Method, device and system for automatically removing tassels by unmanned aerial vehicle based on machine vision

Abstract

The invention discloses a method, device and system for automatically removing tassels by an unmanned aerial vehicle based on machine vision, and the method comprises the steps: obtaining the longitude and latitude information of a target region, a corn RGB image and a corn depth image, carrying out the preprocessing of the corn RGB image, and obtaining the longitude and latitude information of each corn tassel and the average plant height of a corn population; obtaining an unmanned aerial vehicle castration path based on a preset unmanned aerial vehicle flight direction and the longitude and latitude information of each corn tassel; determining a final cutting point of the target corn tassel based on the average plant height of the corn population and the unmanned aerial vehicle emasculation path; if the final cutting point of the target corn tassel is shielded, the real-time pose of the unmanned aerial vehicle is adjusted until a preset condition is met, and then the unmanned aerial vehicle is controlled to remove the target corn tassel according to the final cutting point. According to the unmanned aerial vehicle castration method and the implementation scheme thereof, the corn castration precision is remarkably improved, the castration missing rate is reduced, the stability and adaptability of a castration system are enhanced, and the operation convenience and expansibility are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent application of unmanned aerial vehicles, and in particular to a method, device and system for automatically removing male spikes by unmanned aerial vehicles based on machine vision. Background Art

[0002] Corn detasseling is of great significance to corn scientific research and breeding, gene sequencing, phenotypic research, etc. At present, corn detasseling is mainly done manually. With the development of artificial intelligence and its application in the field of smart agriculture, automatic detasseling by unmanned vehicles combined with machine vision technology is also gradually developing, but both manual detasseling and automatic detasseling by unmanned vehicles have the problem of missing.

[0003] At present, the intelligent application of corn emasculation is mainly used for tassel detection. The Chinese patent (publication number is CN218766644U, and the publication date is March 28, 2023) discloses a emasculation detection device in corn production, which realizes the removal of corn tassels by connecting the rotating column and shearing device of the cylinder on the drone and assisting the visual solution. Due to the high working power and weight of the cylinder rotating device, the power and size requirements of the drone are high. When cutting the tassel at close range, it is easy to cause the corn plant to shake and the stability is poor. The Chinese patent (publication number is CN115713697A, and the publication date is February 24, 2023) discloses an SSD corn tassel detection method, which realizes offline tassel detection by inspecting the field and collecting images by drones. A Chinese patent (publication number CN114402995A, publication date April 29, 2022) discloses a method for removing corn tassels by combining a drone and an unmanned vehicle. The drone inspects and takes sequential images of corn, and offline analysis is used to detect and locate the corn tassels. The tassel position information is then transmitted to the unmanned vehicle, which then moves to the designated location for removal of tassels.

[0004] In the current field of corn detasseling technology, although manual detasseling has long existed as a traditional method, its problems of low efficiency, high labor intensity and high cost are becoming increasingly prominent. With the development of intelligent technology, detasseling technology combining unmanned vehicles and machine vision has gradually emerged, providing new possibilities for automated detasseling. However, these technologies still face many challenges in practical applications: (1) The rate of missing tassels is high. Whether it is manual emasculation or an automatic emasculation system that combines an unmanned vehicle with machine vision, it is difficult to completely avoid the phenomenon of missing tassels. In particular, the accuracy of detection and removal of tassels that are hidden or have special shapes needs to be improved.

[0005] (2) There are problems with the complexity and stability of the detasseling system. Existing unmanned vehicle or drone detasseling systems often rely on complex mechanical devices (such as cylinder rotation devices) and high-precision visual recognition algorithms, which not only increases the weight and power consumption of the system, but also increases the difficulty of implementation and maintenance costs. At the same time, close-range operations can easily cause the plants to shake, affecting the precision and stability of detasseling.

[0006] (3) The emasculation system has poor adaptability. The morphology of corn tassels varies in different growth environments and at different growth stages. The existing system is often unable to flexibly adapt to these changes, resulting in different emasculation effects. Its versatility and robustness need to be improved.

[0007] (4) High hardware requirements. High-power mechanical devices and complex visual processing systems place higher demands on the size, power and endurance of drones, limiting their widespread application in actual field operations. Summary of the invention

[0008] In view of the shortcomings of the prior art, the present invention provides a method, device and system for automatically removing male tassels by drone based on machine vision, which significantly improves the automation level, accuracy and stability of corn detasseling by drone.

[0009] In order to solve the above technical problems, the present invention is solved by the following technical solution: a method for automatically removing male spikes by a drone based on machine vision, comprising the following steps: Acquire the latitude and longitude information of the target area, the corn RGB image and the corn depth image, pre-process the corn RGB image, and obtain a transformed corn RGB image; Based on the transformed corn RGB image, the corn depth image and the longitude and latitude information of the target area, combined with the corn tassel detection model, the longitude and latitude information of each corn tassel and the average plant height of the corn population are obtained; Based on the preset UAV flight direction and the latitude and longitude information of each corn tassel, a UAV detasseling path is obtained; Based on the average plant height of the corn population and the tassel removal path of the drone, the target corn tassel in the target area is located in combination with the real-time position of the drone, and then the final cutting point of the target corn tassel is determined; If the final cutting point of the target corn tassel is blocked, the real-time position and posture of the drone is adjusted until the preset conditions are met, and then the drone is controlled to remove the target corn tassel according to the final cutting point.

[0010] As an implementable method, the method is based on the transformed corn RGB image, the corn depth image and the longitude and latitude information of the target area, combined with the corn tassel detection model, to obtain the longitude and latitude information of each corn tassel and the average plant height of the corn population, including the following steps: Based on the transformed corn RGB image and the corn tassel detection model, the corn tassel rectangular frame is obtained; Based on the rectangular frame of the corn tassel, the center position information of the corn tassel is obtained; Based on the center position information of the corn tassel and the longitude and latitude information of the target area, the longitude and latitude information of each corn tassel is obtained; Based on the center position information of the corn tassel, the average center coordinates of the corn tassel are obtained; The average plant height of the corn population was obtained based on the average center coordinates of the corn tassel and the corn depth image.

[0011] As an implementable method, obtaining the drone emasculation path based on the preset drone flight direction and the latitude and longitude information of each corn tassel comprises the following steps: Based on the preset UAV flight direction, the latitude and longitude information of each corn tassel is sorted to obtain an inspection point queue; Based on the order of the longitude and latitude information of each corn tassel in the inspection landing point queue, the drone detasseling path is obtained.

[0012] As an implementable method, the target corn tassel in the target area is located by the drone detasseling path and combined with the real-time position of the drone, and then the final cutting point of the target corn tassel is determined, including the following steps: Based on the average plant height of the corn population, the tassel removal path of the drone and the real-time position of the drone, the drone is controlled to reach the target position, wherein the target position includes the initial hovering latitude and longitude and the initial hovering height of the drone; Acquire a first RGB image of a target corn tassel and a first depth image of a target corn tassel in a target area; Based on the first RGB image of the target corn tassel, in combination with the corn tassel detection model, a tassel rectangular frame of the target corn tassel is determined to complete the positioning of the target corn tassel; Based on the tassel rectangular frame of the target corn tassel and the first depth image of the target corn tassel, an initial cutting point of the target corn tassel is obtained; Adjust the hovering height and longitude and latitude of the drone based on a first straight-line distance between the target corn tassel and the drone imaging unit and a second straight-line distance between the drone imaging unit and the target corn tassel initial cutting point, wherein the target corn tassel initial cutting point is located at the center of the drone imaging unit; A second RGB image of the target corn tassel and a second depth image of the target corn tassel are obtained, and a final cutting point of the target corn tassel is obtained by combining the corn tassel detection model.

[0013] As an implementable method, if the final cutting point of the target corn tassel is blocked, the real-time posture of the drone is adjusted, specifically: The drone is controlled to rotate around the central axis at a preset movement angle and a preset movement radius to adjust the real-time posture of the drone, wherein the central axis is the central axis perpendicular to the ground where the final cutting point of the target corn tassel is located, and the preset movement radius is the horizontal distance between the drone imaging unit and the final cutting point of the target corn tassel.

[0014] As an implementable method, adjusting the hovering height and longitude and latitude of the drone based on a first straight-line distance between the target corn tassel and the drone imaging unit and a second straight-line distance between the drone imaging unit and the initial cutting point of the corn tassel comprises the following steps: Controlling the UAV to move a first adjustment distance in a vertical direction to complete the adjustment of the hovering height of the UAV; Control the UAV to move in the horizontal direction for a second adjustment distance to complete the adjustment of the longitude and latitude of the UAV; Wherein, the first adjustment distance = the first straight line distance - the second straight line distance * sinθ; The second adjustment distance = the second straight line distance * cosθ; Hovering height = initial hovering height - second adjustment distance; Initial hovering height = average plant height of corn group + drone adjustment coefficient; The θ is the angle between the first straight line and the horizontal plane, and the UAV adjustment coefficient is related to the configuration of the UAV.

[0015] As an implementable method, preprocessing the corn RGB image to obtain a transformed corn RGB image includes the following steps: Acquire width data and height data of the corn RGB image, and determine the longest side of the corn RGB image based on the width data and the height data; Based on the longest side and the preset image size, the image transformation rate is obtained; Performing image transformation based on the image transformation rate to obtain a transformed corn RGB image, specifically: when the image transformation rate=1, keeping the corn RGB image unchanged; when the image transformation rate>1, proportionally reducing the width data and height data of the corn RGB image based on the image transformation rate; when the image transformation rate<1, keeping the resolution unchanged, and enlarging the corn RGB image; The image transformation rate is expressed as follows: R=maxLength / W; Among them, R represents the image transformation rate, maxLength represents the longest side, and W represents the side length of the preset image size.

[0016] As an implementable method, the maize RGB image is enlarged, comprising the following steps: Creating a new image with the same size as the preset image; Based on the size side length of the new image, the width data and the height data of the corn RGB image, a horizontal offset and a vertical offset are obtained, and the corn RGB image is located at the center of the new image; Each pixel of the corn RGB image is copied to the new image according to the horizontal offset and the vertical offset.

[0017] As an implementable method, the target corn tassel is removed by controlling a laser set on a drone, and the laser of the drone is controlled to extend according to a preset extension distance, wherein the drone laser is located on the drone body.

[0018] As an implementable method, the corn RGB image is collected by an imaging unit, and the target corn tassel is removed by controlling a laser set on a drone, wherein an optical axis of the imaging unit is parallel to a laser optical axis emitted by the laser, and a line connecting an optical center of the imaging unit and an emission position of the laser optical axis is perpendicular to both the optical axis of the imaging unit and the laser optical axis.

[0019] A system for automatically removing male tassels by using a drone based on machine vision, comprising an image acquisition and transformation module, a latitude, longitude and altitude calculation module, a tassel removal path calculation module, a tassel cutting point calculation module and a posture adjustment and tassel removal module; The image acquisition and transformation module acquires the latitude and longitude information of the target area, the corn RGB image and the corn depth image, and pre-processes the corn RGB image to obtain a transformed corn RGB image; The latitude, longitude and height calculation module obtains the latitude and longitude information of each corn tassel and the average plant height of the corn population based on the transformed corn RGB image, the corn depth image and the latitude and longitude information of the target area in combination with the corn tassel detection model; The emasculation path calculation module obtains the emasculation path of the drone based on the preset drone flight direction and the longitude and latitude information of each corn tassel; The tassel cutting point calculation module locates the target corn tassel in the target area based on the average plant height of the corn population and the tassel removal path of the drone, and combines the real-time position of the drone to determine the final cutting point of the target corn tassel; The posture adjustment and tassel removal module adjusts the real-time posture of the drone if the final cutting point of the target corn tassel is blocked until the preset conditions are met, and then controls the drone to remove the target corn tassel according to the final cutting point.

[0020] A device for automatically removing male tassels by using a drone based on machine vision comprises a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the following method is implemented: Acquire the latitude and longitude information of the target area, the corn RGB image and the corn depth image, pre-process the corn RGB image, and obtain a transformed corn RGB image; Based on the transformed corn RGB image, the corn depth image and the longitude and latitude information of the target area, combined with the corn tassel detection model, the longitude and latitude information of each corn tassel and the average plant height of the corn population are obtained; Based on the preset UAV flight direction and the latitude and longitude information of each corn tassel, a UAV detasseling path is obtained; Based on the average plant height of the corn population and the tassel removal path of the drone, the target corn tassel in the target area is located in combination with the real-time position of the drone, and then the final cutting point of the target corn tassel is determined; If the final cutting point of the target corn tassel is blocked, the real-time position and posture of the drone is adjusted until the preset conditions are met, and then the drone is controlled to remove the target corn tassel according to the final cutting point.

[0021] The present invention has significant technical effects due to the adoption of the above technical solution: (1) Improve emasculation accuracy and reduce omission rate. By integrating advanced machine vision algorithms and drone platforms, the present invention can achieve accurate detection and positioning of corn tassels and significantly reduce the omission rate. At the same time, combined with the automatic adjustment technology of the drone's posture, the precise docking of the emasculation unit and the tassel cutting point is ensured, thereby improving the accuracy and stability of emasculation.

[0022] (2) Enhance system stability and adaptability. The automatic planning algorithm for the UAV detasseling path designed by the present invention can dynamically adjust the flight path according to the actual height and position information of the corn tassel, avoiding plant shaking caused by close-range operation. In addition, the system can adapt to corn tassels in different growth conditions, improving versatility and robustness.

[0023] (3) Improve operational convenience and expandability. The drone imaging unit and detasseling unit provided by the present invention are reasonably designed and easy to install and debug.

[0024] (4) Reduce hardware requirements and costs. The present invention can use a laser transmitter as a demaleation unit. Compared with traditional mechanical shearing devices, it has the advantages of small size, light weight, and low power consumption, which significantly reduces the requirements for drone hardware. At the same time, the laser removal method simplifies the mechanical structure and reduces the complexity and maintenance cost of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0026] Figure 1 It is a schematic diagram of the overall process of the method of the present invention; Figure 2 This is a schematic diagram of the structure of a drone according to an embodiment of the present invention; Figure 3 It is a schematic diagram of corn detection results of the present invention; Figure 4 This is a schematic diagram of the effect of the secondary flight planning path of the UAV according to an embodiment of the present invention; Figure 5 Schematic diagram of various distances during the posture adjustment process of the drone according to an embodiment of the present invention; Figure 6 This is a diagram showing the positioning effect of the cutting point of the target corn tassel according to an embodiment of the present invention; Reference numerals: 101, RGB image of corn crop; 102, corresponding depth image of corn crop; 301, drone; 303, imaging unit; 304, detasseling unit. DETAILED DESCRIPTION

[0027] The present invention is further described in detail below in conjunction with embodiments. The following embodiments are for explanation of the present invention but the present invention is not limited to the following embodiments.

[0028] Embodiment 1: A method for automatically removing tassels using a drone based on machine vision, such as Figure 1 As shown, the following steps are included S100: Acquire the latitude and longitude information of the target area, the corn RGB image and the corn depth image, and pre-process the corn RGB image to obtain a transformed corn RGB image; S200: Based on the transformed corn RGB image, the corn depth image and the longitude and latitude information of the target area, combined with a corn tassel detection model, the longitude and latitude information of each corn tassel and the average plant height of the corn population are obtained; S300: Obtaining a drone detasseling path based on a preset drone flight direction and the longitude and latitude information of each corn tassel; S400: Based on the average plant height of the corn population and the tassel removal path of the drone, the target corn tassel in the target area is located in combination with the real-time position of the drone, and then the final cutting point of the target corn tassel is determined; S500: If the final cutting point of the target corn tassel is blocked, the real-time posture of the drone is adjusted until the preset conditions are met, and then the drone is controlled to remove the target corn tassel according to the final cutting point.

[0029] By the method of the present invention, the corn RGB image and corn depth image of the target area are obtained, and on this basis, the longitude and latitude information of each corn tassel is obtained, and then the unmanned aerial vehicle emasculation path is obtained, and the unmanned aerial vehicle reaches the target position according to the unmanned aerial vehicle emasculation path and adjusts the posture to realize corn emasculation according to the target corn tassel final cutting point. This method can accurately detect and locate corn tassels by integrating machine vision algorithms and unmanned aerial vehicle platforms, and combines the automatic adjustment technology of unmanned aerial vehicle posture to ensure the accurate docking of emasculation units and tassel cutting points, improve the accuracy of emasculation, and significantly reduce the omission rate. At the same time, the method automatically plans the unmanned aerial vehicle emasculation path, and can dynamically adjust the flight path according to the actual height of the corn tassel and the longitude and latitude information of the tassel, avoid the plant shaking caused by close-range operation, so that the corn tassels of different growth conditions can be adapted, and the versatility and robustness are improved.

[0030] It should be noted that in actual farming, the first detasseling of corn tassels usually relies on manual operation. However, the drone detasseling scheme based on machine vision technology introduced in the present invention is intended to assist in the subsequent detasseling process after the first manual detasseling, whether it is a secondary detasseling or multiple detasseling. In addition, if the drone's endurance is sufficient to support it, the scheme can also be applied to the first detasseling operation. It is worth noting that the present invention does not limit its detasseling method to a specific detasseling stage of corn growth.

[0031] In S100, the latitude and longitude information of the target area, the corn RGB image and the corn depth image are obtained, and the corn RGB image is preprocessed to obtain a transformed corn RGB image, including the following steps: In this embodiment, a drone is provided, such as Figure 2 As shown, it is a schematic diagram of the structure of the drone of this embodiment. The drone 301 includes an imaging unit 303 and a detasseling unit 304. During the first inspection of the drone, the corn RGB image RGB0 and the corresponding corn depth image GS0 are obtained. 101 is a part of the corn RGB image, and 102 is the corn depth image corresponding to the RGB image 101. When the drone takes the corn RGB image, the GPS information (such as the latitude and longitude of the shooting location, altitude, etc.) is recorded at the same time. These GPS information can establish a corresponding relationship with each pixel in the corn RGB image. The grayscale value of the corn depth image represents the vertical distance between the drone imaging unit and the corn tassel or corn.

[0032] In this embodiment, the tassels of corn can be removed by mechanical means, such as cutting off the tassels, pulling out the tassels, etc., or by using lasers, chemical agents, or biological agents to remove the target tassels of corn. This application does not limit this, but the removal of the target tassels of corn by lasers, chemical agents, or biological agents is more effective. The emasculation unit corresponds to a laser emitter or a drug injector. In this embodiment, the emasculation unit does not need to contact the tassels of corn or be released at close range, and is not likely to cause the plants to shake, thereby improving the accuracy and stability of emasculation.

[0033] In the present embodiment, corn tassels can be effectively removed by a variety of methods, and these methods can be traditional mechanical operations, such as cutting and pulling out male tassels, or lasers and applied chemical or biological agents can be used for precise processing. This patent application does not set specific restrictions on this, and is intended to show the diversity and flexibility of technology. It is worth noting that by using lasers or chemical and biological agents of a specific ratio to remove the target corn tassel, it is often possible to show a more significant and efficient effect, because it can be efficiently removed under non-contact or remote operation, thereby significantly improving the accuracy and stability of emasculation operations. Therefore, in practical applications, selecting the optimal solution according to specific conditions and needs will be an important strategy for achieving efficient emasculation.

[0034] In this embodiment, in order to cooperate with the above-mentioned non-contact emasculation strategy, this embodiment is specially designed with an adapted emasculation unit, such as a laser emitter or a finely controlled drug sprayer. These devices are designed to avoid direct physical contact with the corn plants, effectively reducing the shaking of the plants that may be caused by mechanical contact, thereby ensuring the smooth progress of the emasculation process and higher operation accuracy. In addition, this non-contact operation method also helps to reduce additional damage to corn plants, promote the healthy growth of crops, and ultimately improve yield and quality.

[0035] In this embodiment, the emasculation unit is located in the fuselage of the drone, and the drone needs to calibrate the imaging unit 303 and the emasculation unit 304 of the drone before inspection. Taking laser emasculation as an example, the optical axis of the RGB-D camera of the imaging unit 303 is parallel to the laser optical axis emitted by the laser transmitter of the emasculation unit 304 through physical calibration, and the optical center of the imaging unit and the line connecting the initial position of the laser optical axis are perpendicular to the camera optical axis and the laser optical axis. The optical center of the drone imaging unit is the reference point of the imaging unit and the starting point of the camera optical axis, so that the direction of the laser emitted by the emasculation unit can be ensured to be consistent with the imaging center of the RGB-D camera of the imaging unit 303. This implementation uses a laser transmitter as a emasculation unit, which has the advantages of small size, light weight and low power consumption compared to traditional mechanical shearing devices, and significantly reduces the requirements for drone hardware. The laser removal method simplifies the mechanical structure and reduces the complexity and maintenance cost of the system. The same calibration method is also applicable to the calibration of imaging units and drug injectors.

[0036] In this embodiment, since the resolution of the drone image is relatively large, it is usually necessary to preprocess the input corn image to achieve image transformation. The image transformation includes cropping, fusion, filling, etc. to adapt to the size of the preset image of the corn tassel detection model. The image transformation includes the following steps: S110: Obtain width data and height data of the corn RGB image, and use the larger one of the width data and the height data as the longest side of the corn RGB image.

[0037] S120: Based on the longest side and the preset image size, an image transformation rate is obtained. The image transformation rate is expressed as follows: R=maxLength / W; Among them, R represents the image transformation rate, maxLength represents the longest side, and W represents the side length of the preset image size.

[0038] S130: Perform image transformation based on the image transformation rate to obtain a transformed corn RGB image, specifically: When the image transformation rate = 1, the corn RGB image remains unchanged; When the image conversion rate is greater than 1, the width data and height data of the corn RGB image are proportionally reduced based on the image conversion rate, that is, the longest side of the input image is reduced to W; When the image transformation rate is less than 1, the resolution is kept unchanged and the corn RGB image is enlarged.

[0039] In this embodiment, the maize RGB image is enlarged and processed, including the following steps: (1) Create a new image of the same size as the preset image. The initial pixel values ​​of all channels of the new image are set to a uniform value, which can be 0 (that is, black). (2) Based on the size side length of the new image, the width data and height data of the corn RGB image, the horizontal offset and vertical offset are obtained. The corn RGB image is located at the center of the new image. The horizontal offset is (W-width data) / 2, and the vertical offset is (W-height data) / 2. (3) Copy each pixel of the corn RGB image to the new image according to the horizontal offset and vertical offset. In this way, the original corn RGB image is placed at the center of the new image, and the rest of the new image remains black. The new image with the corn RGB image is the enlarged image.

[0040] This magnification processing method does not actually magnify the corn RGB image, but only puts the input original corn RGB image into a larger background. Because the UAV's flight altitude is relatively stable during image acquisition, the relative size of the target object is relatively stable. Therefore, the image size adjustment here is mainly to adjust the image size while ensuring that the target remains unchanged.

[0041] All units representing image sizes in image transformation methods are pixels.

[0042] In S200, based on the transformed corn RGB image, the corn depth image and the longitude and latitude information of the target area, combined with the corn tassel detection model, the longitude and latitude information of each corn tassel and the average plant height of the corn population are obtained, including the following steps: In this embodiment, a trained neural network model can be used as a corn tassel detection model. Since the development of neural networks for target detection is relatively mature, the present invention does not elaborate on the neural network model in detail, and the following suggestions are given for the selection of the corresponding neural network architecture: (1) Select a specific network model according to the actual application scenario. For corn tassel detection applications, since the size, posture, and occlusion of tassels in the same batch vary greatly, it is necessary to select a neural network model that can be applied to targets of different scales and has simple features; (2) Since the resolution of drone images is large, it is necessary to preprocess the original input image, including cropping, fusion, filling, etc. to adapt to the input image size of the neural network (that is, the preset image size of the corn tassel detection model). The corn tassel detection model in this application can determine the longitude and latitude of the corn tassel, locate the target corn tassel through a rectangular frame, and determine the tassel cutting point.

[0043] In this embodiment, the image preprocessing step can be embedded in the corn tassel detection model, and the image preprocessing can be performed at the input stage of the corn tassel detection model, or it can be independent of the corn tassel detection model and input into the corn tassel detection model after processing. The image preprocessing is not only used for the preparation of training data sets and test data sets in the training process of the neural network model, but also for the transformation of input images in the detection process of the corn tassel detection model, which can improve the efficiency of the neural network model training, thereby improving the detection accuracy of the image.

[0044] S210: Based on the transformed corn RGB image and the corn tassel detection model, a corn tassel rectangular frame maizeTasselRects is obtained, where maizeTasselRects={rect1,...,rect i ,...,rect n},i=1,2,...,n,rect iFor each corn tassel rectangular frame, the corn tassel rectangular frame can mark each corn tassel in the corn tassel RGB image after image transformation to achieve accurate spatial positioning of the corn tassel. The marking result schematic diagram is shown in Figure 3 As shown, it provides clear regional boundaries for subsequent image analysis and processing.

[0045] S220: Based on the corn tassel rectangular frame, obtain the corn tassel center position information maizeTasselCenterPoints, where maizeTasselCenterPoints={point1,...,point i ,...,point n}, point i The center position information of each corn tassel; S230: Based on the maize tassel center position information maizeTasselCenterPoints and the latitude and longitude information of the target area, the longitude and latitude information of each maize tassel is obtained.

[0046] S240: Based on the corn tassel center position information maizeTasselCenterPoints, the average center coordinates aveCenterPoint of the corn tassel are obtained.

[0047] S250: Based on the average center coordinates aveCenterPoint of the corn tassel and the corn depth image, the average plant height aveHeight of the corn population is obtained, which is used to determine the flight height of the drone during the emasculation process and the initial hovering height of the drone during emasculation.

[0048] Furthermore, based on the grayscale information of the corn depth image, the height h of each corn tassel from the ground can be obtained. i .

[0049] In S300, if Figure 4 As shown, based on the preset UAV flight direction and the longitude and latitude information of each corn tassel, the UAV detasseling path is obtained, including the following steps: S310: Based on the preset UAV flight direction, the longitude and latitude information of each corn tassel is sorted to obtain an inspection point queue.

[0050] S320: Obtain the UAV detasseling path based on the order of the longitude and latitude information of each corn tassel in the inspection landing point queue.

[0051] Corn fields often present irregular shapes due to factors such as terrain, irrigation conditions, and historical planting habits, which makes it difficult to plan the path for drone emasculation, especially the preset flight direction for drone emasculation. In this embodiment, the selection of a specific preset drone flight direction is not limited. For example, the corn field can be considered as a rectangle, a circle, an ellipse, or other shapes, and the software model can be used to make a selection based on the actual situation of the corn field.

[0052] If the cornfield is considered as a rectangle, the preset drone flight direction can be set to: Based on the corn RGB image, the first column direction W, the second column direction E, the first row direction S, the second row direction N and the ridge lines of each column of corn are obtained, and then the first working position of the drone is determined and added to the inspection landing point queue, wherein the first working position of the drone is the longitude and latitude of the corn tassel in the first column direction E and the first row direction S. The working position includes the longitude and latitude information of each corn tassel, and the inspection landing point queue is a queue arranged according to the order in which the various working positions of the drone are added. The main flight direction of the UAV flight path can be preset as flying from the first row direction S to the second row direction N, and the column flight direction is in the "M" form, and then it is determined whether there are other corn tassels in the same column of the first working position: if so, the longitude and latitude information of the corn tassels in the same column are added to the inspection drop point queue in sequence from the first column direction W to the second column direction E; if not, the nearest column where the tassels exist in the second row direction N is obtained, and the longitude and latitude of the corn tassels in the first column direction W of the nearest column are added to the inspection drop point queue as the next working position of the UAV, until the longitude and latitude information of all corn tassels are added to the inspection drop point queue as the working positions of the UAV.

[0053] It should be noted that the directions W, E, S, and N do not refer to the actual west, east, south, and north. They are relative directions set relative to the rectangular corn field. They are just letters representing a certain direction, but W and E as well as S and N refer to two corresponding directions. It should also be noted that the directions W, E, S, and N do not refer to the actual directions, because the W, E, S, and N of the field in reality do not strictly correspond to the geographical directions, and do not completely match the actual east-west / north-south directions, and together constitute the four directions that define the layout of the corn field. Therefore, the directions W, E, S, and N are only relative directions of the field. For example, in the corn fields in the north, the general planting method is east-west as the row direction and north-south as the column direction. The east can be the first row direction, the west can be the second row direction, the south can be the first column direction, and the north can be the second column direction.

[0054] If the cornfield is regarded as a circle or an ellipse, the preset UAV flight direction can be set to: take the center of the circle or the ellipse as the first working position of the UAV, and then add the longitude and latitude information of all corn tassels as the working positions of the UAV in a spiral manner to the inspection landing point queue in sequence.

[0055] In S400, based on the average plant height of the corn population and the tassel removal path of the drone, and in combination with the real-time position of the drone, the target corn tassel in the target area is positioned, and then the final cutting point of the target corn tassel is determined, including the following steps: S410: Based on the average plant height of the corn group, the emasculation path of the drone and the real-time position of the drone, control the drone to reach the target position and hover. The target position includes the initial hovering latitude and longitude of the drone and the initial hovering height. The initial hovering height of the drone is the initial flight height flyHeight, then flyHeight = the average plant height of the corn group + the drone adjustment coefficient. The drone adjustment coefficient is related to the configuration of the drone and is an empirical value obtained based on the drone model.

[0056] S420: Based on the initial hovering position of the UAV, that is, the target position, a first RGB image of a target corn tassel and a first depth image of a target corn tassel in a target area are acquired.

[0057] S430: Based on the first RGB image of the target corn tassel and in combination with the corn tassel detection model, the tassel rectangular frame of the target corn tassel is determined to complete the positioning of the target corn tassel.

[0058] S440: Based on the tassel rectangular frame of the target corn tassel and the first depth image of the target corn tassel, an initial cutting point of the target corn tassel is obtained, wherein the initial cutting point of the target corn tassel is located at a preset distance from the lower boundary of the tassel rectangular frame of the target corn tassel, and the preset distance is 10-30 mm.

[0059] S450: Based on a first straight-line distance between the target corn tassel and the drone imaging unit and a second straight-line distance between the drone imaging unit and the initial cutting point of the corn tassel, adjusting the hovering height and longitude and latitude of the drone, wherein the initial cutting point of the corn tassel is located at the center of the drone imaging unit, including the following steps: S451: Based on the first RGB image of the target corn tassel and the corn tassel detection model, the center point of the target corn tassel is obtained, and combined with the first depth image of the target corn tassel, the first straight-line distance is obtained. The first straight-line distance is the straight-line distance between the center point of the target corn tassel and the drone imaging unit.

[0060] S452: Based on the UAV imaging unit and the target corn tassel initial cutting point, a second straight-line distance is obtained; the second straight-line distance is the straight-line distance between the UAV imaging unit and the corn tassel cutting point, and the target corn tassel initial cutting point is located at the center position of the UAV imaging unit.

[0061] S453: Based on the first straight-line distance and the second straight-line distance, the UAV is controlled to move in the vertical direction by the first adjustment distance to complete the adjustment of the hovering height of the UAV, and the UAV is controlled to move in the horizontal direction by the second adjustment distance to complete the adjustment of the longitude and latitude of the UAV, wherein: First adjustment distance = first straight line distance - second straight line distance * sinθ; The second adjustment distance = the second straight line distance * cosθ; Hovering height = initial hovering height - second adjustment distance; Initial hovering height = average plant height of corn group + drone adjustment coefficient; Among them, θ is the angle between the first straight line and the horizontal plane. When θ is lower than 90°, the adjustment coefficient of the UAV is related to the configuration of the UAV.

[0062] like Figure 5 As shown, it is a schematic diagram of each distance in the process of adjusting the posture of the drone, and the relationship between each distance can be seen. In this embodiment, if the first adjustment distance is positive, the drone adjusts the distance vertically downward, and if the first adjustment distance is negative, the drone adjusts the distance vertically upward. It should be noted that the calculation method for the first adjustment distance in this embodiment is a fuzzy calculation, and there will be a certain deviation from the actual value that needs to be adjusted, but this deviation can be included in the error range of the drone. Therefore, in order to save computing resources, this embodiment adopts this rough calculation method. Of course, the actual value of the first adjustment distance can also be obtained by precise calculation, and a calculation model of the first adjustment distance that saves computing resources and improves accuracy can also be obtained through an artificial intelligence model. The present invention does not limit the specific method for obtaining the first adjustment distance, as long as the corn tassel can be accurately and efficiently removed in the end.

[0063] S460: Obtaining a second RGB image of a target corn tassel and a second depth image of a target corn tassel, and combining them with a corn tassel detection model to obtain a final cutting point of the target corn tassel, including the following steps: S461: After the drone adjusts the hovering altitude and longitude and latitude, a second RGB image of the target corn tassel and a second depth image of the target corn tassel are obtained.

[0064] S462: Based on the second RGB image of the target corn tassel and the corn tassel detection model, a final rectangular frame of the target corn tassel is obtained.

[0065] S463: Based on the target corn tassel final rectangular frame and the target corn tassel second depth image, the target corn tassel final cutting point is obtained, wherein the target corn tassel final cutting point is located at a preset distance from the lower boundary of the target corn tassel final rectangular frame, and the preset distance is 10-30 mm.

[0066] In this embodiment, since the initial hovering height of the drone is based only on the average plant height of the corn group, the actual height of the target corn tassel is not taken into account. Therefore, after reaching the initial hovering position, the drone will make corresponding adjustments to the hovering height and longitude and latitude according to the detected target corn tassel position. Such adjustments are substantial. After such substantial adjustments, a large amount of emasculation time can be saved and emasculation accuracy can be provided, thereby efficiently completing the emasculation task. This embodiment also saves time and improves accuracy for adjustments in the case of later occlusion, thereby further saving time and improving accuracy for the emasculation method of the present invention.

[0067] In S500, if the final cutting point of the target corn tassel is blocked, the real-time posture of the drone is adjusted until the preset conditions are met, and then the drone is controlled to remove the target corn tassel according to the final cutting point, including the following steps: S510: Control the UAV to rotate around the central axis according to a preset movement angle and a preset movement radius to adjust the real-time position and posture of the UAV, so as to quickly find the final cutting point of the target corn tassel that is blocked, wherein the central axis of rotation is the central axis perpendicular to the ground where the final cutting point of the target corn tassel is located, the preset movement radius of rotation is the horizontal distance between the UAV imaging unit and the final cutting point of the target corn tassel, and the preset movement angle ranges from 15° to 45°.

[0068] S520: Obtain the third RGB image of the target corn tassel and the third depth image of the target corn tassel, and obtain the middle cutting point of the target corn tassel by combining the corn tassel detection model. Use the middle cutting point of the target corn tassel to iterate the final cutting point of the target corn tassel. The specific method for obtaining the middle cutting point of the target corn tassel is the same as the method for obtaining the initial cutting point of the target corn tassel and the final cutting point of the target corn tassel, which will not be repeated here.

[0069] S530: Determine whether the final cutting point of the target corn tassel is blocked. If blocked, enter S510 to adjust the real-time position and posture of the drone again; if not blocked, enter S540.

[0070] S540: Figure 6 As shown, this is a positioning effect diagram of the final cutting point of the target corn tassel in this embodiment. At this time, the final cutting point of the target corn tassel is not blocked, and the drone is controlled to remove the target corn tassel according to the final cutting point.

[0071] In this embodiment, when the drone is controlled to remove the target corn tassel according to the final cutting point, the emasculating unit of the drone can be controlled to extend according to a preset extension distance, so that the distance between the drone emasculating unit and the final cutting point of the target corn tassel can be reduced, thereby improving the emasculating efficiency, the preset extension distance does not exceed the linear distance between the drone imaging unit and the final cutting point of the target corn tassel, and the preset extension distance does not exceed the maximum linear distance that the motion control component of the drone emasculating unit can extend. If emasculation is based on the removal of the target corn tassel by the drone laser, the laser motion control component of the drone is controlled to extend, wherein the drone laser is located on the drone fuselage.

[0072] Embodiment 2: In this embodiment, a DJI Mavic Inspire3 model drone is used. The RGB-D imaging unit carried by the drone is a standard device of the DJI Mavic Inspire3 model drone. The automatic alignment of the imaging information and the positioning device can automatically obtain the longitude and latitude information of each position in the image. A laser is used to remove the corn tassel. The laser is located on the drone body. The working distance of the laser transmitter is 500mm and the power is 150W.

[0073] The steps of image preprocessing are as follows: Step 1: Determine the network input size W*W (pixel*pixel); Step 2: Input GB image inputImage, the image width is width, the height is height, and the longest side of the input image maxLength=max(width,height); Step 3: Calculate the image transformation ratio ratio = W / maxLength; Step 4: Image transformation. The transformation process is as follows: if ratio=1: outputImage=inputImage else if ratio<1: resize(inputImage,outputImage,Size(width*ratio,height*ratio) else: padding(inputImage,outputImage) The padding is used to fill the small image, so as to enlarge the original image while ensuring that the resolution of the original image remains unchanged. The padding algorithm is as follows: enter: inputImage,ratio Algorithm flow: Create an image maskImage with a size of W*W and pixel values ​​of all channels are 0; Traverse inputImage and fill inputImage into the center area of ​​maskImage for i=0 to width for j=0 to height maskImage(i+(W-width) / 2,j+(W-height) / 2)=inputImage(i,j) Output: maskImage outputImage is maskImage.

[0074] The flight path planning for UAV emasculation is as follows: (1) Based on the corn image, the direction of the corn row is east-west, the direction of the row is north-south, and the ridge lines of each corn row are identified. The first working position of the drone is the southeasternmost corn tassel; (2) From south to north, the rows are arranged in an "M" pattern, with the southernmost row and the easternmost tassel as the starting point, that is, Figure 4 The middle line "1" is the starting point, and the flight direction is due east; (3) Determine whether there are other tassels in the same row. If so, add them to the inspection point queue in sequence until the row of tassels is fully covered. If there are no other tassels in the row, determine the nearest tassel row to the north and obtain the westernmost tassel position of the row. After traversing the first row of tassels, plan the drone to enter the westernmost tassel point of the second row, that is, flight route "2". (4) Each time a new row is entered, the UAV will start working from the westernmost position and eventually obtain the key points of the UAV's emasculation flight path.

[0075] Taking the emasculation of one of the corn tassels as an example, when the drone is emasculated, according to the emasculation path, the drone first hovers at the longitude and latitude of the corn plant to be emasculated, the height from the ground is = the average plant height of the corn group + 200mm, the camera optical axis of the drone is 60 ° angle with the ground and is facing the due east direction, and 200mm here is the drone adjustment coefficient, which is determined according to the specific configuration of the drone in the present embodiment. Then the drone collects the RGB picture and depth image of the target corn tassel, and the initial cutting point of the target corn tassel is 20mm from the lower boundary of the tassel rectangular frame. Then the drone is controlled to adjust its hovering height and longitude and latitude, and the RGB image and depth image of the target corn tassel are collected again to obtain the final cutting point of the target corn tassel. If the final cutting point of the target corn tassel is blocked, the UAV will move clockwise with the central axis of the target corn tassel final cutting point perpendicular to the ground as the rotation axis and the horizontal distance between the UAV's imaging unit and the target corn tassel final cutting point as the radius. The angle interval of each movement is 30°. After the UAV moves clockwise for 30°, it will collect images again and obtain the middle cutting point of the target corn tassel to determine whether the target corn tassel cutting point is blocked. If blocked, it will continue to move clockwise until the cutting point of the target corn tassel can be observed normally. During this process, it is also necessary to ensure that the drone automatically rotates around the tassel cutting point until the corn tassel cutting point is 500mm away from the optical center of the drone imaging unit and is at the imaging center of the drone. The corn tassel cutting point and the image center are in four directions: up, down, left, and right. The priority is judged and adjusted in sequence according to up, down, left, and right. When there is a corresponding deviation from the corresponding direction, the drone automatically fine-tunes in the corresponding direction until the tassel cutting point is located in the center of the image. For example, if the tassel cutting point is in the upper right corner of the image, the drone fine-tunes to the upper right corner. When the tassel cutting point coincides with the image center, if the distance between the drone and the tassel cutting point is greater than 500mm, the drone moves distance-500mm along the optical axis, otherwise it moves 500-distance in the opposite direction along the optical axis. Finally, if Figure 6 As shown, the cutting point of the corn tassel is not blocked and can be observed. The laser motion control component that controls the drone extends 500 mm and emits laser to remove the corn tassel.

[0076] Embodiment 3: A system for automatically removing tassels from unmanned aerial vehicles based on machine vision comprises an image acquisition and transformation module, a latitude, longitude and altitude calculation module, a tassel removal path calculation module, a tassel cutting point calculation module and a posture adjustment and tassel removal module; wherein: An image acquisition and transformation module acquires the latitude and longitude information of the target area, a corn RGB image and a corn depth image, and pre-processes the corn RGB image to obtain a transformed corn RGB image; The latitude and longitude and height calculation module obtains the latitude and longitude information of each corn tassel and the average plant height of the corn population based on the transformed corn RGB image, the corn depth image and the latitude and longitude information of the target area in combination with the corn tassel detection model; A tassel removal path calculation module is used to obtain the tassel removal path of the drone based on the preset drone flight direction and the longitude and latitude information of each corn tassel; The tassel cutting point calculation module locates the target corn tassel in the target area based on the average plant height of the corn population and the tassel removal path of the drone, and combines the real-time position of the drone to determine the final cutting point of the target corn tassel; The posture adjustment and tassel removal module adjusts the real-time posture of the drone if the final cutting point of the target corn tassel is blocked until the preset conditions are met, and then controls the drone to remove the target corn tassel according to the final cutting point.

[0077] As an implementable method, the latitude, longitude and altitude calculation module is set to: Based on the transformed corn RGB image and the corn tassel detection model, the corn tassel rectangular frame is obtained; Based on the rectangular frame of the corn tassel, the center position information of the corn tassel is obtained; Based on the center position information of the corn tassel and the longitude and latitude information of the target area, the longitude and latitude information of each corn tassel is obtained; Based on the center position information of the corn tassel, the average center coordinates of the corn tassel are obtained; The average plant height of the corn population was obtained based on the average center coordinates of the corn tassel and the corn depth image.

[0078] As an implementable method, the tassel cutting point calculation module is configured as follows: Based on the average plant height of the corn population, the tassel removal path of the drone and the real-time position of the drone, the drone is controlled to reach the target position, wherein the target position includes the initial hovering latitude and longitude and the initial hovering height of the drone; Acquire a first RGB image of a target corn tassel and a first depth image of a target corn tassel in a target area; Based on the first RGB image of the target corn tassel, in combination with the corn tassel detection model, a tassel rectangular frame of the target corn tassel is determined to complete the positioning of the target corn tassel; Based on the tassel rectangular frame of the target corn tassel and the first depth image of the target corn tassel, an initial cutting point of the target corn tassel is obtained; Adjust the hovering height and longitude and latitude of the drone based on a first straight-line distance between the target corn tassel and the drone imaging unit and a second straight-line distance between the drone imaging unit and the target corn tassel initial cutting point, wherein the target corn tassel initial cutting point is located at the center of the drone imaging unit; A second RGB image of the target corn tassel and a second depth image of the target corn tassel are obtained, and a final cutting point of the target corn tassel is obtained by combining the corn tassel detection model.

[0079] As an implementable method, the posture adjustment in the posture adjustment and detasseling module is set to: Control the drone to rotate around the central axis according to the preset movement angle and preset movement radius to adjust the real-time posture of the drone, where the central axis is the central axis perpendicular to the ground where the final cutting point of the target corn tassel is located, and the preset movement radius is the horizontal distance between the drone imaging unit and the final cutting point of the target corn tassel. Various changes and modifications can be made without departing from the spirit and scope of the present invention, and all equivalent technical solutions also belong to the scope of the present invention.

[0080] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0081] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0082] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0083] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0084] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0085] It should be noted that: The "one embodiment" or "embodiment" mentioned in the specification means that the specific features, structures or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment" or "embodiment" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0086] In addition, it should be noted that the shapes and names of the parts and components of the specific embodiments described in this specification may be different. Any equivalent or simple changes made based on the structure, features and principles described in the patent concept of the present invention are included in the protection scope of the patent of the present invention. The technicians in the technical field of the present invention can make various modifications or supplements to the specific embodiments described or replace them in a similar manner, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A method for automatically removing male tassels using a drone based on machine vision, characterized in that: The following steps are involved: Acquire the latitude and longitude information of the target area, the corn RGB image and the corn depth image, pre-process the corn RGB image, and obtain a transformed corn RGB image; Based on the transformed corn RGB image, the corn depth image and the longitude and latitude information of the target area, combined with the corn tassel detection model, the longitude and latitude information of each corn tassel and the average plant height of the corn population are obtained; Based on the preset UAV flight direction and the latitude and longitude information of each corn tassel, a UAV detasseling path is obtained; Based on the average plant height of the corn population and the tassel removal path of the drone, the target corn tassel in the target area is located in combination with the real-time position of the drone, and then the final cutting point of the target corn tassel is determined; If the final cutting point of the target corn tassel is blocked, the real-time position and posture of the drone is adjusted until the preset conditions are met, and then the drone is controlled to remove the target corn tassel according to the final cutting point.

2. The method for automatically removing tassels by a drone based on machine vision according to claim 1, characterized in that: The method of obtaining the longitude and latitude information of each corn tassel and the average plant height of the corn population based on the transformed corn RGB image, the corn depth image and the longitude and latitude information of the target area in combination with the corn tassel detection model comprises the following steps: Based on the transformed corn RGB image and the corn tassel detection model, the corn tassel rectangular frame is obtained; Based on the rectangular frame of the corn tassel, the center position information of the corn tassel is obtained; Based on the center position information of the corn tassel and the longitude and latitude information of the target area, the longitude and latitude information of each corn tassel is obtained; Based on the center position information of the corn tassel, the average center coordinates of the corn tassel are obtained; The average plant height of the corn population was obtained based on the average center coordinates of the corn tassel and the corn depth image.

3. The method for automatically removing tassels by drone based on machine vision according to claim 1, characterized in that: The method of obtaining a drone emasculation path based on the preset drone flight direction and the latitude and longitude information of each corn tassel comprises the following steps: Based on the preset UAV flight direction, the latitude and longitude information of each corn tassel is sorted to obtain an inspection point queue; Based on the order of the longitude and latitude information of each corn tassel in the inspection landing point queue, the drone detasseling path is obtained.

4. The method for automatically removing tassels by a drone based on machine vision according to claim 1, characterized in that: The method locates the target corn tassel in the target area based on the average plant height of the corn population and the tassel removal path of the drone, and combines the real-time position of the drone to further determine the final cutting point of the target corn tassel, including the following steps: Based on the average plant height of the corn population, the tassel removal path of the drone and the real-time position of the drone, the drone is controlled to reach the target position, wherein the target position includes the initial hovering latitude and longitude and the initial hovering height of the drone; Acquire a first RGB image of a target corn tassel and a first depth image of a target corn tassel in a target area; Based on the first RGB image of the target corn tassel, in combination with the corn tassel detection model, a tassel rectangular frame of the target corn tassel is determined to complete the positioning of the target corn tassel; Based on the tassel rectangular frame of the target corn tassel and the first depth image of the target corn tassel, an initial cutting point of the target corn tassel is obtained; Adjust the hovering height and longitude and latitude of the drone based on a first straight-line distance between the target corn tassel and the drone imaging unit and a second straight-line distance between the drone imaging unit and the target corn tassel initial cutting point, wherein the target corn tassel initial cutting point is located at the center of the drone imaging unit; A second RGB image of the target corn tassel and a second depth image of the target corn tassel are obtained, and a final cutting point of the target corn tassel is obtained by combining the corn tassel detection model.

5. The method for automatically removing tassels by drone based on machine vision according to claim 1, characterized in that: If the final cutting point of the target corn tassel is blocked, the real-time posture of the drone is adjusted, specifically: The drone is controlled to rotate around the central axis at a preset movement angle and a preset movement radius to adjust the real-time posture of the drone, wherein the central axis is the central axis perpendicular to the ground where the final cutting point of the target corn tassel is located, and the preset movement radius is the horizontal distance between the drone imaging unit and the final cutting point of the target corn tassel.

6. The method for automatically removing tassels by a drone based on machine vision according to claim 4, characterized in that: The method of adjusting the hovering height and longitude and latitude of the drone based on a first straight-line distance between the target corn tassel and the drone imaging unit and a second straight-line distance between the drone imaging unit and the initial cutting point of the corn tassel comprises the following steps: Controlling the UAV to move a first adjustment distance in a vertical direction to complete the adjustment of the hovering height of the UAV; Control the UAV to move in the horizontal direction for a second adjustment distance to complete the adjustment of the longitude and latitude of the UAV; Wherein, the first adjustment distance = the first straight line distance - the second straight line distance * sinθ; The second adjustment distance = the second straight line distance * cosθ; Hovering height = initial hovering height - second adjustment distance; Initial hovering height = average plant height of corn group + drone adjustment coefficient; The θ is the angle between the first straight line and the horizontal plane, and the UAV adjustment coefficient is related to the configuration of the UAV.

7. The method for automatically removing tassels by drone based on machine vision according to claim 1, characterized in that: The preprocessing of the corn RGB image to obtain the transformed corn RGB image comprises the following steps: Acquire width data and height data of the corn RGB image, and determine the longest side of the corn RGB image based on the width data and the height data; Based on the longest side and the preset image size, the image transformation rate is obtained; Performing image transformation based on the image transformation rate to obtain a transformed corn RGB image, specifically: when the image transformation rate=1, keeping the corn RGB image unchanged; when the image transformation rate>1, proportionally reducing the width data and height data of the corn RGB image based on the image transformation rate; when the image transformation rate<1, keeping the resolution unchanged, and enlarging the corn RGB image; The image transformation rate is expressed as follows: R=maxLength / W; Among them, R represents the image transformation rate, maxLength represents the longest side, and W represents the side length of the preset image size.

8. The method for automatically removing tassels by drone based on machine vision according to claim 7, characterized in that: The enlarging process of the corn RGB image comprises the following steps: Creating a new image with the same size as the preset image; Based on the size side length of the new image, the width data and the height data of the corn RGB image, a horizontal offset and a vertical offset are obtained, and the corn RGB image is located at the center of the new image; Each pixel of the corn RGB image is copied to the new image according to the horizontal offset and the vertical offset.

9. The method for automatically removing tassels by drone based on machine vision according to claim 1, characterized in that: The target corn tassel is removed by controlling a laser set on a drone, and the laser of the drone is controlled to extend according to a preset extension distance, wherein the drone laser is located on the drone body.

10. The method for automatically removing tassels by drone based on machine vision according to claim 1, characterized in that: The corn RGB image is collected by an imaging unit, and the target corn tassel is removed by controlling a laser set on a drone, wherein the optical axis of the imaging unit is parallel to the laser optical axis emitted by the laser, and the line connecting the optical center of the imaging unit and the emission position of the laser optical axis is perpendicular to both the optical axis of the imaging unit and the laser optical axis.

11. A system for automatically removing male tassels by drone based on machine vision, characterized in that: It includes an image acquisition and transformation module, a latitude, longitude and altitude calculation module, a tassel removal path calculation module, a tassel cutting point calculation module and a posture adjustment and tassel removal module; The image acquisition and transformation module acquires the latitude and longitude information of the target area, the corn RGB image and the corn depth image, and pre-processes the corn RGB image to obtain a transformed corn RGB image; The latitude, longitude and height calculation module obtains the latitude and longitude information of each corn tassel and the average plant height of the corn population based on the transformed corn RGB image, the corn depth image and the latitude and longitude information of the target area in combination with the corn tassel detection model; The emasculation path calculation module obtains the emasculation path of the drone based on the preset drone flight direction and the longitude and latitude information of each corn tassel; The tassel cutting point calculation module locates the target corn tassel in the target area based on the average plant height of the corn population and the tassel removal path of the drone, and combines the real-time position of the drone to determine the final cutting point of the target corn tassel; The posture adjustment and tassel removal module adjusts the real-time posture of the drone if the final cutting point of the target corn tassel is blocked until the preset conditions are met, and then controls the drone to remove the target corn tassel according to the final cutting point.

12. A device for automatically removing male tassels by a drone based on machine vision, comprising a memory, a processor, a computer program stored in the memory and running on the processor, and an SDK running on the drone processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 10 is implemented.

Images