice with the medium-size feature information X4. Finally, the C2f-DBB module (layer 13) improved by DBB performs multi-scale feature fusion, enhancing the small-target detection ability of the model while retaining the original information, and transmits the output feature P to layer 14 and layer 18 respectively. Then, it successively undergoes the same operations of Upsample (layer 14) upsampling + Concat (layer 15) splicing + C2f-DBB module (layer 16), and aggregates the feature P with the large-size feature information X5 to obtain the large-size feature output map Y5. During the top-down transmission of FPN, the large-size feature output map Y5 is halved in size through the Conv module (layer 17), aggregates with the feature P through the Concat module (layer 18), and finally obtains the medium-size feature output map Y4 through the C2f-DBB module (layer 19). Similarly, the medium-size feature output map Y4 is halved in size through the Conv module (layer 20), aggregates with the feature X3 through the Concat module (layer 21), fuses the large, medium, and small-scale feature information through the C2f-DBB module (layer 22), and finally passes through the GAM module (layer 23) to improve the recognition performance of the model for small-scale detection boxes closer to the size of tea leaves, obtaining the small-size feature output map Y3.

[0045] The method for the Detection Head to perform the final regression and classification operations on the feature output maps of three scales is as follows: The Detection Head adopts a decoupled head structure that separates the regression task and the classification task (VFL and DFL). VFL is used as the classification loss function to balance the weights of the target and the background in the small-target detection training and improve the class probability of the predicted target; DFL combines EIOU as the regression loss function to quickly converge to the adjacent area of the target localization and obtain the position information of the predicted bounding box. See Figure 11 As shown, the Detection Head receives the three input features Y5, Y4, and Y3 with different resolutions from the Neck network (FPN + PAN), corresponding to the detection of large, medium, and small targets respectively. The features of each scale are independently predicted by the decoupled head, and finally, the outputs of different scales are weighted and fused to achieve full coverage detection of cross-size targets. At the same time, the Anchor-Free design is used to directly regress the bounding box.

[0046] In step S15, the trained GDE-YOLOv8n model is compared with other object detection algorithms, and the experimental results are shown in Table 1: Table 1 Comparative study of the GDE-YOLOv8n model and other object detection algorithms In Table 1, the accuracy rate represents the ratio of the number of correctly predicted samples in the test set to the number of samples in the predicted values, and also represents the possibility of a sample belonging to a certain class. The recall rate represents the ratio of the number of correctly predicted samples in the test set to the number of samples in the true values, and represents the probability of correctly identifying a certain class. The mean average precision refers to the average precision of all classes during the training process of the neural network model, indicating the overall performance of the model on all classes. To ensure accuracy, the IoU threshold of the mean average precision evaluation index is set to 0.5 to comprehensively evaluate the performance of the model. In addition, the parameter quantity index reflects the performance of the model in terms of resource occupancy, the floating-point operation number index reflects the performance of the model in terms of computational efficiency, the model size index reflects the performance of the model in terms of deployment feasibility, and the frames per second index reflects the performance of the model in terms of detection speed. The calculation formula for the accuracy rate is shown in Equation (9), the calculation formula for the recall rate is shown in Equation (10), the calculation formula for the mean average precision is shown in Equation (11), and the calculation formula for the average precision is shown in Equation (12): In Equations (9) to (12), the true positive is the number of samples correctly predicted by the model as tea leaf diseases, the false positive is the number of samples wrongly predicted by the model as tea leaf diseases, the false negative is the number of samples wrongly predicted by the model as the background, and the true negative is the number of samples correctly predicted by the model as the background. represents the number of classes, in this study the value is set to 3. Where p(r) is the accuracy value corresponding to the recall rate r.

[0047] From the experimental results in Table 1, it can be seen that the recognition accuracy of the GDE-YOLOv8n model reaches 91.7%, which is the best performance among all models, and it is 3.1% higher than the traditional YOLOv8n model. In terms of the parameter quantity, floating-point operation number, and model size of the model, the GDE-YOLOv8n has only a small improvement compared to the YOLOv8n, and the gap can be ignored. It is worth noting that the detection speed reaches 80.0 FPS, which is much higher than the 4.8 FPS required for real-time detection scenarios and has real-time performance. In summary, the GDE-YOLOv8n has better overall performance.

[0048] In step S16, NVIDIA Jetson Orin Nano is a high-performance edge computing platform designed specifically for AI and machine learning applications. It integrates an NVIDIA Ampere architecture GPU with 1024 CUDA cores and 32 Tensor cores; is equipped with a 6-core ARM Cortex-A78AE CPU with a maximum clock frequency of 1.5 GHz; has 8GB of 128-bit LPDDR5 memory, and the development board contains 4 USB interfaces, 1 HDMI interface, 1 RJ45 network interface, and 40 I / O ports. Various peripherals can be connected through these interfaces. Given its compact size and excellent energy efficiency, the GDE-YOLOv8n model is deployed to this device for real-time detection and localization of tea diseases in the tea garden, and for processing and analyzing disease information, providing important technical support for the subsequent control of the spraying execution system.

[0049] In the spraying execution system, the number of solenoid valves and nozzles is 4 each. The solenoid valve is a 12V DC pressure-holding normally closed copper valve, and its working pressure range is 0.02 - 0.8 Mpa. A fine-tuning switch is equipped at the rear end of the nozzle pipeline, and the amount of sprayed medicine can be adjusted by rotating the switch. The spraying pipeline selects a PU pneumatic transparent hose, and the working condition of the spraying pipeline can be directly observed during operation.

[0050] In the monitoring system, the liquid level sensor measures the amount of medicine in the medicine tank through exposed parallel wires and converts it into an analog signal, which can be collected and read by the main control system to achieve medicine amount monitoring. Its working voltage is DC3 - 5V, and the working current is less than 20 mA.

[0051] The intelligent and precise tea garden spraying system based on machine vision included in the method is as follows: It includes an electric motor 1, a frame 2, traveling wheels 3, a gearbox 4, a handrail device 11, and a precise spraying device 12; The electric motor 1 is installed behind the frame 2, the frame 2 is welded to the gearbox 4, and the handrail device 11 is installed above the gearbox 4; The described precise spraying device 12 includes an identification camera 12-1, a camera bracket 12-2, a medicine box bracket 12-3, a medicine box 12-4, an electric pump 12-5, a solenoid valve 12-6, a spraying bracket 12-7, a spraying hose 12-8, a nozzle 12-9, an outer lateral sliding guide rail 12-10, an inner lateral sliding guide rail 12-11, a longitudinal sliding guide rail 12-12, and an NVIDIA Jetson Orin Nano main control system 12-13. The camera bracket 12-2 is welded to the front end of the frame 2. The identification camera 12-1 is installed at both ends of the camera bracket 12-2 for identifying tea pests and diseases. The liquid medicine is transported to the solenoid valve 12-6 by the electric pump 12-5. Both ends of the spraying hose 12-8 are respectively connected to the solenoid valve 12-6 and the nozzle 12-9. One outer lateral sliding guide rail 12-10, one inner lateral sliding guide rail 12-11, and two longitudinal sliding guide rails 12-12 are installed on each side of the spraying bracket 12-7. The end of the extension section of the outer lateral sliding guide rail 12-10 is connected to a longitudinal sliding guide rail 12-12, and the sliding block of the inner lateral sliding guide rail 12-11 is connected to a longitudinal sliding guide rail 12-12. The NVIDIA Jetson Orin Nano main control system 12-13 is used to receive and analyze the information on the tea disease area detected by the identification camera 12-1, and control the nozzle 12-9 to spray medicine on the disease area.

[0052] The spraying method based on the above system is as follows: Use the identification camera 12-1 to collect on-site pictures of the tea garden; Transmit the on-site pictures of the tea garden collected by the identification camera 12-1 to the NVIDIA Jetson Orin Nano main control system, and detect them through the detection model obtained in step S15 to determine whether there are pests and diseases. If so, calculate the location where the pests and diseases occur, control the nozzle 12-9 to move above the location where the pests and diseases occur through the outer lateral sliding guide rail 12-10, the inner lateral sliding guide rail 12-11, and the longitudinal sliding guide rail 12-12, and control the nozzle 12-9 to perform spraying treatment by controlling the opening of the solenoid valve 12-6. If not, there is no need to turn on the nozzle 12-9 for spraying treatment.

[0053] In summary, the intelligent precise spraying system and spraying method for tea gardens based on machine vision provided by the present invention can achieve precise identification and precise spraying of tea garden pests and diseases, and perform pest and disease identification and spraying treatment during the processes of trenching, fertilizing, and soil covering in the tea garden, making the intelligent precise spraying system for tea gardens a system that combines the functions of trenching, fertilizing, soil covering, and precise spraying. It can not only reduce the amount of pesticide used, reduce environmental pollution, but also save energy and protect the environment, and is simple to operate, lightweight and convenient.

[0054] The above are only embodiments of the present invention, and thus do not limit the patent scope of the present invention. Any equivalent transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in the relevant technical fields, shall similarly be included within the patent protection scope of the present invention.

Claims

1. An intelligent and precise tea garden spraying system based on machine vision, characterized in that, It includes an electric motor (1), a frame (2), traveling wheels (3), a gearbox (4), an armrest device (11), and a precise pesticide spraying device (12); The electric motor (1) is installed at the rear of the frame (2), the frame (2) is welded to the gearbox (4), and the armrest device (11) is installed above the gearbox (4); The precise pesticide spraying device (12) includes an identification camera (12-1), a camera bracket (12-2), a medicine box bracket (12-3), a medicine box (12-4), an electric pump (12-5), a solenoid valve (12-6), a pesticide spraying bracket (12-7), a pesticide spraying hose (12-8), a nozzle (12-9), an outer lateral sliding guide rail (12-10), an inner lateral sliding guide rail (12-11), a longitudinal sliding guide rail (12-12), and an NVIDIA Jetson Orin Nano main control system (12-13). The camera bracket (12-2) is welded to the front end of the frame (2), and the identification camera (12-1) is installed at both ends of the camera bracket (12-2) for identifying tea pests and diseases. The liquid medicine is transported to the solenoid valve (12-6) by the electric pump (12-5). The two ends of the pesticide spraying hose (12-8) are respectively connected to the solenoid valve (12-6) and the nozzle (12-9). One outer lateral sliding guide rail (12-10), one inner lateral sliding guide rail (12-11), and two longitudinal sliding guide rails (12-12) are installed on each side of the pesticide spraying bracket (12-7). One longitudinal sliding guide rail (12-12) is connected to the end of the extension section of the outer lateral sliding guide rail (12-10), and one longitudinal sliding guide rail (12-12) is connected to the sliding block of the inner lateral sliding guide rail (12-11). The NVIDIA Jetson Orin Nano main control system (12-13) is used to receive and analyze the information of the tea disease area detected by the identification camera (12-1), and control the nozzle (12-9) to spray pesticides on the disease area.

2. The intelligent and precise tea garden spraying system based on machine vision according to claim 1, wherein It also includes an inclined ditching device (6) and a belt drive device (10); The electric motor (1) transmits power to the gearbox (4) through the belt drive device (10). The gearbox (4) is provided with an output shaft one and an output shaft two. The output shaft one of the gearbox (4) is connected to the traveling wheels (3) to drive the traveling wheels (3) to rotate; The inclined ditching device (6) includes a chain transmission box (6-1), a transmission shaft (6-2), a cross coupling (6-3), a rotary tillage device frame (6-4), a rotary tillage transmission mechanism (6-5), and a rotary tillage mechanism (6-6). A bottom plate is installed below the chain transmission box (6-1), and the front end of the frame (2) is fixedly installed by bolts. Each of the left and right sides of the chain transmission box (6-1) is connected and driven to a rotary tillage transmission mechanism (6-5) through a cross coupling (6-3). Each of the two rotary tillage transmission mechanisms (6-5) is connected to a rotary tillage mechanism (6-6). The output shaft II of the gearbox (4) transmits power to the chain transmission box (6-1) through chain transmission, and drives the cross coupling (6-3), the rotary tillage transmission mechanism (6-5), and the rotary tillage mechanism (6-6) to operate; The axial direction of the transmission shaft (6-2) forms an angle of 70° with the axis of the rotary tillage transmission mechanism (6-5). The axis of the rotary tillage transmission mechanism (6-5) forms an angle greater than 70° and less than 90° with the forward direction. The rotary tillage device frame (6-4) covers the outside of the cross coupling (6-3). The rotary tillage device frame (6-4) is fixedly connected to the chain transmission box (6-1) and the rotary tillage transmission mechanism (6-5) through bolts.

3. The intelligent and precise tea garden spraying system based on machine vision according to claim 2, wherein, The rotary tillage mechanism (6-6) includes a ditching and soil throwing knife (6-61), a mounting cutter disc (6-62), a rotary tillage cutter shaft (6-63), and a soil throwing iron sheet (6-64). The rotary tillage mechanisms (6-6) are installed in a left-right opposed manner. When viewed from the front, the included angle formed by the planes where the mounting cutter discs (6-62) on the left and right sides are located is an acute angle greater than 20° and less than 40°. Five arc-shaped ditching and soil throwing knives (6-61) are circumferentially distributed on each mounting cutter disc (6-62). A circular soil throwing iron sheet (6-64) is welded on the back of the ditching and soil throwing knife (6-61). A rotary tillage cutter shaft (6-63) is coaxially arranged and fixedly connected to a mounting cutter disc (6-62). A rotary tillage cutter shaft (6-63) is fixed on the output shaft of a rotary tillage transmission mechanism (6-5).

4. A spraying method for the intelligent precise spraying system of a tea garden based on machine vision according to any one of claims 1-3, characterized in that, The method is as follows: Step S1: Construct a tea disease detection model, specifically: Step S11: Collect a tea disease dataset containing different environments and types; Step S12: Label the tea disease dataset collected in Step S11, and divide it into a training set, a validation set, and a test set according to a ratio; Step S13: Perform data augmentation and expansion on the training set in Step S12; Step S14: Construct a tea disease detection model based on GDE-YOLOv8n: Introduce a global attention mechanism (GAM) at the end of the Neck of the original YOLOv8n model, use a diverse branch block (DBB) in combination with the C2f module in the Neck network of the original YOLOv8n model, and replace the original CIoU loss function with an EIoU loss function; Step S15: Train and experiment on the model constructed in Step S14 to obtain a detection model; Step S16: Deploy the detection model obtained in Step S15 to run on the NVIDIA Jetson Orin Nano main control system; Step S2: Collect pictures of the tea garden site; Step S3: Transmit the pictures of the tea garden site collected in Step S2 to the NVIDIA Jetson Orin Nano main control system, and perform detection using the detection model obtained in Step S1 to determine whether there are pests and diseases: If so, calculate the location where the pests and diseases occur and then perform spraying treatment.

5. The spraying method of the intelligent and precise tea garden spraying system based on machine vision according to claim 4, wherein In Step S11, collect a tea leaf disease dataset containing different environments and species, specifically: collect pictures of tea leaf pests and diseases in different regions, different types of tea leaf pests and diseases, and different environmental conditions.

6. The spraying method of the intelligent precise spraying system for tea gardens based on machine vision according to claim 4, wherein In Step S12, divide the training set, validation set, and test set and perform annotation, specifically: divide the images collected in Step S11 into the training set, validation set, and test set according to the ratio of 7:2:1, and use the LabelImg picture annotation tool to annotate the training set images in TXT format to form label files.

7. The spraying method of the intelligent and precise tea garden spraying system based on machine vision according to claim 4, characterized in that In Step S13, perform data augmentation operations on the training set divided and annotated in Step S12, specifically: perform transformations on the training set in Step S12, including brightness conversion, noise addition, and random angle rotation, and finally expand the original dataset four times.

8. The spraying method of the intelligent and precise tea garden spraying system based on machine vision according to claim 4, characterized in that, Step S15: Train and experiment on the model constructed in Step S14 to obtain a detection model, specifically: use the Backbone network to extract features from the input image to obtain input features at three scales; in the Neck network, use the Feature Pyramid Network (FPN) and Path Aggregation Network (PAN) to fuse the input features at three scales to obtain feature output maps at three scales; the Detection Head performs final regression and classification operations on the feature output maps at three scales, predicts the bounding box regression values and the confidence of the target presence for each anchor box, and uses the regression prediction values to adjust the anchor boxes to generate the final prediction boxes.

9. The spraying method of the intelligent and precise tea garden spraying system based on machine vision according to claim 4, characterized in that, The method further includes: in Step S2, use the intelligent precise spraying system for tea gardens based on machine vision to collect pictures of the tea garden site; In Step S3, after calculating the location where the pests and diseases occur, use the intelligent precise spraying system for tea gardens based on machine vision to perform spraying treatment.

Images