Plant protection unmanned aerial vehicle operation method integrating weed recognition and real-time variable spraying

Abstract

The invention discloses a plant protection unmanned aerial vehicle operation method fusing weed recognition and real-time variable spraying, and the method comprises the following steps: employing a plant protection unmanned aerial vehicle to carry a multispectral sensor, collecting multispectral data in real time, and generating a high-resolution multispectral image; converting the radiation intensity value into an actual reflectivity value by using a spectral reflectivity standard plate; according to the generated multispectral image, an NDVI value is calculated in real time through an airborne calculation module; in combination with flight path data, mapping the NDVI index and the geographic position of the flight operation; the classification standard of the NDVI value is adjusted in combination with a machine learning algorithm, and a spraying operation prescription map is generated; the electronic control spraying device adjusts the spraying flow in real time; and establishing a spraying model, and calculating an adjustment coefficient alpha to obtain the average pesticide application amount Y per mu. According to the plant protection unmanned aerial vehicle operation method, the spraying amount can be adjusted according to field weed distribution and crop growth vigor, the precision of plant protection operation is improved, the pesticide usage amount is reduced, and environmental pollution is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural plant protection, and in particular to a method for operating a plant protection unmanned aerial vehicle (UAV) based on intelligent recognition and precise pesticide application technology. Specifically, it is an intelligent plant protection operation method that integrates multi-spectral remote sensing image processing, target recognition, and precise variable spraying control. The aim is to solve the problems of low recognition accuracy and poor pesticide application uniformity in traditional plant protection UAV operations, improve the intelligent level of operations, and promote the development of agricultural plant protection towards precision. Background Art

[0002] In recent years, as a new type of agricultural aviation equipment, plant protection UAVs have demonstrated significant technical advantages and application values in the fields of field management and monitoring. Compared with traditional ground mechanical operation methods, plant protection UAVs have outstanding characteristics such as high operation efficiency, low operation cost, and less crop damage in the early monitoring of crop diseases and pests, precise pesticide spraying, and fertilizer spreading. They have been widely recognized by agricultural growers. However, the existing plant protection UAV spraying operations usually adopt a uniform spraying method across the whole field, with insufficient recognition ability for complex weeds and crops in the field. Their operation planning and pesticide application decisions have not yet achieved "one flight for multiple pest control", that is, they cannot perform targeted variable spraying of herbicidal pesticides on the discovered weeds in the field during the process of route planning and pesticide application for crops. Traditional field weed identification and control mainly rely on manual labor. Operators need to observe the growth status of field weeds with the naked eye and manually control the plant protection UAV for aerial spraying operations. This operation method has obvious limitations: one is the difficulty in achieving real-time dynamic response during flight, resulting in low operation efficiency; the second is the limited accuracy of manual identification, prone to weed omission; the third is the inaccurate control of the pesticide application amount, often resulting in over-application or under-application of pesticides. These problems not only increase the operation cost but also lead to waste of pesticide resources, poor pesticide application uniformity, and further cause secondary problems such as environmental pollution. Therefore, there is an urgent need for a method for operating a plant protection UAV that can achieve intelligent weed identification and real-time variable spraying to improve the operation accuracy and resource utilization efficiency of plant protection UAVs and promote the development of agricultural plant protection towards intelligence and precision. Summary of the Invention

[0003] In order to overcome the deficiencies existing in the prior art, the present invention proposes a method for operating a plant protection UAV that integrates weed identification and real-time variable spraying, to solve the problems of inaccurate spraying, inaccurate identification, and resource waste during the operation of plant protection UAVs in the prior art.

[0004] The technical solution of the present invention to solve the above technical problems is as follows:

[0005] A method for operating a plant protection UAV that integrates weed identification and real-time variable spraying, characterized by comprising the following steps:

[0006] a. Utilize a multi-rotor plant protection UAV equipped with a multi-spectral camera sensor, which can collect multi-spectral data including the red light band (RED) and the near-infrared band (NIR) in real time. When the plant protection UAV conducts cruise operations above the field according to the preset flight path, the multi-spectral camera obtains the spectral reflectance of each area of the field vegetation canopy through high-frame-rate imaging, generating a high-resolution multi-spectral image.

[0007] b. Use a spectral reflectance standard plate for reflectance standardization to convert the radiation intensity value in the multi-spectral image into the actual reflectance value.

[0008] c. According to the collected multi-spectral data, through the on-board computing module, calculate the NDVI value in real time during the flight of the plant protection UAV. Combine the flight trajectory (GPS data) and flight path data of the plant protection UAV to map the NDVI index to the geographical location of the operation field block, generating an instant NDVI distribution map.

[0009] d. Combine machine learning algorithms to dynamically adjust the classification criteria of NDVI values. For different crop growth stages and environmental conditions, the field can be divided into several NDVI regions of different grades to achieve accurate identification and classification.

[0010] e. Establish a spraying model according to the corresponding relationship between the NDVI value and the spraying amount. The formula is:

[0011] Y = α × A,

[0012] where Y is the average liquid medicine spraying amount per mu, α is the adjustment coefficient, A is the standard liquid medicine spraying amount per mu, and T weed is the NDVI threshold.

[0013] When it is higher or lower than the threshold, conduct weeding or defoliant spraying operations, and calculate the adjustment coefficient α according to the T weed threshold under the NDVI grade classification, and calculate the average liquid medicine spraying amount Y per mu in the weed area and the crop variable spraying area.

[0014] f. Generate a spraying operation prescription map according to the NDVI classification result. The prescription map is dynamically associated with the UAV flight path to ensure real-time adjustment during flight. Mark the areas that need to be sprayed with variable amounts at fixed points, and perform variable spraying of weeding pesticides on the abnormal NDVI value areas covered with weeds.

[0015] g. In combination with the flight path of the unmanned aerial vehicle, the plant protection unmanned aerial vehicle is equipped with an electronically controlled spraying device, which includes an electronically adjustable flow control valve and a nozzle. The electronically adjustable flow control valve automatically adjusts the opening and closing state of the nozzle and the spraying flow rate according to the NDVI regional data, automatically opens the nozzle for spraying herbicides in the weed identification area, and adjusts the spraying flow rate in real time according to the weed density to achieve the spraying operation.

[0016] In a preferred embodiment of the present invention, in step (a), the multispectral camera sensor can simultaneously collect spectral data in the red light band (RED) and the near-infrared band (NIR), and these two bands help to improve the distinguishability between crops and weeds.

[0017] In a preferred embodiment of the present invention, in step (a), the high-frame rate imaging technology can collect high-resolution images in real time, can capture subtle vegetation changes, and ensure accurate monitoring of the crop health status and weed distribution.

[0018] In a preferred embodiment of the present invention, in step (b), the spectral reflectance standard plate can perform reflectance normalization to ensure the consistency of all collected image data.

[0019] In a preferred embodiment of the present invention, in step (c), the on-board computing module can perform lightweight real-time processing of multispectral data to generate an instant prescription map. The calculation formula for the NDVI value is:

[0020]

[0021] Where NDVI is the normalized difference vegetation index, NIR is the reflectance value in the near-infrared band, and Red is the reflectance value in the red light band.

[0022] In a preferred embodiment of the present invention, in step (d), through a machine learning algorithm, the remote sensing images are trained to accurately identify the weeds in the crops under different scenarios and different growth stages of the crops. By using the difference between the NDVI values of the normal crops around the weeds and the NDVI values of the weed areas, a threshold is set. When the difference between the two is greater than or less than the threshold, the weed area is identified.

[0023] In a preferred embodiment of the present invention, in step (e), the spraying model can divide the interval according to the NDVI threshold obtained from the remote sensing information on the basis of the standard liquid application rate per mu, calculate the adjustment coefficient α, and thus obtain the average liquid application rate per mu Y.

[0024] In a preferred embodiment of the present invention, in step (f), according to the RKT positioning of the plant protection UAV, the area where the abnormal NDVI value is located is determined, the operation prescription map is associated with the flight path of the plant protection UAV, and variable spraying operation is carried out on the area where the abnormal NDVI value is located.

[0025] In a preferred embodiment of the present invention, in step (g), the electronic control spraying device includes an electronic flow control valve and a nozzle, and is controlled by the on-board main controller. The electronic flow control valve is connected to the nozzle and the herbicide tank mounted on the plant protection UAV through a liquid pipe. When a weed area is identified, the main controller issues an instruction to the electronic control spraying device, and the electronic flow control valve automatically opens, closes and adjusts the nozzle according to the operation prescription map to carry out weeding operation.

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

[0027] 1. The present invention collects data in the red (RED) and near-infrared (NIR) bands through a multispectral camera sensor, calculates the NDVI value in real time, and generates an instant NDVI distribution map in combination with the flight path data to achieve precise differentiation between crops and weeds. Compared with the prior art, the present invention provides higher recognition accuracy and real-time performance, ensures accurate recognition of weeds and crops, and thus realizes precise variable spraying.

[0028] 2. The present invention combines machine learning algorithms to dynamically adjust the classification criteria of NDVI values according to different crop growth stages and field environmental conditions to ensure accurate recognition of weeds and crops. Compared with the fixed criteria of traditional technologies, the dynamic adjustment mechanism of the present invention can effectively cope with changes in different seasons and environments, and improves the flexibility and adaptability of recognition.

[0029] 3. By generating a spraying operation prescription map and dynamically associating it with the UAV flight path, the present invention realizes precise adjustment of the spraying amount. Combining real-time NDVI data, the UAV can automatically adjust the spraying flow rate during flight to ensure that pesticides are only used in areas with dense weeds, effectively reducing pesticide waste and improving operation efficiency. Compared with traditional technologies, the present invention can more precisely control the use of pesticides and reduce environmental pollution. Brief Description of the Drawings

[0030] Figure 1 It is a working flow chart of the plant protection UAV operation method for integrating weed recognition and real-time variable spraying of the present invention. Detailed Embodiments

[0031] The present invention will be further described in detail below in conjunction with embodiments and the accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0032] In the mechanized cotton harvesting operations in Xinjiang Uygur Autonomous Region, weeds in the fields can entangle the spindles of cotton pickers during the mechanized harvesting process, causing mechanical failures and a decline in harvesting efficiency, thus hindering the operation of cotton pickers. Therefore, weed control operations need to be carried out during the stage of spraying defoliants. Currently, in cotton fields, the common weed is Solanum nigrum, which is densely distributed among normally growing cotton plants, forming a serious weed infestation. During the film-covering application of pesticides in the cotton seedling stage, it is often difficult to effectively control such weed infestations. Therefore, it is particularly important to conduct remote sensing monitoring of weeds and variable spraying during the boll-opening stage of cotton.

[0033] See Figure 1 , the plant protection UAV is equipped with a multispectral camera sensor to simultaneously collect spectral information in the red light band (RED, 650 - 680 nm) and the near-infrared band (NIR, 750 - 900 nm), and conduct cruise operations according to the preset flight path (flight altitude is 2 meters, flight speed is 5 m / s), and the flight path is accurately positioned by GPS. The standard board is placed in a specific area of the cotton field to ensure that its reflectance value is consistent with the actual environment. Through the reflectance value of the standard board, the radiation intensity value in the multispectral image is converted into the actual reflectance value to ensure the consistency of all collected image data. The normalized difference vegetation index (NDVI) is calculated in real time by the on-board computing module. Combining with the GPS data, the NDVI value is mapped to the geographical location of the cotton field to generate an instant NDVI distribution map. The convolutional neural network (CNN) model is used to train the multispectral image, dynamically adjust the NDVI classification standard, and accurately identify Solanum nigrum weeds.

[0034] Before large-scale defoliant spraying operations are carried out on cotton, the canopy colors of weeds and cotton in the fields are similar, the NDVI differences are not significant, and the values are concentrated in the range of 0.7 - 1, mainly for variable spraying operations of cotton defoliants. The standard application liquid volume of cotton defoliants per mu is 2 liters. In the T weed area where T is between 0.9 - 1, the vegetation coverage of the cotton canopy is high, and the adjustment coefficient α = 1. According to the spraying model, Y = 2 liters / mu. In the T weed area where T is between 0.8 - 0.9, the vegetation coverage of the cotton canopy is relatively high, and the adjustment coefficient α = 0.9. According to the spraying model, Y = 1.8 liters / mu. In the T weedIn the area between 0.7 and 0.8, the vegetation coverage of the cotton canopy decreases, and the natural defoliation rate of the cotton in this area increases. The use of defoliant should be reduced. The adjustment coefficient α = 0.8. According to the spraying model, Y = 1.6 liters per mu. According to the generated spraying operation prescription map and combined with the UAV flight path, the electronic flow control valve in the electronic control spraying device adjusts the liquid medicine flow rate in real time and controls the on / off state of the nozzle. In the first operation of spraying cotton defoliant, three application parameter specifications are executed in areas with different canopy coverage degrees. When the second spraying operation of cotton defoliant is carried out at an interval of 10 - 15 days, at this time, the cotton leaves have fallen off significantly, while the weeds have high vegetation coverage due to their resistance to cotton defoliant, and there is a large color difference between them and the surrounding cotton plants. At this time, targeted weeding operations are carried out on areas with a higher NDVI threshold. The standard application liquid volume of cotton herbicide per mu is 1 liter, at T weed In the area between 0.9 and 1, the vegetation coverage of the weeds is high. The adjustment coefficient α = 1. According to the spraying model, Y = 1 liter per mu, at T weed In the area between 0.8 and 0.9, the vegetation coverage of the weeds is relatively high. The adjustment coefficient α = 0.85. According to the spraying model, Y = 0.85 liters per mu. Based on the parameters of the spraying weed operation prescription map and combined with the real-time UAV flight path data, the liquid medicine flow rate is dynamically adjusted through the electronic flow control valve, and the opening and closing states of the nozzles are accurately controlled synchronously. Compared with the traditional plant protection operation method, this method accurately controls the amount of pesticide used, improves the operation accuracy and resource utilization efficiency of the plant protection UAV, and reduces environmental pollution.

[0035] The above is a preferred embodiment of the present invention, but the embodiments of the present invention are not limited to the above content. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. A plant protection drone operation method integrating weed identification and real-time variable spraying, characterized in that: The following steps are involved: a. Use a multi-rotor agricultural drone equipped with a multispectral camera sensor, which can collect multispectral data including red light band (RED) and near infrared band (NIR) in real time. When the agricultural drone cruises over the field according to the preset flight path, the multispectral camera obtains the spectral reflectance of each area of ​​the field vegetation canopy through high frame rate imaging, generating high-resolution multispectral images. b. Use spectral reflectance standard plates to standardize reflectance and convert the radiation intensity values ​​in multispectral images into actual reflectance values. c. Based on the collected multispectral data, the NDVI value is calculated in real time during the flight of the plant protection UAV through the airborne computing module. Combined with the flight trajectory (GPS data) and flight path data of the plant protection UAV, the NDVI index is mapped with the geographical location of the working field to generate a real-time NDVI distribution map. d. Combined with machine learning algorithms to adjust the classification of NDVI values, the fields can be divided into several different levels of NDVI areas according to different crop growth stages and environmental conditions, so as to achieve accurate identification and classification. e. Generate a spraying prescription map based on the NDVI classification results. The prescription map is dynamically associated with the UAV flight path to ensure real-time adjustment during the flight. Mark the areas that require fixed-point variable spraying, and perform variable spraying of herbicides in areas with abnormal NDVI values ​​covered by weeds. f. According to the corresponding relationship between NDVI value and spraying amount, a spraying model is established, and its formula is: Where Y is the average amount of liquid applied per mu, α is the adjustment coefficient, A is the standard amount of liquid applied per mu, T weed is the NDVI threshold. When the threshold is above or below, weed control or defoliant spraying is carried out, and the T weed The threshold value is used to calculate the adjustment coefficient α, and the average amount of liquid sprayed per mu Y in the weed area and the crop variable spraying area is calculated. g. Combined with the flight path of the drone, the plant protection drone is equipped with an electronically controlled spraying device, which includes an electronic flow control valve and a nozzle that can be adjusted in real time. The electronic flow control valve automatically adjusts the switch state and spraying flow of the nozzle according to the NDVI area data, automatically opens the nozzle for spraying herbicides in the weed identification area, and adjusts the spraying flow in real time according to the density of the weeds to achieve the spraying operation.

2. The plant protection drone operation method integrating weed identification and real-time variable spraying according to claim 1 is characterized in that: In step a, when the plant protection UAV cruises above the field according to a preset flight path, the multispectral sensor camera it carries can collect multispectral data including the red light band (RED) and near-infrared band (NIR) in real time, and generate high-resolution multispectral images after lightweight processing by the onboard computing module.

3. The plant protection drone operation method integrating weed identification and real-time variable spraying according to claim 1 is characterized in that: In step c, according to the RKT positioning of the plant protection UAV, the NDVI index is mapped to the geographical location of the working field, so that the working prescription map is associated with the flight path of the plant protection UAV to generate a real-time NDVI distribution map.

4. The plant protection drone operation method integrating weed identification and real-time variable spraying according to claim 1 is characterized in that: In step d, remote sensing images are trained through machine learning so that the plant protection drone can accurately identify weeds in crops under different environments and in different crop categories.

5. The plant protection drone operation method integrating weed identification and real-time variable spraying according to claim 1 is characterized in that: In step g, the plant protection drone is equipped with an electronically controlled spraying device, which includes an electronic flow control valve and a nozzle that can be adjusted in real time. The electronic flow control valve is connected to the nozzle and the herbicide box mounted on the plant protection drone through a liquid pipe. When weeds are identified, the main controller issues instructions to the electronically controlled spraying device, and the electronic flow control valve automatically switches and adjusts the nozzle according to the operation prescription map.

Images