Tomato leaf trialeurodes vaporariorum insect pest pesticide application method based on visual detection

Abstract

The invention discloses a tomato leaf trialeurodes vaporariorum insect pest pesticide application method based on visual detection, which comprises the following steps: in the moving process of a chassis of an automatic pesticide application device, detecting the disease part of a tomato leaf through a trained YOLOv9 network, if a tomato leaf insect pest area is detected, stopping moving, then carrying an industrial fan on a leaf disturbance module, and carrying out automatic pesticide application on the tomato leaf insect pest area. The pesticide spraying module performs leaf disturbance on the identified insect pest area to expose trialeurodes vaporariorum on the back of the leaf, and meanwhile, the pesticide spraying module processes insect pest position area information, rotates the direction of a spray head, aligns the direction of the spray head to the leaf insect pest area, and then accurately applies pesticide to the tomato leaf infected with the trialeurodes vaporariorum; therefore, the purpose of treating trialeurodes vaporariorum parasitizing on the backs of tomato leaves is achieved. The problems that a traditional tomato trialeurodes vaporariorum pest control method depends on a manual backpack sprayer, a large amount of pesticide liquid is sprayed, the air humidity in a greenhouse is remarkably improved, multiple hygrophilous germs are bred, and diseases such as gray mold, early blight, late blight and leaf mold are caused are avoided.

Description

Technical Field

[0001] The invention belongs to the technical field of visual detection control, and in particular to a method for applying pesticides to whitefly pests on tomato leaves based on visual detection. Background Art

[0002] Whiteflies are small in size and have strong reproductive capacity. They usually appear on the back of leaves where pesticides are less distributed. Modern greenhouses provide them with an ideal living environment. Whiteflies suck crop juice, which will lead to nutrient loss and virus transmission, affecting tomato yield and quality. Since whiteflies often attach to the back of leaves, frequent and large-scale spraying of liquid medicine is required to effectively control them. The traditional method of tomato whitefly pest control relies on manual backpack sprayers. Although it can effectively kill whiteflies, a large amount of liquid medicine spraying will significantly increase the air humidity in the greenhouse, and then breed a variety of moisture-loving pathogens, causing diseases such as gray mold, early blight, late blight, and leaf mold. In addition, the traditional method has low operating efficiency, high labor intensity, difficult working conditions, uneven spraying, and the effective utilization rate of liquid medicine is usually less than 10%, resulting in a large amount of liquid medicine waste and pollution to the environment. There is also a risk of poisoning for sprayers. Therefore, it is particularly important to develop a spraying method that is efficient, uses less medicine, and is precisely attached to the back of leaves to reduce the burden on farmers and reduce the negative impact of liquid medicine on the environment and plants. In recent years, target detection technology has developed rapidly and has gradually been applied to agriculture. Combining target detection technology with automatic spraying for plant disease and pest control has become a research focus. Summary of the invention

[0003] The present invention provides a method for applying pesticides to tomato leaves whitefly pests based on visual detection, so as to solve the problems existing in the above-mentioned background technology.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is: a method for applying pesticides to tomato leaves against whitefly pests based on visual detection, which specifically comprises the following steps:

[0005] Step S1: constructing an automatic pesticide application device for tomato whitefly pests, integrating a mobile module, a data acquisition module, a visual inspection module, a leaf disturbance module and a spraying module;

[0006] Step S2: YOLOv9 network training;

[0007] Step S3: placing an automatic pesticide application device for tomato whitefly pests between tomato ridges, and controlling the automatic pesticide application device to move forward through a controller;

[0008] Step S4: The data acquisition module collects leaf data between tomato ridges in real time, and transmits the collected data to the visual detection module, and the leaf data is detected in real time through the trained YOLOv9 network carried by the visual detection module;

[0009] Step S5: When tomato leaves infected with whiteflies are detected, a signal is transmitted to a control module in the industrial computer to stop the vehicle from moving;

[0010] Step S6: obtaining the position coordinates of the tomato leaves infected with whiteflies in the image by calculation, transmitting the position coordinate information to the spraying module, and rotating the spray head to aim at the target leaves;

[0011] Step S7: Start the leaf disturbance module, and the flowing gas will disturb the tomato leaves infected with whiteflies, exposing the back of the leaves to the spraying module;

[0012] Step S8: The spraying module sprays the processed medicine gelatin mixed emulsion on the target leaves; 5 seconds after spraying, the automatic spraying device will continue to travel along the ridges, continue to detect other leaves and automatically spray.

[0013] Preferably, in step S2, the YOLOv9 network training specifically includes the following steps:

[0014] Step S20: collecting data of tomato leaves infected with whiteflies as a data set for YOLOv9 network training, including images of leaves infected with whiteflies in the conditions of non-overlapping and good light, overlapping and good light, non-overlapping and poor light, and overlapping and poor light; ensuring the diversity of the data set;

[0015] Step S21: Based on the Pycharm platform, a leaf random splicing transformation enhancement method is used to perform data enhancement on the collected image to improve the robustness and generalization ability of the detection network model;

[0016] Step S22: Labelimg is used to annotate the data set, and xml tags are generated at the same time. The xml tags are converted into txt tags using Python scripts, and the data set is divided into a training set, a validation set, and a test set in a ratio of 8:1:1;

[0017] Step S23: Adjust the parameters in the YOLOv9 file, set the image input size to 640 pixels × 640 pixels, batch to 16, workers to 0, devices to 0, maximum number of iterations to 300, initial learning rate to 0.01, and modify the image input path and the path for saving the training results;

[0018] Step S24: Run the train.py file to train the data set, and obtain the trained weight best.pt after the training is completed;

[0019] Step S25: In the detect file, modify the weight path to the path of best.pt, import the test set in the above data set into the network, and check the detection effect;

[0020] Step S26: Import the trained network into the embedded EPC-S101 industrial computer.

[0021] Preferably, in step S21, the data enhancement method specifically includes:

[0022] a. Random stitching: First, the original image is randomly cropped and stitched to generate new image samples; this method can simulate the different perspectives and partial occlusions of leaves in the real world, thereby improving the model's ability to adapt to new perspectives;

[0023] b. Contrast adjustment: By changing the contrast of the image, the visual scenes under different lighting conditions are simulated; this helps the training model maintain stable recognition performance under various lighting environments;

[0024] c. Brightness change: Brightness adjustment can further increase the diversity of image samples, simulate image features at different times, such as day, night, or under different weather conditions, and enable the model to better generalize to various practical application scenarios;

[0025] d. Noise addition: Adding random noise to the image simulates the interference encountered during real-world image acquisition, increasing the complexity of the dataset.

[0026] Preferably, in step S1, the mobile module is a front-drive mobile chassis based on a STM32F103 microprocessor, and its core components mainly include a control handle and a VCU controller, which are used for the smooth movement of the automatic spraying device between tomato ridges.

[0027] Preferably, in step S1, the data acquisition module is an Intel RealSense D455 depth camera, which is used to capture images of tomato leaves in the greenhouse.

[0028] Preferably, in step S1, the visual inspection module is an embedded EPC-S101 industrial computer, in which a trained YOLOv9 network is implanted for data processing and analysis of the collected images to detect the location of pests and diseases and locate the pests and diseases.

[0029] Preferably, in step S1, the blade disturbance module is a small industrial fan, which is used to disturb the tomato leaves so that the back of the tomato leaves infected with whiteflies are exposed to the spray head, so that the device can apply pesticides on the back of the leaves.

[0030] Preferably, in step S1, the spraying module is composed of a pesticide bottle, a gelatin bottle, an air pump and a nozzle, so that the pesticide and gelatin are gathered in the nozzle, and a slow-release liquid mixed with high-efficiency chlorflucyanate and gelatin is formed for precise spraying of tomato leaves.

[0031] The beneficial effects of adopting the above technical solution are:

[0032] 1. A method for applying pesticides to tomato leaves for whitefly pests based on visual detection is provided by the present invention. A chassis trolley equipped with an automatic pesticide application device moves between tomato ridges. During the movement, the diseased part of the tomato leaves is detected by using a trained YOLOv9 network. If a tomato leaf pest area is detected, the movement will be stopped. Then, an industrial fan is installed on the blade disturbance module. The target tomato leaves are flipped by adjusting parameters such as the fan height, angle and wind force. The identified pest area is subjected to blade disturbance to expose the whiteflies parasitic on the back of the leaves. At the same time, the spraying module rotates the direction of the nozzle within a certain range by processing the information of the pest location area. After the nozzle direction is aligned with the leaf pest area, the tomato leaves infected with whiteflies are accurately sprayed with pesticides, thereby achieving the purpose of controlling the whiteflies parasitic on the back of the tomato leaves; the traditional tomato whitefly pest control method is avoided, which relies on a manual backpack sprayer, spraying a large amount of liquid medicine, significantly increasing the air humidity in the greenhouse, resulting in the breeding of a variety of moisture-loving pathogens, and causing diseases such as gray mold, early blight, late blight, and leaf mold.

[0033] 2. The present invention provides a method for applying pesticides to tomato leaves for whitefly pests based on visual detection, which can realize automatic identification of the application liquid. Not only can the liquid be sprayed more evenly, the effective utilization rate of the liquid can be improved, the waste and loss of the liquid and the pollution to the environment can be reduced, but the operation efficiency is also high, which greatly reduces the labor intensity of farmers and reduces the risk of poisoning for sprayers.

[0034] 3. The present invention provides a method for applying pesticides to tomato leaves for whitefly pests based on visual detection, which uses a gelatin liquid emulsion, better adapts to the leaf disturbance module, and effectively increases the adhesion rate of pesticides on the back of leaves infected with whiteflies. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is an assembly diagram of the automatic pesticide application device for tomato whitefly pests of the present invention;

[0036] Figure 2 It is an indication of leaf disturbance and spraying;

[0037] Figure 3 This is the effect diagram of some images after data enhancement;

[0038] Figure 4 This is the detection effect diagram of the YOLOv9 network;

[0039] Figure 5 The heat map obtained by using the Grad-CAM heat map visualization method to visualize the heat map of the seventh layer in the YOLOv9 network;

[0040] Figure 6 The workflow diagram of the method of applying pesticides to tomato leaves for whitefly pests based on visual detection;

[0041] 1. Blade disturbance module; 2. Spraying module; 3. Data acquisition module; 4. Mobile module; 5. Visual inspection module. DETAILED DESCRIPTION

[0042] The specific implementation methods of the present invention are further explained in detail below by describing the embodiments with reference to the accompanying drawings, with the aim of helping those skilled in the art to have a more complete, accurate and in-depth understanding of the concept and technical solution of the present invention and facilitating its implementation.

[0043] like Figures 1 to 6 As shown, the present invention is a method for applying pesticides to tomato leaves for whitefly pests based on visual detection. A chassis trolley equipped with an automatic pesticide application device moves between tomato ridges. During the movement, the diseased parts of tomato leaves are detected by using a trained YOLOv9 network. If a tomato leaf pest area is detected, the movement will stop. Then, an industrial fan is installed on the blade disturbance module 1. By adjusting the fan height, angle, wind speed and other parameters, the target tomato leaves are flipped, and the identified pest area is disturbed to expose the parasites on the back of the leaves. At the same time, the spraying module 2 processes the pest location information and rotates the nozzle direction within a certain range. After the nozzle direction is aimed at the leaf pest area, the tomato leaves infected with whiteflies are precisely sprayed, thereby achieving the purpose of controlling the whiteflies parasitic on the back of tomato leaves; avoiding the traditional tomato whitefly pest control method that relies on manual backpack sprayers, spraying a large amount of liquid medicine, significantly increasing the air humidity in the greenhouse, causing the breeding of a variety of moisture-loving pathogens, and causing gray mold, early blight, late blight, leaf mold and other diseases. .

[0044] The specific working method is described in the following embodiments:

[0045] Embodiment 1:

[0046] A method for applying pesticides to tomato leaves against whitefly pests based on visual detection, specifically comprising the following steps:

[0047] Step S1: constructing an automatic pesticide application device for tomato whitefly pests, integrating a mobile module 4, a data acquisition module 3, a visual detection module 5, a leaf disturbance module 1 and a spraying module 2;

[0048] Step S2: YOLOv9 network training;

[0049] Step S20: collecting data of tomato leaves infected with whiteflies as a data set for YOLOv9 network training, including 400 images of leaves infected with whiteflies in the conditions of non-overlapping and good light, overlapping and good light, non-overlapping and poor light, and overlapping and poor light, to ensure the diversity of the data set;

[0050] Step S21: Based on the Pycharm platform, a leaf random splicing transformation enhancement method is used to perform data enhancement on the collected images; the number of images is expanded to 800; and the robustness and generalization ability of the detection network model are improved;

[0051] In step S21, the data enhancement method specifically includes:

[0052] a. Random stitching: First, the original image is randomly cropped and stitched to generate new image samples; this method can simulate the different perspectives and partial occlusions of leaves in the real world, thereby improving the model's ability to adapt to new perspectives;

[0053] b. Contrast adjustment: By changing the contrast of the image, the visual scenes under different lighting conditions are simulated; this helps the training model maintain stable recognition performance under various lighting environments;

[0054] c. Brightness change: Brightness adjustment can further increase the diversity of image samples, simulate image features at different times, such as day, night, or under different weather conditions, and enable the model to better generalize to various practical application scenarios;

[0055] d. Noise addition: Adding random noise to the image simulates the interference encountered in the real-world image acquisition process, increasing the complexity of the data set.

[0056] The effect of this data enhancement method is shown in the figure Figure 3 shown.

[0057] Step S22: Use Labelimg to annotate the dataset and generate xml tags. Use Python script to convert the xml tags into txt tags and divide the dataset into training set train, validation set val and test set text in a ratio of 8:1:1, which are 640, 80 and 80 images respectively.

[0058] Step S23: Adjust the parameters in the YOLOv9 file, set the image input size to 640 pixels × 640 pixels, batch to 16, workers to 0, devices to 0, maximum number of iterations to 300, initial learning rate to 0.01, and modify the image input path and the path for saving the training results;

[0059] Step S24: Run the train.py file to train the data set, and obtain the trained weight best.pt after the training is completed;

[0060] Step S25: In the detect file, modify the weight path to the path of best.pt, import the test set text in the above data set into the network, and check the detection effect. The detection effect is as follows Figure 4 As shown;

[0061] Step S26: Import the trained network into the embedded EPC-S101 industrial computer.

[0062] Step S3: placing an automatic pesticide application device for tomato whitefly pests between tomato ridges, and controlling the automatic pesticide application device to move forward through a controller;

[0063] Step S4: the data acquisition module 3 collects leaf data between tomato ridges in real time, and transmits the collected data to the visual detection module 5, and the leaf data is detected in real time through the trained YOLOv9 network carried by the visual detection module 5;

[0064] Step S5: When tomato leaves infected with whiteflies are detected, a signal is transmitted to a control module in the industrial computer to stop the vehicle from moving;

[0065] Step S6: obtaining the position coordinates of the tomato leaves infected with whiteflies in the image by calculation, transmitting the position information to the spraying module 2, and rotating the spray head to aim at the target leaves;

[0066] Step S7: Start the leaf disturbance module 1. The flowing gas will disturb the tomato leaves infected with whiteflies detected by the YOLOv9 network, exposing the back of the leaves to the spraying module 2. The leaf disturbance and spraying demonstration is as follows: Figure 2 shown.

[0067] Step S8: the spraying module 2 sprays the processed medicine gelatin mixed emulsion on the target leaves; 5 seconds after spraying, the automatic spraying device will continue to travel along the ridges, continue to detect other leaves and automatically spray.

[0068] Embodiment 2:

[0069] The automatic pesticide application device for tomato whitefly pests of the present invention is as follows: Figure 1As shown, a mobile module 4, a data acquisition module 3, a visual detection module 5, a leaf disturbance module 1 and a spraying module 2 are integrated to achieve real-time detection and intelligent management of whitefly pests in complex environments. Among them, the mobile module 4 is used for the smooth movement of the device between tomato ridges; the data acquisition module 3 is used to collect images of tomato leaves in the greenhouse; the visual detection module 5 is used to process and analyze the collected images to detect the location of pests and diseases and realize the location of pests and diseases; the leaf disturbance module 1 is used to disturb the tomato leaves so that the back of the tomato leaves infected with whiteflies are exposed to the nozzle; the spraying module 2 is used to accurately spray the tomato leaves.

[0070] The chassis trolley equipped with an automatic pesticide application device moves between tomato ridges. During the movement, the diseased parts of tomato leaves are detected by using the trained YOLOv9 network. If the pest area of ​​tomato leaves is detected, the movement will stop. Then, an industrial fan is installed on the blade disturbance module 1. The target tomato leaves are flipped by adjusting the fan height, angle, wind force and other parameters, and the blades of the identified pest areas are disturbed to expose the whiteflies parasitic on the back of the leaves. At the same time, the spraying module 2 will rotate the direction of the nozzle within a certain range by processing the information of the pest location area. After the nozzle direction is aimed at the leaf pest area, the tomato leaves infected with whiteflies are accurately sprayed with pesticides, thereby achieving the purpose of controlling the whiteflies parasitic on the back of tomato leaves; the traditional tomato whitefly pest control method is avoided by relying on manual backpack sprayers, and a large amount of liquid spraying significantly increases the air humidity in the greenhouse, resulting in the breeding of a variety of moisture-loving pathogens, causing gray mold, early blight, late blight, leaf mold and other diseases.

[0071] Mobile Module

[0072] The mobile module 4 is a front-drive mobile chassis based on the STM32F103 microprocessor, and its core components mainly include a control handle and a VCU controller.

[0073] Among them, the control handle is used to realize the control of the mobile module 4. In addition, the VCU vehicle controller is used as the central nervous system of the mobile module 4 to realize the collection of input signals, the output of load control signals, and the management and coordination of the vehicle ecosystem, including the power management unit (PMU), motor control unit (MCU), brake control unit (BCU), steering control unit (SCU) and other subsystems. It collects the status information of each component and integrates and analyzes this information to determine the status and operation of the mobile module 4. In addition, the VCU controller can perform real-time fault diagnosis and early warning by real-time monitoring of the working status of the agricultural machinery, sensor data, actuator feedback and other information, and record and store various data during the operation of the agricultural machinery, such as operation time, location, operation parameters, etc., for subsequent operation analysis and decision support, helping farmers and agricultural professionals to optimize and manage agricultural machinery operations.

[0074] Among them, MCU is an STM32F103 microprocessor, which obtains the control command sent by the vehicle motion control handle and converts the command data into motor motion control data, and controls the movement of the vehicle motor through the motor driver to achieve speed and rotation control of the mobile module 4.

[0075] The microprocessor runs at a main frequency of 72MHz and has high-performance processing capabilities, capable of processing high-speed data and calculating complex algorithms. At the same time, the microcontroller integrates a wealth of peripherals, including multiple general timers, serial port interfaces, SPI interfaces, I2C interfaces, CAN interfaces, etc., to meet the control requirements of the mobile module 4.

[0076] In addition, a high-power battery is integrated in the mobile module 4, which can provide a suitable operating voltage for each component. In addition, shock-absorbing devices are provided on all four wheels, which greatly improves the stability and operability of the mobile module 4 during operation.

[0077] Data acquisition module

[0078] Data acquisition module 3 is an Intel RealSense D455 depth camera, which increases the distance between depth sensors to 95 mm, and improves the depth error to less than 2% at 4 meters. The depth FOV (height × width) is 87 × 58, and the depth resolution is up to 1280 × 720. In addition, this depth camera integrates an IMU, which enables the application to improve the depth perception capability in any situation where the camera moves to obtain high-quality data.

[0079] Visual inspection module

[0080] Visual inspection module 5 is an embedded EPC-S101 mini industrial computer equipped with Intel 5th generation ATOME8000 processor, fanless ultra-thin system, supporting win7 / 10, windows sever 2012 / 2016 / 2019, Linx system. The trained YOLOv9 network is implanted in the industrial computer to achieve accurate real-time detection of tomato leaves infected with whiteflies.

[0081] Blade disturbance module

[0082] The blade disturbance module 1 uses a small industrial fan to blow the tomato leaves. When the airflow contacts the tomato leaves, the blades will swing and flip. At the same time, the targeted mist droplets of the medicine will be captured by the flipping blades, so as to achieve accurate application and efficient use of the medicine. The present invention studies the fan gas disturbance part based on the knowledge of fluid mechanics, solves the control equation of the fan gas disturbance by numerical discretization method, and finds that the wind speed of 1.5m / s can efficiently disturb the blades without disturbing the spray direction of the medicine and effectively allow the droplets of the medicine to adhere.

[0083] Spraying module

[0084] The spraying module 2 is mainly composed of a pesticide bottle, a gelatin bottle, an air pump and a nozzle, so that the liquid medicine and gelatin are gathered in the nozzle and a slow-release liquid medicine mixed with high-efficiency chlorofluorocyanate and gelatin is formed, which has higher adhesion and can be better adapted to the spraying method.

[0085] The present invention is described above by way of example in conjunction with the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-mentioned method. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention; or the above-mentioned concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A method for applying pesticides to tomato leaves against whitefly pests based on visual detection, characterized in that: The specific steps include: Step S1: constructing an automatic pesticide application device for tomato whitefly pests, integrating a mobile module (4), a data acquisition module (3), a visual detection module (5), a leaf disturbance module (1) and a pesticide spraying module (2); Step S2: YOLOv9 network training; Step S3: placing the automatic pesticide application device for tomato whitefly pests between tomato ridges and controlling the device to move forward through the controller; Step S4: the data acquisition module (3) collects leaf data between tomato ridges in real time, and transmits the collected data to the visual detection module (5). The leaf data is detected in real time through the trained YOLOv9 network carried by the visual detection module (5); Step S5: When tomato leaves infected with whiteflies are detected, a signal is transmitted to a control module in the industrial computer to stop the vehicle from moving; Step S6: obtaining the position coordinates of the tomato leaves infected with whiteflies in the image by calculation, transmitting the position information to the spraying module (2), and rotating the spray head to aim at the target leaves; Step S7: Start the leaf disturbance module (1), and the flowing gas will disturb the tomato leaves infected with whiteflies, exposing the back of the leaves to the spraying module (2); Step S8: The spraying module (2) sprays the treated medicine gelatin mixed emulsion on the target leaves; 5 seconds after spraying, the automatic spraying device will continue to move along the ridges, continue to detect other leaves and automatically spray.

2. A method for applying pesticides to tomato leaves against whitefly pests based on visual detection according to claim 1, characterized in that: In step S2, the YOLOv9 network training specifically includes the following steps: Step S20: collecting data of tomato leaves infected with whiteflies as a data set for YOLOv9 network training, including images of leaves infected with whiteflies in the conditions of no overlap and good light, overlap and good light, no overlap and poor light, and overlap and poor light; Step S21: Based on the Pycharm platform, a leaf random splicing transformation enhancement method is used to perform data enhancement on the collected image; Step S22: Labelimg is used to annotate the data set, and xml tags are generated at the same time. The xml tags are converted into txt tags using Python scripts, and the data set is divided into a training set (train), a validation set (val), and a test set (text), with a ratio of 8:1:1; Step S23: Adjust the parameters in the YOLOv9 file, set the image input size to 640 pixels × 640 pixels, batch to 16, workers to 0, devices to 0, maximum number of iterations to 300, initial learning rate to 0.01, and modify the image input path and the path for saving the training results; Step S24: Run the train.py file to train the data set, and obtain the trained weight best.pt after the training is completed; Step S25: In the detect file, modify the weight path to the path of best.pt, import the test set (text) in the above data set into the network, and check the detection effect; Step S26: Import the trained network into the embedded EPC-S101 industrial computer.

3. A method for applying pesticides to tomato leaves against whitefly pests based on visual detection according to claim 2, characterized in that: In step S21, the data enhancement method specifically includes: a. Random stitching: First, the original image is randomly cropped and stitched to generate new image samples; this method can simulate the different perspectives and partial occlusions of leaves in the real world, thereby improving the model's ability to adapt to new perspectives; b. Contrast adjustment: By changing the contrast of the image, the visual scenes under different lighting conditions are simulated; this helps the training model maintain stable recognition performance under various lighting environments; c. Brightness change: Brightness adjustment can further increase the diversity of image samples, simulate image features at different times, such as day, night, or under different weather conditions, and enable the model to better generalize to various practical application scenarios; d. Noise addition: Adding random noise to the image simulates the interference encountered during real-world image acquisition, increasing the complexity of the dataset.

4. The method for applying pesticides to tomato leaves against whitefly pests based on visual detection according to claim 1, characterized in that: In the step S1, the mobile module (4) is a front-drive mobile chassis based on a STM32F103 microprocessor, and its core components mainly include a control handle and a VCU controller, which are used for the smooth movement of the automatic pesticide application device between tomato ridges.

5. The method for applying pesticides to tomato leaves against whitefly pests based on visual detection according to claim 1, characterized in that: In step S1, the data acquisition module (3) is an Intel RealSense D455 depth camera, which is used to capture images of tomato leaves in the greenhouse.

6. The method for applying pesticides to tomato leaves against whitefly pests based on visual detection according to claim 1, characterized in that: In step S1, the visual inspection module (5) is an embedded EPC-S101 industrial computer, in which a trained YOLOv9 network is implanted, which is used to process and analyze the collected images to detect the location of pests and diseases and achieve the location of pests and diseases.

7. The method for applying pesticides to tomato leaves against whitefly pests based on visual detection according to claim 1, characterized in that: In step S1, the blade disturbance module (1) is a small industrial fan, which is used to disturb the tomato leaves so that the back of the tomato leaves infected with whiteflies are exposed to the spray head, so that the device can apply pesticides on the back of the leaves.

8. The method for applying pesticides to tomato leaves against whitefly pests based on visual detection according to claim 1, characterized in that: In step S1, the spraying module (2) is composed of a pesticide bottle, a gelatin bottle, an air pump and a spray head, so that the pesticide and gelatin are gathered in the spray head and a slow-release liquid mixed with high-efficiency chlorflucyanate and gelatin is formed for precise spraying of tomato leaves.

Images