A cassava growth control method by unmanned aerial vehicle covering spraying

Abstract

The application discloses a kind of unmanned plane covering spraying cassava control method, including the following steps: obtaining cassava planting area aerial image, calculating vegetation index, and identifying abnormal area according to the comparison result with preset threshold value;Planting area is divided into two-dimensional grid array, and binary state matrix is generated according to the comparison result of the area proportion of abnormal area in grid and fault tolerance threshold value;Abnormal array point is extracted based on state matrix, and the area of connected region formed by adjacent abnormal array point is calculated, when the area is less than the circular area corresponding to the minimum turning radius of unmanned plane, keep the route and close the spraying system when passing, when the area is greater than or equal to the minimum turning radius, bypass;Real-time positioning coordinates of unmanned plane are obtained and compared with state matrix to control the start-stop of spraying system, and the advance trigger distance is calculated according to the flight speed and mechanical delay time, and the start-stop instruction is issued in advance to compensate for mechanical delay.The application realizes accurate removal of cassava abnormal area spraying, saves liquid medicine and reduces energy consumption.

Description

Technical Field

[0001] This invention relates to agricultural operation methods using drones, specifically to a method for controlling excessive growth of cassava by covering and spraying with drones. Background Technology

[0002] Cassava, as an important root crop, often exhibits excessive vegetative growth in its above-ground stems and leaves due to favorable water and fertilizer conditions or varietal characteristics. This excessive growth consumes a large amount of photosynthetic products, leading to a decline in the yield and quality of underground tubers. Therefore, in production practice, it is often necessary to control this excessive growth by applying plant growth regulators such as paclobutrazol and uniconazole to inhibit stem elongation and promote nutrient transfer to the tubers.

[0003] Currently, the application of growth regulators for cassava mainly relies on manual labor or traditional ground machinery for uniform spraying. However, manual backpack spraying is labor-intensive, inefficient, and makes it difficult to ensure uniform application. Traditional ground machinery is prone to crushing plants or missing areas in cassava fields with wide row spacing. Furthermore, cassava is quite sensitive to growth regulators such as paclobutrazol. If the concentration exceeds a certain threshold (e.g., 250 mg / L), it can easily lead to severe stunting of plants, smaller and thicker leaves, dark green leaves, and even reduced tuber numbers, deformities, or cracking—symptoms of phytotoxicity that severely impact final yield. Therefore, precise application is crucial for controlling growth in cassava.

[0004] With the development of agricultural drone technology, using drones for large-area spraying has become an important means of improving efficiency. However, existing drone spraying solutions typically employ a "full coverage spraying" mode, which involves uniformly spraying the entire plot along a pre-set fixed flight path, or a "continuous variable application" mode (which adjusts the spraying volume in real time based on sensor data). However, considering the unique agronomical characteristics of cassava cultivation, these conventional operating modes have significant drawbacks:

[0005] First, cassava planting involves wide row spacing, and the field is prone to localized poor growth, missing seedlings, or seedling death due to root rot, uneven fertilization and irrigation, and other reasons. Full-coverage spraying will spray the pesticide onto abnormal areas with weak growth, which not only wastes the pesticide but may also further inhibit the growth of weak seedlings. Second, traditional continuous variable application (such as "spray more when the plant is growing well and less when the plant is growing poorly") requires complex calculations by the onboard computer, resulting in a large computational load. In actual field operations, it is easily affected by external wind fields, and the spraying response is delayed, making it difficult to achieve truly precise control.

[0006] In addition, existing drone operation methods often suffer from a single path dependency when dealing with abnormal areas: either modifying the flight path or adjusting the spraying volume. If facing a large area of ​​continuous abnormality, simply traversing the flight path without continuously spraying will still result in ineffective spraying. Conversely, if facing scattered abnormal points, frequently modifying the flight path and detouring will cause the drone to maneuver frequently, significantly increasing additional energy consumption. Summary of the Invention

[0007] The purpose of this invention is to overcome the above-mentioned problems and provide a method for controlling excessive growth of cassava by covering and spraying with drones. This method achieves intelligent and precise removal and spraying of abnormal cassava growth areas through a binary array model and a dual strategy of "interrupted pesticide crossing + detour avoidance". It not only significantly reduces the onboard computing power requirements and operational energy consumption, but also saves pesticide dosage and effectively avoids the risk of pesticide damage, thereby significantly improving operational efficiency and economic benefits.

[0008] The objective of this invention is achieved through the following technical solution:

[0009] A method for controlling excessive growth of cassava by covering and spraying with drones includes the following steps:

[0010] (1) Obtain aerial images of the cassava planting area, extract spectral or morphological features of the cassava canopy based on the aerial images, calculate the vegetation index reflecting the growth of cassava, and identify areas with weak growth, missing seedlings, dead plants or severe diseases in the cassava planting area as abnormal areas based on the comparison results of the vegetation index and the preset threshold.

[0011] (2) Divide the cassava planting area into a two-dimensional grid array composed of several grid units. For each grid unit, determine whether the grid unit is an abnormal array point based on the ratio of the area of ​​the abnormal region inside it to the total area of ​​the grid unit. If the ratio is greater than the preset fault tolerance threshold, mark the grid unit as an abnormal array point and assign a value of 0. If the ratio is not greater than the fault tolerance threshold, mark the grid unit as a normal array point and assign a value of 1. Generate a binary state matrix from the values ​​of all grid units.

[0012] (3) Based on the binarized state matrix, extract all abnormal array points and calculate the area of ​​the connected region formed by adjacent abnormal array points. Based on the relationship between the area of ​​the connected region and the area of ​​the circle corresponding to the minimum turning radius of the UAV, select the avoidance strategy:

[0013] When the area of ​​the connected region is smaller than the circular area corresponding to the minimum turning radius of the drone, the original flight path of the drone is maintained. When the drone passes through the grid cell corresponding to the connected region, the spraying system is turned off. After the drone leaves the grid cell corresponding to the connected region, the spraying system is turned on.

[0014] When the area of ​​the connected region is greater than or equal to the circular area corresponding to the minimum turning radius of the drone, the connected region is set as a flight obstacle zone, and the flight path of the drone is replanned to bypass the flight obstacle zone.

[0015] (4) When performing spraying operations, the real-time positioning coordinates of the UAV are obtained and the position is compared with the binary state matrix. The start and stop of the spraying system are controlled according to the comparison results. At the same time, the advance trigger distance is calculated based on the current flight speed of the UAV and the mechanical delay time of the spraying system. The shutdown or start command is issued in advance before the UAV reaches the boundary of the abnormal array point to compensate for the mechanical delay of the spraying system.

[0016] In a preferred embodiment of the present invention, in step (1), the aerial image is acquired by a drone equipped with a multispectral camera or a high-resolution visible light camera;

[0017] The vegetation index is the normalized vegetation index or the greenness index, which is characterized by the ratio obtained by calculating the difference between the green band and the red band and then dividing it by the sum of the green band, the red band and the blue band.

[0018] In a preferred embodiment of the present invention, in step (1), after identifying the abnormal region, a morphological filtering algorithm is used to process the identification result. The morphological filtering algorithm includes dilation and erosion operations, which are used to remove background noise from field weeds and bare soil, and to accurately locate the physical coordinate boundary of the abnormal growth of cassava.

[0019] In a preferred embodiment of the present invention, in step (2), the side length of the grid cell of the two-dimensional grid array is set according to the effective spray width of the plant protection drone, and the effective spray width is used as the side length of the grid cell.

[0020] In a preferred embodiment of the present invention, in step (2), the fault tolerance threshold is 15%; when the area ratio of the abnormal region in a certain grid cell is greater than 15%, the grid cell is marked as an abnormal array point and assigned a value of 0; when the area ratio of the abnormal region in a certain grid cell is less than or equal to 15%, the grid cell is marked as a normal array point and assigned a value of 1.

[0021] In a preferred embodiment of the present invention, in step (3), when the area of ​​the connected region is greater than or equal to the circular area corresponding to the minimum turning radius of the UAV, and the flight route needs to be replanned to bypass the flight obstacle area, a multi-objective optimization function is used to plan the detour path. The multi-objective optimization function uses the sum of the product of the total flight time and its weight coefficient and the product of the total energy consumption and its weight coefficient as the optimization objective to optimize the detour path.

[0022] Furthermore, when planning the detour path using the multi-objective optimization function, kinematic constraints for the agricultural drone are also introduced, specifically including: constraining the instantaneous flight speed of the drone within a preset maximum flight speed, constraining the instantaneous acceleration within a preset maximum acceleration, and constraining the curvature of the flight path within the reciprocal of a preset minimum turning radius; and ensuring the drone's stable flight attitude during the detour by setting smooth trajectory transition point coordinates.

[0023] In a preferred embodiment of the present invention, in step (3), when the area of ​​the connected region is greater than or equal to the circular area corresponding to the minimum turning radius of the UAV, the connected region is marked as a flight obstacle zone using an improved ant colony algorithm, and the three-dimensional flight path of the UAV is replanned to bypass the flight obstacle zone.

[0024] In a preferred embodiment of the present invention, in step (4), real-time dynamic differential positioning is used to obtain the centimeter-level three-dimensional coordinates of the UAV, and the centimeter-level three-dimensional coordinates are compared with the binarized state matrix at high frequency.

[0025] In a preferred embodiment of the present invention, in step (4), the advance triggering distance is determined based on the product of the current forward speed of the UAV and the mechanical delay time of the spraying system; when the UAV approaches the boundary of the abnormal array point and the distance is equal to the advance triggering distance, the shutdown command is issued in advance, and when the UAV leaves the boundary of the abnormal array point and the distance is equal to the advance triggering distance, the start command is issued in advance.

[0026] Compared with the prior art, the present invention has the following advantages:

[0027] 1. This invention maps the cassava growth recognition results into binary "normal array points" and "abnormal array points", simplifying complex agronomic judgments into 0 / 1 logical decisions, greatly reducing the demand for onboard computer computing power, avoiding interference from external wind fields on the continuous variable application accuracy, and significantly improving the response speed of spraying commands.

[0028] 2. This invention adaptively selects either "discontinuing drug crossing" or "route detour" strategy based on the area of ​​the abnormal connected region: scattered abnormal points adopt the method of crossing to discontinuing drug to avoid frequent maneuvering and energy consumption; large areas of continuous abnormal regions adopt the method of route detour to save flight time; dual avoidance and collaborative optimization effectively solve the defects of a single avoidance method that cannot take into account both energy consumption and accuracy.

[0029] 3. This invention addresses the agronomical characteristics of cassava, such as wide row spacing and susceptibility to localized seedling loss and death, by constructing a complete closed-loop scheme of "identification-mapping-avoidance". This scheme saves on pesticide usage while ensuring the effect of controlling excessive growth, and fills the technological gap in targeted removal and pesticide application by drones for large root crops such as cassava. Detailed Implementation

[0030] To enable those skilled in the art to fully understand the technical solutions of the present invention, the present invention will be further described below with reference to embodiments, but the embodiments of the present invention are not limited thereto.

[0031] Example 1

[0032] The method for controlling excessive growth of cassava by covering and spraying with drones in this embodiment includes the following steps:

[0033] I. Identification of areas with abnormal cassava growth.

[0034] Before carrying out spraying operations, it is necessary to first identify areas of abnormal growth within the cassava planting area. Specifically, a drone equipped with a multispectral camera or a high-resolution visible light camera is used to conduct preliminary aerial photography of the target cassava planting area to obtain orthophotos of the cassava canopy. The flight altitude is set at 10 to 15 meters above the cassava canopy to ensure that the ground resolution of the image is within 5 centimeters per pixel.

[0035] After acquiring aerial images, the spectral characteristics of the cassava canopy are extracted, and vegetation indices reflecting cassava growth are calculated. In this embodiment, the vegetation indices used are the Normalized Difference Vegetation Index (NDVI) or the Greenness Index (VARI). The formula for calculating the Greenness Index (VARI) is as follows:

[0036] ;

[0037] In the formula, Green is the reflectance value of the green band, Red is the reflectance value of the red band, and Blue is the reflectance value of the blue band.

[0038] Based on the comparison between the calculated vegetation index and a preset threshold, areas in cassava planting regions exhibiting weak growth, missing seedlings, dead plants, or severe disease are identified as abnormal areas. Specifically, a standard vegetation index threshold V for healthy cassava growth in the local area is set. th When the vegetation index V of a certain area <V th When this happens, the area is determined to be an abnormal area.

[0039] Furthermore, after identifying the abnormal areas, a morphological filtering algorithm is used to process the identification results. This morphological filtering algorithm includes dilation and erosion operations to remove background noise from field weeds and bare soil, and to accurately locate the physical coordinate boundaries of abnormal cassava growth.

[0040] II. Constructing a two-dimensional drug application array.

[0041] The target cassava planting area is divided into a two-dimensional grid array consisting of several grid units. The side length of the grid unit is set according to the effective spray width W of the agricultural drone, that is, the effective spray width W is used as the side length of the grid unit. For example, when the effective spray width W of the agricultural drone is 2.0 meters, the target operation area is divided into a two-dimensional square grid of M rows × N columns from due north to due east, with a side length of 2.0 meters.

[0042] For each grid cell C i,j Calculate the area ratio P of the abnormal region within the grid cell, which is the ratio of the area of ​​the abnormal region within the grid cell to the total area of ​​the grid cell; set the fault tolerance threshold P. th (Preferred percentage in this embodiment is 15%), each grid cell is binarized and assigned a value according to the following rules:

[0043] When P>P th When this occurs, the grid cell is marked as an abnormal array point, and its matrix elements are assigned a value of 0.

[0044] When P≤P th When this happens, the grid cell is marked as a normal array point, and its matrix elements are assigned a value of 1.

[0045] The complete binary state matrix S is generated by assigning values ​​to all grid cells. M×N The system stores the state matrix in the UAV's onboard flight control motherboard for subsequent steps to access.

[0046] III. Dual avoidance strategy based on state matrix.

[0047] Based on the binary state matrix S M×N Extract all grid cells with a value of 0 (i.e., anomalous array points), and calculate the area A of the connected regions formed by adjacent anomalous array points. err Introducing the minimum turning radius R of the drone min Calculate the area of ​​the corresponding circle. .

[0048] Based on the area A of the connected region err and Given the size relationship between them, choose one of the following two avoidance strategies:

[0049] (a) Micro-management avoidance – crossing the road after stopping medication.

[0050] When A err < When the abnormal connectivity area is small, adopting a detour strategy would lead to frequent drone maneuvers and increased energy consumption. In this case, the drone's original parallel flight path remains unchanged. The drone flies along its original flight path. When its centroid coordinates enter the boundary of the grid cell corresponding to the abnormal connectivity area, the control system sends a shutdown command to the spraying system, shutting down the water pump and nozzle solenoid valves. When the drone's centroid coordinates leave the boundary of the grid cell corresponding to the abnormal connectivity area, the control system sends an activation command to resume spraying operations.

[0051] (ii) Macro-level route avoidance - detour avoidance.

[0052] When A err ≥ In such cases, the contiguous abnormal area is designated as a rigid flight obstacle zone, and the flight path of the UAV is replanned to bypass the flight obstacle zone.

[0053] Furthermore, when replanning the detour route, a multi-objective optimization function is used to plan the detour route. This multi-objective optimization function takes the sum of the product of the total flight time T and its weight coefficient ω1, and the product of the total energy consumption E and its weight coefficient ω2 as the optimization objective, that is, the objective function J is expressed as:

[0054] ;

[0055] By optimizing the above objective function, a comprehensive optimization is achieved that maximizes operational efficiency and minimizes battery energy consumption during the obstacle avoidance process.

[0056] Furthermore, when planning the detour path using a multi-objective optimization function, kinematic constraints for the agricultural drone are also introduced, specifically including:

[0057] ;

[0058] ;

[0059] ;

[0060] Where v(t) is the instantaneous flight speed of the UAV, V max The maximum permissible flight speed is preset; a(t) is the instantaneous acceleration, A max The preset maximum allowable acceleration; κ(t) is the trajectory curvature, R min This is the minimum turning radius for the drone. By setting smooth trajectory transition point coordinates, the drone's flight attitude is ensured to remain stable during maneuvering, avoiding center of gravity shift caused by liquid sloshing in the tank due to sharp turns.

[0061] Furthermore, when replanning flight routes, an improved ant colony algorithm is used to mark connected regions as flight obstacle zones, and the UAV's three-dimensional flight path is replanned to bypass these obstacle zones.

[0062] IV. Precise drug application and compensation for delays.

[0063] During spraying operations, Real-time Dynamic Differential Positioning (RTK) is activated to acquire the UAV's centimeter-level three-dimensional coordinates, and these centimeter-level three-dimensional coordinates are compared with the onboard binarized state matrix S. M×N High-frequency position comparison is performed, and the start and stop of the spraying system are controlled in real time based on the comparison results.

[0064] Meanwhile, considering the inherent mechanical delay Δt between the flight control system issuing a command and the water pump and nozzle solenoid valve fully responding, the system calculates the advance trigger distance D based on the UAV's current forward speed v:

[0065] ;

[0066] When the drone approaches the boundary of the abnormal array point and the distance is equal to D, the system issues a shutdown command in advance; when the drone leaves the boundary of the abnormal array point and the distance is equal to D, the system issues an activation command in advance. Through the above delay compensation mechanism, the physical sedimentation boundary of the pesticide solution is ensured to fit perfectly with the preset grid array boundary, achieving precise coverage of normal cassava plants and zero application of pesticides to abnormal areas.

[0067] Furthermore, during spraying operations, the drone utilizes a downward-looking sensor mounted on it for real-time terrain-following flight. The downward-looking sensor can be either a lidar or an ultrasonic sensor to ensure that the drone maintains a constant flight altitude under complex terrain conditions.

[0068] Example 2

[0069] The drone-based cassava over-spraying method for controlling excessive growth in this embodiment is operated as follows in a real-world scenario:

[0070] (a) Experimental design.

[0071] The experiment was conducted at a cassava base, using the local main cassava variety. A randomized block design was used, with treatment group T1 (paclobutrazol treatment), treatment group T2 (uniconazole treatment), and a blank control group CK. Each treatment was repeated three times.

[0072] Treatment Group T1: A drone-based coverage spraying scheme was used to spray 30% paclobutrazol wettable powder diluted 800-1000 times. Based on the abnormal areas identified by previous aerial photography, a binary state matrix was generated. The drone carried out spraying operations according to the method of this invention, applying the pesticide only to normal array points and avoiding abnormal array points (interrupting the pesticide passage or detouring).

[0073] Treatment group T2: Using the same coverage spraying path as T1, spray 5% uniconazole wettable powder at a dilution ratio of 2000-3000 times, and perform precise spraying in accordance with the method of the present invention.

[0074] Blank control group (CK): No growth control treatment was given; routine management was performed.

[0075] (ii) Spraying timing and operating parameters.

[0076] The first spraying should be carried out when all cassava plants in the treatment plots reach a height of 60-100 cm. The drone operation parameters should be set as follows: flight altitude 3-4 meters above the canopy, flight speed 1-2 m / s to ensure sufficient droplet settling time. Apply 30-40 kg of pesticide solution per acre, evenly spraying it onto the stems and leaves, focusing on the upper new leaves and tender shoots. If excessive vegetative growth persists later, a second spray can be applied 15-20 days after the first application, with a slightly lower concentration (e.g., 80-120 mg / L).

[0077] (III) Data collection and effect evaluation.

[0078] The following indicators were observed before application, and 7, 14, and 30 days after application, as well as before harvest:

[0079] Plant height: Measured with a tape measure from the ground to the growing point. 10 plants were marked in each plot, and the same plant was measured.

[0080] Stem diameter: Measure the diameter at the third node above the base of the main stem using vernier calipers.

[0081] Internode length: The total length of five consecutive internodes in the middle of the main stem is measured with a measuring tape, and the average value is taken.

[0082] Chlorophyll content (SPAD value): The chlorophyll content of the fully expanded leaves at the top was measured using a chlorophyll meter, and the average value was taken from 3 leaves per plant.

[0083] Growth control rate: calculated as (average plant height of CK - average plant height of treatment) / average plant height of CK × 100%.

[0084] During the harvest period, the number of tubers per plant, the weight of tubers per plant, the yield of the plot, and the starch content were investigated. The costs of pesticides and operations were also recorded, and a comprehensive economic benefit analysis was conducted.

[0085] Experimental results show that the method of the present invention for controlling excessive growth in cassava can accurately identify and avoid abnormal growth areas in the cassava planting area, effectively saving pesticide dosage while ensuring the effect of controlling excessive growth, and avoiding the risk of pesticide damage caused by accidental spraying of pesticides into weak seedling areas.

[0086] The above are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above content. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for controlling excessive growth of cassava by covering and spraying with a drone, characterized in that, Includes the following steps: (1) Obtain aerial images of the cassava planting area, extract spectral or morphological features of the cassava canopy based on the aerial images, calculate the vegetation index reflecting the growth of cassava, and identify areas with weak growth, missing seedlings, dead plants or severe diseases in the cassava planting area as abnormal areas based on the comparison results of the vegetation index and the preset threshold. (2) Divide the cassava planting area into a two-dimensional grid array composed of several grid units. For each grid unit, determine whether the grid unit is an abnormal array point based on the ratio of the area of ​​the abnormal region inside it to the total area of ​​the grid unit. If the ratio is greater than the preset fault tolerance threshold, mark the grid unit as an abnormal array point and assign a value of 0. If the ratio is not greater than the fault tolerance threshold, mark the grid unit as a normal array point and assign a value of 1. Generate a binary state matrix from the values ​​of all grid units. (3) Based on the binarized state matrix, extract all abnormal array points and calculate the area of ​​the connected region formed by adjacent abnormal array points. Based on the relationship between the area of ​​the connected region and the area of ​​the circle corresponding to the minimum turning radius of the UAV, select the avoidance strategy: When the area of ​​the connected region is smaller than the circular area corresponding to the minimum turning radius of the drone, the original flight path of the drone is maintained. When the drone passes through the grid cell corresponding to the connected region, the spraying system is turned off. After the drone leaves the grid cell corresponding to the connected region, the spraying system is turned on. When the area of ​​the connected region is greater than or equal to the circular area corresponding to the minimum turning radius of the drone, the connected region is set as a flight obstacle zone, and the flight path of the drone is replanned to bypass the flight obstacle zone. (4) When performing spraying operations, the real-time positioning coordinates of the UAV are obtained and the position is compared with the binary state matrix. The start and stop of the spraying system are controlled according to the comparison results. At the same time, the advance trigger distance is calculated based on the current flight speed of the UAV and the mechanical delay time of the spraying system. The shutdown or start command is issued in advance before the UAV reaches the boundary of the abnormal array point to compensate for the mechanical delay of the spraying system.

2. The method for controlling excessive growth of cassava by drone spraying according to claim 1, characterized in that, In step (1), the aerial images are acquired by a drone equipped with a multispectral camera or a high-resolution visible light camera; The vegetation index is the normalized vegetation index or the greenness index, which is characterized by the ratio obtained by calculating the difference between the green band and the red band and then dividing it by the sum of the green band, the red band and the blue band.

3. The method for controlling excessive growth of cassava by drone spraying according to claim 1, characterized in that, In step (1), after identifying the abnormal region, a morphological filtering algorithm is used to process the identification result. The morphological filtering algorithm includes dilation and erosion operations, which are used to remove background noise from field weeds and bare soil, and to accurately locate the physical coordinate boundary of the abnormal cassava growth.

4. The method for controlling excessive growth of cassava by drone spraying according to claim 1, characterized in that, In step (2), the side length of the grid cell of the two-dimensional grid array is set according to the effective spray width of the plant protection drone, and the effective spray width is used as the side length of the grid cell.

5. The method for controlling excessive growth of cassava by drone spraying according to claim 1, characterized in that, In step (2), the fault tolerance threshold is 15%; when the area ratio of the abnormal region in a certain grid cell is greater than 15%, the grid cell is marked as an abnormal array point and assigned a value of 0; when the area ratio of the abnormal region in a certain grid cell is less than or equal to 15%, the grid cell is marked as a normal array point and assigned a value of 1.

6. The method for controlling excessive growth of cassava by drone spraying according to claim 1, characterized in that, In step (3), when the area of ​​the connected region is greater than or equal to the circular area corresponding to the minimum turning radius of the UAV, and the flight route needs to be replanned to bypass the flight obstacle area, a multi-objective optimization function is used to plan the detour path. The multi-objective optimization function uses the sum of the product of the total flight time and its weight coefficient and the product of the total energy consumption and its weight coefficient as the optimization objective to optimize the detour path.

7. The method for controlling excessive growth of cassava by drone spraying according to claim 6, characterized in that, When planning the detour path using the multi-objective optimization function, kinematic constraints for the agricultural drone are also introduced, specifically including: constraining the instantaneous flight speed of the drone within a preset maximum flight speed, constraining the instantaneous acceleration within a preset maximum acceleration, and constraining the curvature of the flight path within the reciprocal of a preset minimum turning radius; and ensuring the drone's flight attitude remains stable during the detour by setting smooth trajectory transition point coordinates.

8. The method for controlling excessive growth of cassava by drone spraying according to claim 1, characterized in that, In step (3), when the area of ​​the connected region is greater than or equal to the circular area corresponding to the minimum turning radius of the UAV, the connected region is marked as a flight obstacle zone using an improved ant colony algorithm, and the three-dimensional flight path of the UAV is replanned to bypass the flight obstacle zone.

9. The method for controlling excessive growth of cassava by drone spraying according to claim 1, characterized in that, In step (4), real-time dynamic differential positioning is used to obtain the centimeter-level three-dimensional coordinates of the UAV, and the centimeter-level three-dimensional coordinates are compared with the binarized state matrix at high frequency.

10. The method for controlling excessive growth of cassava by drone spraying according to claim 1, characterized in that, In step (4), the advance trigger distance is determined based on the product of the current forward speed of the UAV and the mechanical delay time of the spraying system; when the UAV approaches the boundary of the abnormal array point and the distance is equal to the advance trigger distance, the shutdown command is issued in advance, and when the UAV leaves the boundary of the abnormal array point and the distance is equal to the advance trigger distance, the start command is issued in advance.