Unmanned aerial vehicle for monitoring farmland diseases and insect pests

Abstract

The invention relates to the technical field of unmanned aerial vehicles, and provides a farmland pest monitoring unmanned aerial vehicle which comprises an unmanned aerial vehicle body, a bottom shell is fixed to the bottom of the unmanned aerial vehicle body, a through groove is formed in one side of the bottom shell, and a high-definition camera is installed at the position, corresponding to the through groove, in the unmanned aerial vehicle body through a protection assembly. A GPS positioning module is arranged in the unmanned aerial vehicle body; the cleaning assembly is arranged on the protection assembly; the marking assembly is arranged on the side, away from the through groove, of the bottom shell. By means of the technical scheme, the problems that an existing unmanned aerial vehicle camera is exposed in the outdoor environment for a long time and is prone to being blocked by dust and the like, the shooting quality is affected, an automatic cleaning mechanism is lacked, the camera does not have effective protection measures in the non-working state and is prone to being damaged in the transportation or landing process, and an intelligent recognition system for diseases and insect pests is lacked are solved. And automatic analysis and treatment of pest and disease damage types cannot be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles (UAVs), and in particular to a UAV for monitoring farmland pests and diseases. Background Art

[0002] With the development of modern agriculture, farmland pest and disease monitoring, a key component of agricultural production management, directly impacts crop yield and quality. Traditional pest and disease monitoring methods, which mostly rely on manual inspections, visual inspections, and empirical judgment, suffer from low efficiency, limited coverage, and poor accuracy, making them unable to meet the demands of modern agricultural precision management.

[0003] In recent years, with the continuous development of drone and image recognition technologies, drone-based farmland monitoring has gradually gained popularity. Using high-definition cameras to capture and analyze crop growth has become a key area of intelligent agriculture. However, existing drones for farmland monitoring still have the following technical drawbacks: For example, existing drone cameras are exposed to the outdoors for extended periods, easily obscured by dust and other factors, affecting image quality; they lack automatic cleaning mechanisms, have no effective protection measures when not in use, and are susceptible to damage during transport or landing; and they lack intelligent pest and disease identification systems, making it impossible to automatically analyze and address pest types. Summary of the Invention

[0004] The present invention proposes a drone for monitoring farmland pests and diseases, which solves the problems of existing drone cameras being exposed to outdoor environments for a long time, easily obscured by dust and other factors affecting the shooting quality, lacking an automatic cleaning mechanism, having no effective protection measures when the camera is not in working state, and easily damaged during transportation or landing, as well as lacking an intelligent recognition system for pests and diseases, making it impossible to realize automatic analysis and processing of pest and disease types.

[0005] The technical solutions of the present invention are as follows: A farmland pest and disease monitoring drone, comprising: The drone body has a bottom shell fixed to the bottom of the drone body, a through slot is formed on one side of the bottom shell, a high-definition camera is installed in the drone body at a position corresponding to the through slot through a protective component, and a GPS positioning module is installed in the drone body; A cleaning component, the cleaning component being mounted on the protection component and being used to clean the surface of the high-definition camera; a marking assembly, the marking assembly being mounted on a side of the bottom shell away from the through slot, and being used to mark areas in farmland where pests and diseases occur; A pest and disease identification system can generate the types and distribution locations of pests and diseases based on images taken by a high-definition camera, and control a marking component to mark areas where pests and diseases occur.

[0006] Preferably, the protection assembly includes a mounting block, the mounting block is installed in the drone body, the top wall of the mounting block is connected to the top of the inner cavity of the drone body through an electric push rod, and push rods are rotatably installed on both sides of the mounting block; The other end of the push rod is rotatably mounted with a sliding plate; The sliding plate is slidably mounted on the top wall of the bottom shell; Two groups of L-shaped limiting plates are symmetrically fixed to the top wall of the bottom shell, and the sliding plate is slidably installed between the corresponding group of L-shaped limiting plates.

[0007] Preferably, the high-definition camera is fixed at the center of the bottom wall of the mounting block, an annular groove is provided in the mounting block, and the cleaning component is installed in the annular groove.

[0008] Preferably, the cleaning assembly includes a servo motor, the servo motor is fixed on the top wall of the annular groove, and a driving gear is fixed to the power output end at the bottom of the servo motor; The driving gear is meshingly connected with an outer gear ring; The outer gear ring is rotatably mounted on the inner wall of the annular groove, and an arc-shaped scraper is fixed to the bottom wall of the outer gear ring via a connecting rod; The inner wall of the arc-shaped scraper is in contact with the outer wall of the high-definition camera; An arc-shaped limiting through groove communicating with the annular groove is formed through the bottom wall of the mounting block, and the connecting rod is slidably mounted in the arc-shaped limiting through groove.

[0009] Preferably, the marking assembly includes a liquid storage tank, the liquid storage tank is fixed to the bottom of the bottom shell, the liquid storage tank stores marking liquid dye, a mounting groove is opened on one side of the liquid storage tank, and a micro water pump is fixed on the top wall of the mounting groove; The liquid inlet of the micro water pump is connected to the bottom of the inner cavity of the liquid storage tank through a connecting pipe, and the liquid outlet of the micro water pump is connected to a nozzle through a connecting pipe; The nozzle is fixed between the side walls of the mounting groove, and a plurality of nozzles are evenly fixed on the bottom wall of the nozzle; A feeding port is provided on the top of the side wall of the liquid storage tank, and a threaded cover is screwed on the feeding port.

[0010] Preferably, the pest identification system includes: An image acquisition module is used to capture images of farmland areas using a high-definition camera carried by a drone; The image preprocessing module is used to perform normalization and image denoising on the collected images. The normalization formula is:

[0011] in, x is the original pixel value of the image, x norm is the normalized pixel value; The image denoising adopts Gaussian filtering, and its filtering formula is:

[0012] in, G(x,y) is the Gaussian kernel, σ is the standard deviation, which controls the width of the Gaussian distribution. x and y is the offset of the pixel in the image; The image recognition module is used to identify pests and diseases on the pre-processed images using a deep convolutional neural network. The convolution calculation formula is:

[0013] in, I is the input image, K is the convolution kernel, S ( i, j ) is the output feature map, m and n is the dimension of the convolution kernel; The output feedback module is used to output the category and distribution information of the pests and diseases according to the identification results, and is connected to the external terminal through the wireless transmission module.

[0014] Preferably, the image preprocessing module further includes a data enhancement module that supports image rotation, scaling, flipping and other operations. The image rotation transformation is calculated by the following matrix:

[0015] in,( x,y ) is the coordinate of the original image, ( , ) is the rotated coordinate, and θ is the rotation angle.

[0016] Preferably, the pooling operation in the image recognition module adopts a maximum pooling method, and the pooling calculation formula is:

[0017] in, P is the size of the pooling window, I ( i, j ) is the input feature map, S ( i, j ) is the output after pooling.

[0018] Preferably, the deep learning model used by the image recognition module supports transfer learning, and the training loss function used is:

[0019] in, is the loss function, f ( x i ， ) is the output of the model, x i and y i are input and target labels, are the parameters of the model, N is the sample size.

[0020] Preferably, the image recognition module uses a Softmax function in the last layer to perform probability prediction on the pest and disease category, and the classification probability calculation formula is:

[0021] in, P is a category c k The predicted probability of z k is the score output by the neural network, C is the total number of categories.

[0022] The beneficial effects of the present invention are: 1. The protective component provided in the present invention can automatically retract the camera into the drone body when it is not in use, thereby preventing damage caused by external collisions or environmental erosion and extending the service life of the equipment. The cleaning component composed of an arc scraper and a servo motor can regularly remove dust and stains from the camera surface, ensuring image clarity and improving recognition accuracy.

[0023] 2. The marking component provided in the present invention can mark the areas where pests and diseases occur in the farmland, so that agricultural personnel can quickly find the pest and disease areas. The pest and disease identification system provided can generate the types and distribution locations of pests and diseases based on the pictures taken by the high-definition camera, and control the marking component to mark the areas where pests and diseases occur, so that agricultural personnel can intuitively see the pest and disease areas.

[0024] 3. The present invention introduces a deep convolutional neural network, which can achieve high-precision identification of pests and diseases in complex backgrounds and under various environmental conditions. Experiments have shown that the model's recognition accuracy exceeds 92%. Through data enhancement (such as image rotation, scaling, and brightness changes) and transfer learning, the model can maintain stable recognition capabilities in a variety of crops (such as rice, wheat, and corn) and under various weather and lighting conditions, significantly improving the system's applicability and robustness in actual agricultural scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] Figure 1 This is a top view of the external structure of the present invention; Figure 2 This is a three-dimensional diagram of the external structure of the present invention when viewed from above; Figure 3 This is a three-dimensional diagram of the internal structure of the bottom shell of the present invention; Figure 4 This is a perspective view of the protective assembly structure of the present invention; Figure 5 This is a three-dimensional diagram of the cleaning component structure of the present invention; Figure 6 This is a three-dimensional diagram of the marking assembly structure of the present invention; Figure 7 This is a structural block diagram of the pest identification system of the present invention.

[0027] In the figure: 1. UAV body; 2. Bottom shell; 3. Through slot; 4. Protection component; 41. Mounting block; 42. Electric push rod; 43. Push rod; 44. Sliding plate; 45. L-shaped limit plate; 5. HD camera; 6. Cleaning component; 61. Servo motor; 62. Drive gear; 63. Outer gear ring; 64. Connecting rod; 65. Arc scraper; 7. Annular groove; 8. Marking component; 81. Liquid storage tank; 82. Mounting slot; 83. Micro water pump; 84. Nozzle; 85. Nozzle; 86. Connecting pipe; 87. Feeding port; 88. Threaded cover; 9. Arc limit slot; 10. Image acquisition module; 11. Image preprocessing module; 12. Image recognition module; 13. Output feedback module; 14. Wireless transmission module; 15. GPS positioning module. DETAILED DESCRIPTION

[0028] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0029] Example 1 like Figures 1 to 4 As shown, this embodiment proposes: A farmland pest and disease monitoring drone, comprising: The drone body 1 has a bottom shell 2 fixed to the bottom of the drone body 1. A through slot 3 is formed on one side of the bottom shell 2. A high-definition camera 5 is installed inside the drone body 1 at a position corresponding to the through slot 3 through a protective component 4. A GPS positioning module 15 is installed inside the drone body 1. Cleaning assembly 6, cleaning assembly 6 is mounted on the protection assembly 4, and is used to clean the surface of the high-definition camera 5; The marking assembly 8 is installed on the side of the bottom shell 2 away from the through groove 3. The marking assembly 8 is used to mark the area where pests and diseases occur in the farmland; The pest and disease identification system can generate the types and distribution locations of pests and diseases based on the pictures taken by the high-definition camera 5, and control the marking component 8 to mark the areas where pests and diseases occur.

[0030] In this embodiment, the protection component 4 is provided to protect the high-definition camera 5. When the high-definition camera 5 is not needed, the protection component 4 can store the high-definition camera 5 inside the drone body 1 for protection, effectively preventing the high-definition camera 5 from being accidentally damaged. The cleaning component 6 is provided to clean the shooting end of the high-definition camera 5 to prevent dust and stains from covering the high-definition camera 5, resulting in unclear image acquisition. The marking component 8 is provided to mark areas where pests and diseases occur in farmland, allowing agricultural personnel to quickly find areas with pests and diseases. The pest and disease identification system is provided to generate the types and distribution locations of pests and diseases based on the pictures taken by the high-definition camera 5, and control the marking component 8 to mark areas where pests and diseases occur.

[0031] Example 2 like Figures 3 to 5 As shown, based on the same concept as the above embodiment 1, this embodiment further proposes: The protection assembly 4 includes a mounting block 41, which is installed in the drone body 1. The top wall of the mounting block 41 is connected to the top of the inner cavity of the drone body 1 through an electric push rod 42. Push rods 43 are rotatably installed on both sides of the mounting block 41. The other end of the push rod 43 is rotatably mounted with a sliding plate 44; The sliding plate 44 is slidably mounted on the top wall of the bottom shell 2; Two groups of L-shaped limiting plates 45 are symmetrically fixed to the top wall of the bottom shell 2 , and the sliding plate 44 is slidably installed between the corresponding group of L-shaped limiting plates 45 .

[0032] The high-definition camera 5 is fixed at the center of the bottom wall of the mounting block 41 . An annular groove 7 is defined in the mounting block 41 , and the cleaning component 6 is mounted in the annular groove 7 .

[0033] The cleaning assembly 6 includes a servo motor 61, which is fixed to the top wall of the annular groove 7. A driving gear 62 is fixed to the power output end at the bottom of the servo motor 61; The driving gear 62 is meshedly connected with an outer gear ring 63; The outer gear ring 63 is rotatably mounted on the inner wall of the annular groove 7, and an arc-shaped scraper 65 is fixed to the bottom wall of the outer gear ring 63 via a connecting rod 64; The inner wall of the arc scraper 65 fits in with the outer wall of the high-definition camera 5; An arcuate limiting groove 9 communicating with the annular groove 7 is formed through the bottom wall of the mounting block 41 , and the connecting rod 64 is slidably mounted in the arcuate limiting groove 9 .

[0034] In this embodiment, the protective assembly 4 is provided to protect the high-definition camera 5. When the high-definition camera 5 is not in use, the protective assembly 4 can store the high-definition camera 5 inside the drone body 1 for protection, effectively preventing the high-definition camera 5 from being accidentally damaged. When the high-definition camera 5 is needed, the electric push rod 42 is controlled to extend, and the electric push rod 42 drives the mounting block 41 downward. During the downward movement of the mounting block 41, the push rods 43 on both sides can push the sliding plate 44 outward, so that the slot 3 is opened. At this time, the high-definition camera 5 at the bottom of the mounting block 41 can be moved through the slot 3 to the bottom of the drone body 1 for shooting operations. The storage process is the opposite of the above process.

[0035] The cleaning assembly 6 is designed to clean the camera end of the HD camera 5, preventing dust and dirt from obscuring the camera 5 and causing unclear images. When the camera end of the HD camera 5 needs to be cleaned, the servo motor 61 is driven to rotate back and forth. The servo motor 61 drives the curved scraper 65 below via the drive gear 62 and the outer gear ring 63, thereby cleaning the HD camera 5 and ensuring the best possible image quality.

[0036] Example 3 like Figure 6 As shown, based on the same concept as the above embodiment 1, this embodiment further proposes: The marking assembly 8 includes a liquid storage tank 81, which is fixed to the bottom of the bottom shell 2. The liquid storage tank 81 stores marking liquid dye. A mounting groove 82 is opened on one side of the liquid storage tank 81, and a micro water pump 83 is fixed to the top wall of the mounting groove 82. The liquid inlet of the micro water pump 83 is connected to the bottom of the inner cavity of the liquid storage tank 81 through the connecting pipe 86, and the liquid outlet of the micro water pump 83 is connected to the nozzle 84 through the connecting pipe 86; The nozzle 84 is fixed between the side walls of the mounting groove 82, and a plurality of nozzles 85 are evenly fixed on the bottom wall of the nozzle 84; A feeding port 87 is provided on the top of the side wall of the liquid storage box 81 , and a threaded cover 88 is screwed to the feeding port 87 .

[0037] In this embodiment, marking assembly 8 can mark areas of farmland where pests and diseases occur, allowing agricultural workers to quickly locate the pest areas. When the pest identification system detects an area where pests and diseases have occurred, it controls micro-pump 83 to operate. Micro-pump 83 draws harmless marking liquid dye from reservoir 81 into nozzle 84 and sprays it onto the corresponding pest and disease area through nozzle 85, allowing agricultural workers to visually identify the pest and disease area. A feeding port 87 provided on reservoir 81 facilitates the replenishment of the marking liquid dye into reservoir 81.

[0038] Example 4 like Figure 7 As shown, based on the same concept as the above embodiment 1, this embodiment further proposes: In this embodiment, the pest identification system includes: Image acquisition module 10, the image acquisition module 10 is used to take flight photos of the farmland area through the high-definition camera 5 carried by the drone, collect image data, and perform real-time positioning through the set GPS positioning module 15; The image preprocessing module 11 is used to perform normalization processing and image denoising on the collected image. The normalization processing formula is:

[0039] in, x is the original pixel value of the image, x norm is the normalized pixel value; Image denoising uses Gaussian filtering, and its filtering formula is:

[0040] in, G(x,y) is the Gaussian kernel, σ is the standard deviation, which controls the width of the Gaussian distribution. x and y is the offset of the pixel points in the image. In this embodiment, σ=1.2 is used to remove noise in the image; The image preprocessing module 11 further includes a data enhancement module that supports image rotation, scaling, flipping and other operations. The image rotation transformation is calculated by the following matrix:

[0041] in,( x,y ) is the coordinate of the original image, ( , ) are the rotated coordinates, and θ is the rotation angle. In this embodiment, image enhancement methods (horizontal image flipping, random rotation of 15° to 45°, and brightness variation) are used to expand the training set sample size and improve the robustness of the recognition model.

[0042] The image recognition module 12 is used to use a deep convolutional neural network to identify pests and diseases on the pre-processed image. The convolution calculation formula is:

[0043] in, I is the input image, K is the convolution kernel, S ( i, j ) is the output feature map, m and n is the dimension of the convolution kernel; The output feedback module 13 is used to output the pest category and distribution information according to the identification result, and is connected to the external terminal through the wireless transmission module 14.

[0044] The pooling operation in the image recognition module adopts the maximum pooling method, and the pooling calculation formula is:

[0045] in, P is the size of the pooling window, I ( i, j ) is the input feature map, S ( i, j ) is the output after pooling.

[0046] The deep learning model used by the image recognition module 12 supports transfer learning, and the training loss function used is:

[0047] in, is the loss function, f ( x i ， ) is the output of the model, x i and y i are input and target labels, are the parameters of the model, N is the sample size.

[0048] The image recognition module 12 uses the Softmax function in the last layer to predict the probability of pest and disease categories. The classification probability calculation formula is:

[0049] in, P is a category c k The predicted probability of z k is the score output by the neural network, C is the total number of categories.

[0050] In this embodiment, a pre-trained ResNet50 network is used as the backbone network structure. After pre-training on ImageNet, the model parameters are fine-tuned through transfer learning. The new data set contains 5 types of common crop pests and diseases: leaf spot, aphids, leaf rollers, rice planthoppers, and rust, totaling 5,000 labeled images. The model output layer uses Softmax for classification, outputs the probability values of the 5 types of pests and diseases, adopts the cross entropy loss function as the objective function, uses the Adam optimizer, the initial learning rate is 0.0001, and the number of training rounds is 80 rounds. The model of the pest and disease recognition system is obtained and tested in an area of 300 acres of farmland. In the test results, the model recognition accuracy rate reached 92.7%, and the false alarm rate was less than 4.5%. The pest and disease prevention response time was shortened by 36 hours, significantly reducing economic losses.

[0051] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A drone for monitoring farmland pests and diseases, characterized by: include: A drone body (1), wherein a bottom shell (2) is fixed to the bottom of the drone body (1), a through slot (3) is provided on one side of the bottom shell (2), a high-definition camera (5) is installed in the drone body (1) at a position corresponding to the through slot (3) through a protective component (4), and a GPS positioning module (15) is installed in the drone body (1); A cleaning component (6), the cleaning component (6) being mounted on the protective component (4), and the cleaning component (6) being used to clean the surface of the high-definition camera (5); a marking assembly (8), the marking assembly (8) being mounted on a side of the bottom shell (2) away from the through slot (3), and the marking assembly (8) being used to mark areas in farmland where pests and diseases occur; A pest and disease identification system is capable of generating the types and distribution locations of pests and diseases based on images taken by a high-definition camera (5), and controlling a marking component (8) to mark areas where pests and diseases occur.

2. The farmland pest monitoring drone according to claim 1, characterized in that: The protection assembly (4) includes a mounting block (41), the mounting block (41) is installed in the drone body (1), the top wall of the mounting block (41) is connected to the top of the inner cavity of the drone body (1) via an electric push rod (42), and push rods (43) are rotatably installed on both sides of the mounting block (41); The other end of the push rod (43) is rotatably mounted with a sliding plate (44); The sliding plate (44) is slidably mounted on the top wall of the bottom shell (2); Two groups of L-shaped limiting plates (45) are symmetrically fixed to the top wall of the bottom shell (2), and the sliding plate (44) is slidably installed between the corresponding group of L-shaped limiting plates (45).

3. The farmland pest monitoring drone according to claim 2, characterized in that: The high-definition camera (5) is fixed at the center of the bottom wall of the mounting block (41); an annular groove (7) is provided in the mounting block (41); and the cleaning component (6) is installed in the annular groove (7).

4. The farmland pest monitoring drone according to claim 3, characterized in that: The cleaning assembly (6) comprises a servo motor (61), the servo motor (61) being fixed on the top wall of the annular groove (7), and a driving gear (62) being fixed to the power output end at the bottom of the servo motor (61); The driving gear (62) is meshingly connected with an outer gear ring (63); The outer gear ring (63) is rotatably mounted on the inner wall of the annular groove (7), and an arc-shaped scraper (65) is fixed to the bottom wall of the outer gear ring (63) via a connecting rod (64); The inner wall of the arc-shaped scraper (65) is in contact with the outer wall of the high-definition camera (5); An arc-shaped limiting groove (9) communicating with the annular groove (7) is formed through the bottom wall of the mounting block (41), and the connecting rod (64) is slidably mounted in the arc-shaped limiting groove (9).

5. The farmland pest monitoring drone according to claim 1, characterized in that: The marking assembly (8) includes a liquid storage tank (81), the liquid storage tank (81) is fixed to the bottom of the bottom shell (2), the liquid storage tank (81) stores marking liquid dye, a mounting groove (82) is provided on one side of the liquid storage tank (81), and a micro water pump (83) is fixed to the top wall of the mounting groove (82); The liquid inlet of the micro water pump (83) is connected to the bottom of the inner cavity of the liquid storage tank (81) through a connecting pipe (86), and the liquid outlet of the micro water pump (83) is connected to a nozzle (84) through a connecting pipe (86); The nozzle (84) is fixed between the side walls of the mounting groove (82), and a plurality of nozzles (85) are evenly fixed on the bottom wall of the nozzle (84); A feeding port (87) is provided at the top of the side wall of the liquid storage tank (81), and a threaded cover (88) is screwed to the feeding port (87).

6. The farmland pest monitoring drone according to claim 1, characterized in that: The pest identification system includes: An image acquisition module (10), the image acquisition module (10) is used to perform flight photography of the farmland area using a high-definition camera (5) carried by a drone to collect image data; An image preprocessing module (11) is used to perform normalization processing and image denoising on the collected image. The normalization processing formula is: in, x is the original pixel value of the image, x norm is the normalized pixel value; The image denoising adopts Gaussian filtering, and its filtering formula is: in, G(x,y) is the Gaussian kernel, σ is the standard deviation, which controls the width of the Gaussian distribution. x and y is the offset of the pixel in the image; An image recognition module (12) is used to identify pests and diseases on the pre-processed image using a deep convolutional neural network, wherein the convolution calculation formula is: in, I is the input image, K is the convolution kernel, S ( i, j ) is the output feature map, m and n is the dimension of the convolution kernel; An output feedback module (13) is used to output the category and distribution information of the pests and diseases according to the identification result, and is connected to an external terminal via a wireless transmission module (14).

7. The farmland pest monitoring drone according to claim 6, characterized in that: The image preprocessing module (11) further includes a data enhancement module that supports image rotation, scaling, flipping and other operations. The image rotation transformation is calculated by the following matrix: in,( x,y ) is the coordinate of the original image, ( , ) is the rotated coordinate, and θ is the rotation angle.

8. The farmland pest monitoring drone according to claim 6, characterized in that: The pooling operation in the image recognition module adopts the maximum pooling method, and the pooling calculation formula is: in, P is the size of the pooling window, I ( i, j ) is the input feature map, S ( i, j ) is the output after pooling.

9. The farmland pest monitoring drone according to claim 6, characterized in that: The deep learning model used in the image recognition module (12) supports transfer learning, and the training loss function used is: in, is the loss function, f ( x i ， ) is the output of the model, x i and y i are input and target labels, are the parameters of the model, N is the sample size.

10. The farmland pest monitoring drone according to claim 6, characterized in that: The image recognition module (12) uses the Softmax function in the last layer to perform probability prediction on the pest and disease category, and the classification probability calculation formula is: in, P is a category c k The predicted probability of z k is the score output by the neural network, C is the total number of categories.

Images