Unmanned aerial vehicle group control crop density adaptive patrol method based on fractal geometric analysis

Abstract

The invention discloses an unmanned aerial vehicle group control crop density adaptive patrol method based on fractal geometric analysis. The method comprises the following steps: S1, collecting ground crop density distribution information image data; s2, performing preliminary classification on the density region according to the fractal dimension information; s3, the high-density area, the medium-density area and the low-density area are processed; s4, generating a multi-scale distribution information graph of crop density; s5, the unmanned aerial vehicle group control system distributes patrol tasks and flight paths of all the unmanned aerial vehicles according to density data of the multi-scale distribution information graph, and a self-adaptive patrol system adjusts flight parameters according to the distributed tasks; s6, in the patrol process of the unmanned aerial vehicle, executing the steps S1 to S5 circularly; s7, performing multi-scale fractal geometric analysis on the image data of each region after patrol is completed; and S8, in the multilevel crop density analysis report, identifying and marking an abnormal density region. According to the invention, the patrol efficiency, the data precision and the system collaboration are obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles, and in particular to an unmanned aerial vehicle group-controlled crop density adaptive inspection method based on fractal geometry analysis. Background Art

[0002] In the existing technology, drone inspections have been widely used in the agricultural field to monitor crop growth, assess crop density, and detect abnormal problems. However, most drone inspection technologies rely on fixed routes and static image analysis, that is, collecting data through preset flight paths and collection points. Although the existing technology can achieve a certain degree of inspection coverage in farmland areas with uniform density, it has obvious shortcomings in farmland with uneven crop density. Fixed routes are difficult to adjust the inspection path according to the actual density distribution, resulting in insufficient inspections in high-density areas and repeated collection in low-density areas, resulting in low inspection efficiency. In addition, the static image analysis method lacks real-time adjustment capabilities when facing dynamic changes in crop density, and cannot provide timely and accurate data support.

[0003] Another technical attempt is to use UAV dynamic patrol based on image analysis and simple adaptive algorithms. By combining real-time collected image data, the flight altitude and speed of the UAV are appropriately adjusted. This type of method is often based on local data and lacks in-depth analysis of the global distribution of crop density, resulting in insufficient adjustment accuracy. In scenarios with complex farmland terrain or highly differentiated crop density, traditional dynamic patrol technology is difficult to capture detailed changes in high-density areas, and it is also difficult to quickly cover low-density areas, and the patrol effect is still limited. In addition, when dealing with the multi-scale characteristics of crop density, existing technologies lack effective analysis tools and cannot fully reflect the fine-grained information of crop density distribution.

[0004] For crop density analysis, the commonly used method in the prior art is density assessment based on statistics or simple image processing. The density of the local area is calculated by the threshold segmentation method of the pixel gray value. Although it can preliminarily reflect the overall density distribution of the crop, it lacks the ability to effectively describe the complexity of density changes, fine-grained characteristics and multi-scale distribution. When high-density and low-density areas coexist, the evaluation results are easily distorted. In addition, the density analysis algorithm in the prior art is usually difficult to form a coordinated optimization with the drone patrol system, resulting in the separation of data collection and analysis processes, further affecting the patrol efficiency and the actual utilization value of the data. Summary of the invention

[0005] One object of the present invention is to propose a UAV group-controlled crop density adaptive inspection method based on fractal geometry analysis, which significantly improves the inspection efficiency, data accuracy and system coordination.

[0006] According to an embodiment of the present invention, a method for adaptively inspecting crop density by controlling a group of drones based on fractal geometry analysis comprises the following steps:

[0007] S1. Use multiple drones in the drone group to carry out initial inspections of the preset farmland area, respectively equipped with image acquisition sensors, to collect image data of ground crop density distribution information, and dynamically adjust the inspection frequency according to the crop growth cycle. In the late stage of crop growth, the inspection frequency is increased according to the crop growth rate;

[0008] S2. The ground data processing system preliminarily processes the image data collected by each drone based on the fractal geometry analysis model, uses the box counting method to perform fractal geometry analysis on the global distribution of crop density to generate the fractal dimension of density distribution, and preliminarily classifies the density area based on the fractal dimension information;

[0009] S3. For high-density areas, a fine-grained fractal geometry analysis model is applied, a smaller fractal box size is set in the fractal geometry box counting method, the minute density changes of crop groups in high-density areas are captured, crop density distribution information in high-density areas is obtained, and fine-grained fractal density distribution data is generated;

[0010] For medium-density areas, a medium-sized fractal box is set to perform fractal geometry analysis to extract the density distribution characteristics of the medium-density area;

[0011] For low-density areas, a larger fractal box is set to simplify fractal geometry analysis and obtain simplified fractal density distribution data of low-density areas;

[0012] S4. The ground data processing system integrates the fine-grained fractal density distribution data of high-density areas, the fractal density distribution data of medium-density areas, and the simplified fractal density distribution data of low-density areas collected by multiple drones to generate a multi-scale distribution information map of crop density, and distributes the multi-scale distribution information map to the adaptive inspection system of each drone;

[0013] S5. The drone group control system assigns the inspection tasks and flight trajectories of each drone according to the density data of the multi-scale distribution information map, and the adaptive inspection system adjusts the flight parameters according to the assigned tasks. In high-density areas, the drone is adjusted to lower the flight altitude, slow down the flight speed and reduce the shooting angle; in medium-density areas, the drone is maintained at a medium altitude and speed for inspection; in low-density areas, the flight altitude is increased, the flight speed is increased and the shooting angle is expanded;

[0014] S6. During the drone inspection process, S1 to S5 are executed cyclically, so that the drone group continuously updates and adjusts the flight parameters and task allocation according to the real-time collected image data and fractal geometry analysis results, dynamically responds to the distribution changes of crop density in the farmland area, and generates and updates the multi-scale distribution information map of crop density in real time;

[0015] S7. After the inspection is completed, the image data of the entire area collected by the drone group is transmitted to the ground data processing system, and the image data of each area is subjected to multi-scale fractal geometry analysis to generate a multi-level crop density analysis report including crop-dense areas, sparse areas and their fine-grained distribution characteristics. The multi-level crop density analysis report provides global to fine-grained crop density information based on the fractal geometry analysis results of different density levels;

[0016] S8. In the multi-level crop density analysis report, identify and mark abnormal density areas, including areas with abnormally high density and areas with abnormally sparse density. The system automatically generates corresponding abnormal warning information and sends it to the agricultural management end. At the same time, the abnormal areas are prioritized as the inspection focus of the drone group.

[0017] Optionally, the S1 specifically includes the following sub-steps:

[0018] S11, using multiple drones in the drone group to carry out initial inspections of the preset farmland area, respectively equipped with image acquisition sensors, to obtain image data of ground crop density distribution information;

[0019] S12, setting the initial flight parameters of the UAV, the initial flight parameters including the flight altitude H0, the flight speed V0 and the shooting interval I0;

[0020] S13. Dynamically adjust the drone inspection frequency F(t) according to the crop growth cycle and real-time crop density distribution information collected in the farmland area:

[0021]

[0022] Among them, F min is the lowest frequency of inspection, used for image acquisition in the early stage of crop growth, F max is the highest inspection frequency, which is used for intensive collection in the late stage of crop growth. k is the growth rate adjustment coefficient, which is used to control the variation range of inspection frequency. t is the current crop growth time. c It is the turning point time of crop growth rate, which is used to indicate that crops have entered a rapid growth period;

[0023] S14. In the late stage of crop growth, according to the high-density area ratio P in the real-time density distribution information h Dynamically adjust the inspection frequency. If P h >P th, then the inspection frequency F(t) increases to the maximum value F max , otherwise, keep the current frequency unchanged, where P th It is the preset high-density area ratio threshold, which is used to determine whether the crops have entered a dense distribution state;

[0024] S15. The UAV performs continuous image acquisition of crop density distribution information according to the adjusted patrol frequency F(t), and sends the acquired image data to the ground data processing system in real-time transmission.

[0025] Optionally, S2 specifically includes the following sub-steps:

[0026] S21, receiving ground crop density distribution information image data collected by the drone group, inputting the image data into a ground data processing system, and constructing an initial image data set for crop density distribution analysis;

[0027] S22, preprocessing the initial image data set, including grayscale conversion, noise removal, geometric correction and orthographic projection transformation of the image, to generate standard image data suitable for fractal geometry analysis, wherein the standard image data is represented in the form of a two-dimensional pixel matrix, and the grayscale value G(x, y) of each pixel reflects the density of the corresponding ground crop;

[0028] S23, based on the improved multi-scale box counting method, the standard image data is subjected to fractal geometry analysis, and the initial box size r0 and scale factor s are set, and the box size sequence is r i =r0 / s i (i=0,1,2,…,n), for each scale r i Divide the standard image data into r i × i Box B k (r i ), for each box B k (r i ) Calculate the cumulative sum S of the grayscale values ​​inside the box k (r i ):

[0029]

[0030] Calculation scale r i The total gray value S under total (r i ):

[0031]

[0032] Among them, N(r i ) is the scale r i The total number of boxes under;

[0033] Calculate the scaling function F(q,r i ):

[0034]

[0035] Among them, q is the multifractal order parameter, which is used to capture the multi-scale characteristics of crop density distribution;

[0036] S24. Calculate the generalized scaling exponent τ(q) and the generalized fractal dimension D for different values ​​of q q :

[0037]

[0038] Generalized fractal dimension D q Reflects the complexity of crop density distribution at different scales and in different density areas;

[0039] S25. According to the generalized fractal dimension D q Construct a multifractal spectrum and calculate the characteristic parameters of the fractal spectrum, including the spectrum width ΔD and the spectrum peak position:

[0040]

[0041] Among them, D qmin and D qmax are the generalized fractal dimensions corresponding to the selected minimum and maximum q values ​​respectively;

[0042] Crop density areas are classified according to the fractal spectrum width ΔD:

[0043] When the fractal spectrum width ΔD>ΔD high When , the area is determined to be a high-density area;

[0044] When the fractal spectrum width ΔD low ≤ΔD≤ΔD high When , the area is determined to be a medium-density area;

[0045] When the fractal spectrum width ΔD<ΔD low When , the area is determined to be a low-density area;

[0046] Where, ΔD high and ΔD low It is a preset fractal spectrum width threshold, which is used to distinguish crop areas with different density levels.

[0047] Optionally, the S4 specifically includes the following sub-steps:

[0048] S41, receiving the fine-grained fractal density distribution data of high-density areas, the fractal density distribution data of medium-density areas, and the simplified fractal density distribution data of low-density areas collected by multiple drones, fusing the data of each drone based on a unified geographic coordinate system, and constructing an initial set of crop density data from multiple drones

[0049] S42, for the initial set Perform spatial registration processing to achieve spatial alignment of each drone data by minimizing the registration error function ∈(x,y). The error function is defined as:

[0050]

[0051] Among them, M and N represent the number of UAVs involved in data fusion and the number of pixels of each UAV respectively; G i (x,y) and G j (x, y) are the density distribution values ​​of the i-th and j-th drones at the geographic coordinates (x, y), and after optimization, the fused density distribution data D is obtained. f (x,y);

[0052] S43, according to the fused density distribution data D f (x, y) Multi-scale feature extraction is performed on the data of high-density areas, medium-density areas, and low-density areas to generate a fine-grained distribution feature matrix M for high-density areas h , medium density region feature matrix M m and the low-density region feature matrix M l , the feature extraction formula is as follows:

[0053]

[0054] where z∈{h,m,l} represents high-density, medium-density, and low-density areas; G z (x, y) is the density distribution value of region z; R is the scale range of the fractal box size, which is used for multi-scale feature extraction; [x1, x2] and [y1, y2] are the coordinate ranges of a specific region;

[0055] S44. Generate a multi-scale distribution information map of crop density using the fine-grained distribution feature matrix of high-density areas, the feature matrix of medium-density areas, and the feature matrix of low-density areas

[0056]

[0057] Among them, α z (r) is the regional weight function, which is used to adjust the weights of different density regions in the multi-scale distribution information map;

[0058] S45. Generate a multi-scale distribution information map The adaptive patrol system assigned to the drones adjusts the flight parameters and patrol tasks of each drone according to the characteristics and weight distribution of high, medium and low density areas in the information map, so that the drones can dynamically respond to changes in the density of farmland crops.

[0059] Optionally, S5 specifically includes the following sub-steps:

[0060] S51. The UAV group control system receives the multi-scale distribution information map of crop density generated by the ground data processing system. The density data in the multi-scale distribution information map is classified into high-density areas, medium-density areas and low-density areas, and each classification result is used as the input of the UAV task allocation;

[0061] S52. According to the classification results of crop density distribution and the flight capability parameters of each UAV, inspection tasks are assigned. The task assignment follows the following optimization objectives:

[0062]

[0063] Among them, Φ is the total inspection cost function, U is the number of drones, T i,z represents the patrol time of the i-th drone in area z, C i is the energy consumption of the i-th UAV, ω1 and ω2 are the weight coefficients of time and energy consumption;

[0064] S53, generating optimized flight trajectories for each drone. The flight trajectories in high-density areas are mainly based on covering key areas. When planning flight trajectories, the drone's residence time and route overlap rate in high-density areas are increased. The flight trajectories in medium-density areas use a gridding method to balance coverage and collection efficiency. The flight trajectories in low-density areas use a sparse coverage mode to maximize the coverage of the drone's flight path.

[0065] S54. The adaptive inspection system dynamically adjusts the flight parameters of the drone, including flight altitude, flight speed and shooting angle, according to the task allocation results and flight trajectory. In high-density areas, the drone is adjusted to a lower flight altitude and slows down the flight speed, while narrowing the shooting angle to obtain detailed data. In medium-density areas, the drone maintains a medium flight altitude and speed, taking into account both collection accuracy and efficiency. In low-density areas, the drone is adjusted to a higher flight altitude, increases the flight speed and expands the shooting angle.

[0066] S55. The UAV performs inspection tasks according to the optimized task allocation and flight trajectory, and responds to changes in crop density distribution information in real time during the inspection process.

[0067] The beneficial effects of the present invention are:

[0068] (1) The present invention introduces a multi-scale fractal geometry analysis model into the processing of UAV image data, and uses an improved box counting method and multi-fractal spectrum analysis technology to capture the complex characteristics of crop density distribution from multiple scales and multiple levels. Unlike the traditional single density calculation method, the present invention can dynamically adjust the analysis accuracy based on different fractal box sizes, apply a fine-grained fractal geometry model to high-density areas to capture slight changes in crop density, and use a simplified model to reduce the calculation complexity in low-density areas. Multi-scale fractal geometry analysis makes up for the shortcomings of traditional methods in dealing with complex and irregular crop distribution, and significantly improves the accuracy and reliability of crop density distribution data.

[0069] (2) The present invention maps the distribution characteristics of high, medium and low density areas to geographic coordinates by generating a multi-scale distribution information map, and then dynamically allocates inspection tasks and generates optimized flight trajectories based on regional characteristics and UAV performance parameters. Compared with the traditional fixed-route or simple adaptively adjusted UAV inspection method, the present invention innovatively realizes the efficient coordination between the UAV group control system and the ground data processing system, so that the UAV prioritizes low-altitude inspections in high-density areas to obtain fine data, and quickly covers low-density areas to save time and resources, thereby realizing globally optimized inspection path planning. The inspection task allocation is adjusted through a multi-scale distribution information map and a weight function, which significantly improves the inspection coverage efficiency and the spatial uniformity of the inspection data.

[0070] (3) The present invention receives density distribution update data from the ground processing system in real time and dynamically adjusts flight parameters in combination with the UAV adaptive patrol system, so that the UAV can respond to changes in crop density distribution in real time during the patrol process. Differentiated parameter adjustment strategies are adopted for different density areas. Low-altitude, slow and precise patrols in high-density areas, grid patrols in medium-density areas, and high-altitude, rapid coverage in low-density areas are dynamically controlled to ensure the adaptability of patrol tasks to crop distribution characteristics. The real-time linkage between task allocation and flight parameter adjustment significantly reduces the repeated operations and blind spot coverage of UAVs, thereby improving energy efficiency and patrol quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0072] Figure 1 This is a flow chart of a method for adaptively inspecting crop density by group-controlled UAVs based on fractal geometry analysis proposed by the present invention. DETAILED DESCRIPTION

[0073] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0074] refer to Figure 1 , a crop density adaptive inspection method of drone group control based on fractal geometry analysis, comprising the following steps:

[0075] S1. Use multiple drones in the drone group to carry out initial inspections of the preset farmland area, respectively equipped with image acquisition sensors, to collect image data of ground crop density distribution information, and dynamically adjust the inspection frequency according to the crop growth cycle. In the late stage of crop growth, the inspection frequency is increased according to the crop growth rate;

[0076] S2. The ground data processing system preliminarily processes the image data collected by each drone based on the fractal geometry analysis model, uses the box counting method to perform fractal geometry analysis on the global distribution of crop density to generate the fractal dimension of density distribution, and preliminarily classifies the density area based on the fractal dimension information;

[0077] S3. For high-density areas, a fine-grained fractal geometry analysis model is applied, a smaller fractal box size is set in the fractal geometry box counting method, the minute density changes of crop groups in high-density areas are captured, crop density distribution information in high-density areas is obtained, and fine-grained fractal density distribution data is generated;

[0078] For medium-density areas, a medium-sized fractal box is set to perform fractal geometry analysis to extract the density distribution characteristics of the medium-density area;

[0079] For low-density areas, a larger fractal box is set to simplify fractal geometry analysis and obtain simplified fractal density distribution data of low-density areas;

[0080] S4. The ground data processing system integrates the fine-grained fractal density distribution data of high-density areas, the fractal density distribution data of medium-density areas, and the simplified fractal density distribution data of low-density areas collected by multiple drones to generate a multi-scale distribution information map of crop density, and distributes the multi-scale distribution information map to the adaptive inspection system of each drone;

[0081] S5. The drone group control system assigns the inspection tasks and flight trajectories of each drone according to the density data of the multi-scale distribution information map, and the adaptive inspection system adjusts the flight parameters according to the assigned tasks. In high-density areas, the drone is adjusted to lower the flight altitude, slow down the flight speed and reduce the shooting angle; in medium-density areas, the drone is maintained at a medium altitude and speed for inspection; in low-density areas, the flight altitude is increased, the flight speed is increased and the shooting angle is expanded;

[0082] S6. During the drone inspection process, S1 to S5 are executed cyclically, so that the drone group continuously updates and adjusts the flight parameters and task allocation according to the real-time collected image data and fractal geometry analysis results, dynamically responds to the distribution changes of crop density in the farmland area, and generates and updates the multi-scale distribution information map of crop density in real time;

[0083] S7. After the inspection is completed, the image data of the entire area collected by the drone group is transmitted to the ground data processing system, and the image data of each area is subjected to multi-scale fractal geometry analysis to generate a multi-level crop density analysis report including crop-dense areas, sparse areas and their fine-grained distribution characteristics. The multi-level crop density analysis report provides global to fine-grained crop density information based on the fractal geometry analysis results of different density levels;

[0084] S8. In the multi-level crop density analysis report, identify and mark abnormal density areas, including areas with abnormally high density and areas with abnormally sparse density. The system automatically generates corresponding abnormal warning information and sends it to the agricultural management end. At the same time, the abnormal areas are prioritized as the inspection focus of the drone group.

[0085] In this implementation, S1 specifically includes the following sub-steps:

[0086] S11, using multiple drones in the drone group to carry out initial inspections of the preset farmland area, respectively equipped with image acquisition sensors, to obtain image data of ground crop density distribution information;

[0087] S12, setting the initial flight parameters of the UAV, the initial flight parameters including the flight altitude H0, the flight speed V0 and the shooting interval I0;

[0088] S13. Dynamically adjust the drone inspection frequency F(t) according to the crop growth cycle and real-time crop density distribution information collected in the farmland area:

[0089]

[0090] Among them, F min is the lowest frequency of inspection, used for image acquisition in the early stage of crop growth, F max is the highest inspection frequency, which is used for intensive collection in the late stage of crop growth. k is the growth rate adjustment coefficient, which is used to control the variation range of inspection frequency. t is the current crop growth time. c It is the turning point time of crop growth rate, which is used to indicate that crops have entered a rapid growth period;

[0091] S14. In the late stage of crop growth, according to the high-density area ratio P in the real-time density distribution information h Dynamically adjust the inspection frequency. If P h >P th, then the inspection frequency F(t) increases to the maximum value F max , otherwise, keep the current frequency unchanged, where P th It is the preset high-density area ratio threshold, which is used to determine whether the crops have entered a dense distribution state;

[0092] S15. The UAV performs continuous image acquisition of crop density distribution information according to the adjusted patrol frequency F(t), and sends the acquired image data to the ground data processing system in real-time transmission.

[0093] In this implementation, S2 specifically includes the following sub-steps:

[0094] S21, receiving ground crop density distribution information image data collected by the drone group, inputting the image data into a ground data processing system, and constructing an initial image data set for crop density distribution analysis;

[0095] S22, preprocessing the initial image data set, including grayscale conversion, noise removal, geometric correction and orthographic projection transformation of the image, to generate standard image data suitable for fractal geometry analysis, the standard image data is represented in the form of a two-dimensional pixel matrix, and the grayscale value G(x, y) of each pixel reflects the density of the corresponding ground crop;

[0096] S23, based on the improved multi-scale box counting method, the standard image data is subjected to fractal geometry analysis, and the initial box size r0 and scale factor s are set, and the box size sequence is r i =r0 / s i (i=0,1,2,…,n), for each scale r i Divide the standard image data into r i × i Box B k (r i ), for each box B k (r i ) Calculate the cumulative sum S of the grayscale values ​​inside the box k (r i ):

[0097]

[0098] Calculation scale r i The total gray value S under total (r i ):

[0099]

[0100] Among them, N(r i ) is the scale r i The total number of boxes under;

[0101] Calculate the scaling function F(q,r i ):

[0102]

[0103] Among them, q is the multifractal order parameter, which is used to capture the multi-scale characteristics of crop density distribution;

[0104] S24. Calculate the generalized scaling exponent τ(q) and the generalized fractal dimension D for different values ​​of q q :

[0105]

[0106] Generalized fractal dimension D q Reflects the complexity of crop density distribution at different scales and in different density areas;

[0107] S25. According to the generalized fractal dimension D q Construct a multifractal spectrum and calculate the characteristic parameters of the fractal spectrum, including the spectrum width ΔD and the spectrum peak position:

[0108]

[0109] in, and are the generalized fractal dimensions corresponding to the selected minimum and maximum q values ​​respectively;

[0110] Crop density areas are classified according to the fractal spectrum width ΔD:

[0111] When the fractal spectrum width ΔD>ΔD high When , the area is determined to be a high-density area;

[0112] When the fractal spectrum width ΔD low ≤ΔD≤ΔD high When , the area is determined to be a medium-density area;

[0113] When the fractal spectrum width ΔD<ΔD low When , the area is determined to be a low-density area;

[0114] Where, ΔD high and ΔD low It is a preset fractal spectrum width threshold, which is used to distinguish crop areas with different density levels.

[0115] In this implementation, S4 specifically includes the following sub-steps:

[0116] S41, receiving the fine-grained fractal density distribution data of high-density areas, the fractal density distribution data of medium-density areas, and the simplified fractal density distribution data of low-density areas collected by multiple drones, fusing the data of each drone based on a unified geographic coordinate system, and constructing an initial set of crop density data from multiple drones

[0117] S42, for the initial set Perform spatial registration processing to achieve spatial alignment of each drone data by minimizing the registration error function ∈(x,y). The error function is defined as:

[0118]

[0119] Among them, M and N represent the number of UAVs involved in data fusion and the number of pixels of each UAV respectively; G i (x,y) and G j (x, y) are the density distribution values ​​of the i-th and j-th drones at the geographic coordinates (x, y), and after optimization, the fused density distribution data D is obtained. f (x,y);

[0120] S43, according to the fused density distribution data D f (x, y) Multi-scale feature extraction is performed on the data of high-density areas, medium-density areas, and low-density areas to generate a fine-grained distribution feature matrix M for high-density areas h , medium density region feature matrix M m and the low-density region feature matrix M l , the feature extraction formula is as follows:

[0121]

[0122] where z∈{h,m,l} represents high-density, medium-density, and low-density areas; G z (x, y) is the density distribution value of region z; R is the scale range of the fractal box size, which is used for multi-scale feature extraction; [x1, x2] and [y1, y2] are the coordinate ranges of a specific region;

[0123] S44. Generate a multi-scale distribution information map of crop density using the fine-grained distribution feature matrix of high-density areas, the feature matrix of medium-density areas, and the feature matrix of low-density areas

[0124]

[0125] Among them, α z (r) is the regional weight function, which is used to adjust the weights of different density regions in the multi-scale distribution information map;

[0126] S45. Generate a multi-scale distribution information map The adaptive patrol system assigned to the drones adjusts the flight parameters and patrol tasks of each drone according to the characteristics and weight distribution of high, medium and low density areas in the information map, so that the drones can dynamically respond to changes in the density of farmland crops.

[0127] In this implementation, S5 specifically includes the following sub-steps:

[0128] S51. The UAV group control system receives the multi-scale distribution information map of crop density generated by the ground data processing system. The density data in the multi-scale distribution information map is classified into high-density area, medium-density area and low-density area, and each classification result is used as the input of UAV task allocation;

[0129] S52. According to the classification results of crop density distribution and the flight capability parameters of each UAV, inspection tasks are assigned. The task assignment follows the following optimization objectives:

[0130]

[0131] Among them, Φ is the total inspection cost function, U is the number of drones, T i,z represents the patrol time of the i-th drone in area z, C i is the energy consumption of the i-th UAV, ω1 and ω2 are the weight coefficients of time and energy consumption;

[0132] S53, generating optimized flight trajectories for each drone. The flight trajectories in high-density areas are mainly based on covering key areas. When planning flight trajectories, the drone's residence time and route overlap rate in high-density areas are increased. The flight trajectories in medium-density areas use a gridding method to balance coverage and collection efficiency. The flight trajectories in low-density areas use a sparse coverage mode to maximize the coverage of the drone's flight path.

[0133] S54. The adaptive inspection system dynamically adjusts the flight parameters of the drone, including flight altitude, flight speed and shooting angle, according to the task allocation results and flight trajectory. In high-density areas, the drone is adjusted to a lower flight altitude and slows down the flight speed, while narrowing the shooting angle to obtain detailed data. In medium-density areas, the drone maintains a medium flight altitude and speed, taking into account both collection accuracy and efficiency. In low-density areas, the drone is adjusted to a higher flight altitude, increases the flight speed and expands the shooting angle.

[0134] S55. The UAV performs inspection tasks according to the optimized task allocation and flight trajectory, and responds to changes in crop density distribution information in real time during the inspection process.

[0135] Embodiment 1:

[0136] Example 1 In a high-yield agricultural area, farmers introduced drone inspection technology to improve farmland management efficiency. The area mainly grows corn, covering an area of ​​2,000 hectares. The terrain is complex, including plains, hills and some low-lying land, and the crop planting density is uneven. The high-density areas are mainly concentrated in plains, medium-density areas are distributed in hilly areas, and low-density areas are mainly low-lying and marginal areas. Due to different crop growth cycles, the density distribution of corn varies greatly. Traditional fixed-route drone inspections are insufficient in high-density areas and repeated in low-density areas, with low efficiency and incomplete data. Farmers are in urgent need of an efficient and accurate inspection method.

[0137] To solve this problem, we conducted an experiment on the UAV group-controlled crop density adaptive inspection method based on the present invention in the middle of corn growth (early June). The test site was an experimental farmland in the area with a total area of ​​200 hectares, including 50 hectares of high-density areas, 100 hectares of medium-density areas and 50 hectares of low-density areas.

[0138] Before the inspection begins, the ground data processing system receives historical meteorological data, crop planting information, and farmland terrain data to initialize the inspection parameters. The drone swarm consists of 6 high-performance drones, each equipped with a multispectral camera and a high-resolution RGB camera. The drone swarm control system communicates with the ground data processing system in real time through a wireless network, and can dynamically adjust flight parameters during the inspection.

[0139] In the experiment, we limited the UAV's flight altitude to between 10-80 meters, set the flight speed to 3-10 meters per second, and captured images once per second in high-density areas, once every 3 seconds in medium-density areas, and once every 5 seconds in low-density areas. The ground data processing system used a fractal geometry analysis model to process the collected data in real time and generate multi-scale distribution information maps, dynamically allocating UAV tasks and flight trajectories.

[0140] After the inspection mission begins, the drone swarm first conducts an initial inspection over the farmland according to the initial parameters. The ground data processing system receives real-time image data, performs multi-scale analysis on the image based on the box counting method of fractal geometry, calculates the fractal dimension of crop density, and divides the farmland area into high-density, medium-density and low-density areas according to the fractal dimension.

[0141] High-density areas are mainly distributed in plains. The drone group control system assigns two drones to conduct key inspections in the area. The drones reduce their flight altitude to 20 meters, their flight speed to 3 meters per second, and their shooting angle to 30 degrees to ensure the precision of the collected images. The ground data processing system analyzes the collected image data in real time, captures subtle changes in crop leaf coverage, and generates high-resolution density distribution maps.

[0142] In the medium-density area, which is mainly located in the hilly areas, three drones were assigned to patrol. The flight altitude was set to 40 meters, the speed was maintained at 5 meters per second, and the shooting angle was adjusted to 60 degrees to ensure the balance between data collection efficiency and image quality. During the patrol, the drones covered the entire area in a grid and dynamically adjusted the path to optimize the patrol efficiency.

[0143] Low-density areas, distributed in marginal plots and low-lying areas of farmland, are quickly inspected by a drone. The flight altitude is increased to 60 meters, the speed is increased to 10 meters per second, and the shooting angle is expanded to 90 degrees to cover a larger area. The collected data is used to verify the basic distribution characteristics of low-density areas.

[0144] Due to the fixed flight routes, the traditional method has low collection accuracy in high-density areas and some details are not captured; there is repeated collection in medium-density areas, which is inefficient; in low-density areas, some areas are not inspected due to limited coverage. In contrast, the present invention significantly improves the collection accuracy in high-density areas by real-time adjustment of flight parameters and task allocation, optimizes the inspection path in medium-density areas, reduces redundant data, and achieves rapid and comprehensive coverage in low-density areas.

[0145] In addition, the fractal dimension results of the collected data showed that the average fractal dimension of the high-density area was 1.85, that of the medium-density area was 1.65, and that of the low-density area was 1.35, which was highly consistent with the actual planting density distribution, verifying the effectiveness of fractal geometry analysis.

[0146] This Example 1 verifies the feasibility of the present invention through application in actual farmland. The results show that the method of the present invention is significantly superior to traditional methods in terms of patrol efficiency, data collection accuracy and coverage, providing accurate and efficient technical support for farmland crop density monitoring, while significantly reducing patrol costs and time consumption, and has broad application prospects.

[0147] The present invention introduces a multi-scale fractal geometry analysis model in the processing of UAV image data, and uses an improved box counting method and multi-fractal spectrum analysis technology to capture the complex characteristics of crop density distribution from multiple scales and multiple levels. Unlike the traditional single density calculation method, the present invention can dynamically adjust the analysis accuracy based on different fractal box sizes, apply fine-grained fractal geometry models to high-density areas to capture slight changes in crop density, and use simplified models to reduce calculation complexity in low-density areas. Multi-scale fractal geometry analysis makes up for the shortcomings of traditional methods in dealing with complex and irregular crop distributions, and significantly improves the accuracy and reliability of crop density distribution data.

[0148] The present invention maps the distribution characteristics of high, medium and low density areas to geographic coordinates by generating a multi-scale distribution information graph, dynamically allocates inspection tasks and generates optimized flight trajectories based on regional characteristics and UAV performance parameters. Compared with traditional fixed routes or simple adaptive adjustment UAV inspection methods, the present invention innovatively achieves efficient collaboration between the UAV group control system and the ground data processing system, allowing UAVs to prioritize low-altitude inspections in high-density areas to obtain fine data, quickly cover low-density areas to save time and resources, and achieve globally optimized inspection path planning. The inspection task allocation is adjusted through multi-scale distribution information graphs and weight functions, which significantly improves the inspection coverage efficiency and the spatial uniformity of inspection data.

[0149] The present invention receives density distribution update data from the ground processing system in real time and dynamically adjusts flight parameters in combination with the UAV adaptive patrol system, so that the UAV can respond to changes in crop density distribution in real time during the patrol process, and adopts differentiated parameter adjustment strategies for different density areas. Dynamic regulation and control of low-altitude, slow and fine patrols in high-density areas, grid patrols in medium-density areas, and high-altitude rapid coverage in low-density areas ensures the adaptability of patrol tasks to crop distribution characteristics. The real-time linkage of task allocation and flight parameter adjustment significantly reduces the repeated operations and blind area coverage of UAVs, and improves energy utilization efficiency and patrol quality.

[0150] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for crop density adaptive inspection by drone group control based on fractal geometry analysis, characterized in that: The steps include: S1. Use multiple drones in the drone group to carry out initial inspections of the preset farmland area, respectively equipped with image acquisition sensors, to collect image data of ground crop density distribution information, and dynamically adjust the inspection frequency according to the crop growth cycle. In the late stage of crop growth, the inspection frequency is increased according to the crop growth rate; S2. The ground data processing system preliminarily processes the image data collected by each drone based on the fractal geometry analysis model, uses the box counting method to perform fractal geometry analysis on the global distribution of crop density to generate the fractal dimension of density distribution, and preliminarily classifies the density area based on the fractal dimension information; S3. For high-density areas, a fine-grained fractal geometry analysis model is applied, a smaller fractal box size is set in the fractal geometry box counting method, the minute density changes of crop groups in high-density areas are captured, crop density distribution information in high-density areas is obtained, and fine-grained fractal density distribution data is generated; For medium-density areas, a medium-sized fractal box is set to perform fractal geometry analysis to extract the density distribution characteristics of the medium-density area; For low-density areas, a larger fractal box is set to simplify fractal geometry analysis and obtain simplified fractal density distribution data of low-density areas; S4. The ground data processing system integrates the fine-grained fractal density distribution data of high-density areas, the fractal density distribution data of medium-density areas, and the simplified fractal density distribution data of low-density areas collected by multiple drones to generate a multi-scale distribution information map of crop density, and distributes the multi-scale distribution information map to the adaptive inspection system of each drone; S5. The drone group control system assigns the inspection tasks and flight trajectories of each drone according to the density data of the multi-scale distribution information map, and the adaptive inspection system adjusts the flight parameters according to the assigned tasks. In high-density areas, the drone is adjusted to lower the flight altitude, slow down the flight speed and reduce the shooting angle; in medium-density areas, the drone is maintained at a medium altitude and speed for inspection; in low-density areas, the flight altitude is increased, the flight speed is increased and the shooting angle is expanded; S6. During the drone inspection process, S1 to S5 are executed cyclically, so that the drone group continuously updates and adjusts the flight parameters and task allocation according to the real-time collected image data and fractal geometry analysis results, dynamically responds to the distribution changes of crop density in the farmland area, and generates and updates the multi-scale distribution information map of crop density in real time; S7. After the inspection is completed, the image data of the entire area collected by the drone group is transmitted to the ground data processing system, and the image data of each area is subjected to multi-scale fractal geometry analysis to generate a multi-level crop density analysis report including crop-dense areas, sparse areas and their fine-grained distribution characteristics. The multi-level crop density analysis report provides global to fine-grained crop density information based on the fractal geometry analysis results of different density levels; S8. In the multi-level crop density analysis report, identify and mark abnormal density areas, including areas with abnormally high density and areas with abnormally sparse density. The system automatically generates corresponding abnormal warning information and sends it to the agricultural management end. At the same time, the abnormal areas are prioritized as the inspection focus of the drone group.

2. The method for adaptive inspection of crop density by drone group control based on fractal geometry analysis according to claim 1 is characterized in that: The S1 specifically includes the following sub-steps: S11, using multiple drones in the drone group to carry out initial inspections of the preset farmland area, respectively equipped with image acquisition sensors, to obtain image data of ground crop density distribution information; S12, setting the initial flight parameters of the UAV, the initial flight parameters including the flight altitude H0, the flight speed V0 and the shooting interval I0; S13. Dynamically adjust the drone inspection frequency F(t) according to the crop growth cycle and real-time crop density distribution information collected in the farmland area: Among them, F min is the lowest frequency of inspection, used for image collection in the early stage of crop growth, F max is the highest inspection frequency, which is used for intensive collection in the late stage of crop growth. k is the growth rate adjustment coefficient, which is used to control the variation range of inspection frequency. t is the current crop growth time. c It is the turning point time of crop growth rate, which is used to indicate that crops have entered a rapid growth period; S14. In the late stage of crop growth, according to the high-density area ratio P in the real-time density distribution information h Dynamically adjust the inspection frequency. If P h >P th , then the inspection frequency F(t) increases to the maximum value F max , otherwise, keep the current frequency unchanged, where P th It is the preset high-density area ratio threshold, which is used to determine whether the crops have entered a dense distribution state; S15. The UAV performs continuous image acquisition of crop density distribution information according to the adjusted patrol frequency F(t), and sends the acquired image data to the ground data processing system in real-time transmission.

3. The method for crop density adaptive inspection by drone group control based on fractal geometry analysis according to claim 1 is characterized in that: The S2 specifically includes the following sub-steps: S21, receiving ground crop density distribution information image data collected by the drone group, inputting the image data into a ground data processing system, and constructing an initial image data set for crop density distribution analysis; S22, preprocessing the initial image data set, including grayscale conversion, noise removal, geometric correction and orthographic projection transformation of the image, to generate standard image data suitable for fractal geometry analysis, wherein the standard image data is represented in the form of a two-dimensional pixel matrix, and the grayscale value G(x, y) of each pixel reflects the density of the corresponding ground crop; S23, based on the improved multi-scale box counting method, the standard image data is subjected to fractal geometry analysis, and the initial box size r0 and scale factor s are set, and the box size sequence is r i =r0 / s i (i=0,1,2,…,n), for each scale r i Divide the standard image data into r i × i Box B k (r i ), for each box B k (r i ) Calculate the cumulative sum S of the grayscale values ​​inside the box k (r i ): Calculation scale r i The total gray value S under total (r i ): Among them, N(r i ) is the scale r i The total number of boxes under; Calculate the scaling function F(q,r i ): Among them, q is the multifractal order parameter, which is used to capture the multi-scale characteristics of crop density distribution; S24. Calculate the generalized scaling exponent τ(q) and the generalized fractal dimension D for different values ​​of q q : Generalized fractal dimension D q Reflects the complexity of crop density distribution at different scales and in different density areas; S25. According to the generalized fractal dimension D q Construct a multifractal spectrum and calculate the characteristic parameters of the fractal spectrum, including the spectrum width ΔD and the spectrum peak position: in, and are the generalized fractal dimensions corresponding to the selected minimum and maximum q values ​​respectively; Crop density areas are classified according to the fractal spectrum width ΔD: When the fractal spectrum width ΔD>ΔD high When , the area is determined to be a high-density area; When the fractal spectrum width ΔD low ≤ΔD≤ΔD high When , the area is determined to be a medium-density area; When the fractal spectrum width ΔD<ΔD low When , the area is determined to be a low-density area; Where, ΔD high and ΔD low It is a preset fractal spectrum width threshold, which is used to distinguish crop areas with different density levels.

4. The method for crop density adaptive inspection by drone group control based on fractal geometry analysis according to claim 1 is characterized in that: The S4 specifically includes the following sub-steps: S41, receiving the fine-grained fractal density distribution data of high-density areas, the fractal density distribution data of medium-density areas, and the simplified fractal density distribution data of low-density areas collected by multiple drones, fusing the data of each drone based on a unified geographic coordinate system, and constructing an initial set of crop density data from multiple drones S42, for the initial set Perform spatial registration processing to achieve spatial alignment of each drone data by minimizing the registration error function ∈(x,y). The error function is defined as: Among them, M and N represent the number of UAVs involved in data fusion and the number of pixels of each UAV respectively; G i (x,y) and G j (x, y) are the density distribution values ​​of the i-th and j-th drones at the geographic coordinates (x, y), and after optimization, the fused density distribution data D is obtained. f (x,y); S43, according to the fused density distribution data D f (x, y) Multi-scale feature extraction is performed on the data of high-density areas, medium-density areas, and low-density areas to generate a fine-grained distribution feature matrix M for high-density areas h , medium density region feature matrix M m and the low-density region feature matrix M l , the feature extraction formula is as follows: where z∈{h,m,l} represents high-density, medium-density, and low-density areas; G z (x, y) is the density distribution value of region z; R is the scale range of the fractal box size, which is used for multi-scale feature extraction; [x1, x2] and [y1, y2] are the coordinate ranges of a specific region; S44. Generate a multi-scale distribution information map of crop density using the fine-grained distribution feature matrix of high-density areas, the feature matrix of medium-density areas, and the feature matrix of low-density areas Among them, α z (r) is the regional weight function, which is used to adjust the weights of different density regions in the multi-scale distribution information map; S45. Generate a multi-scale distribution information map The adaptive patrol system assigned to the drones adjusts the flight parameters and patrol tasks of each drone according to the characteristics and weight distribution of high, medium and low density areas in the information map, so that the drones can dynamically respond to changes in the density of farmland crops.

5. The method for adaptive inspection of crop density by drone group control based on fractal geometry analysis according to claim 1 is characterized in that: The S5 specifically includes the following sub-steps: S51. The UAV group control system receives the multi-scale distribution information map of crop density generated by the ground data processing system. The density data in the multi-scale distribution information map is classified into high-density area, medium-density area and low-density area, and each classification result is used as the input of UAV task allocation; S52. According to the classification results of crop density distribution and the flight capability parameters of each UAV, inspection tasks are assigned. The task assignment follows the following optimization objectives: Among them, Φ is the total inspection cost function, U is the number of drones, T i,z represents the patrol time of the i-th drone in area z, C i is the energy consumption of the i-th UAV, ω1 and ω2 are the weight coefficients of time and energy consumption; S53, generating optimized flight trajectories for each drone. The flight trajectories in high-density areas are mainly based on covering key areas. When planning flight trajectories, the drone's residence time and route overlap rate in high-density areas are increased. The flight trajectories in medium-density areas use a gridding method to balance coverage and collection efficiency. The flight trajectories in low-density areas use a sparse coverage mode to maximize the coverage of the drone's flight path. S54. The adaptive inspection system dynamically adjusts the flight parameters of the drone, including flight altitude, flight speed and shooting angle, according to the task allocation results and flight trajectory. In high-density areas, the drone is adjusted to a lower flight altitude and slows down the flight speed, while narrowing the shooting angle to obtain detailed data. In medium-density areas, the drone maintains a medium flight altitude and speed, taking into account both collection accuracy and efficiency. In low-density areas, the drone is adjusted to a higher flight altitude, increases the flight speed and expands the shooting angle. S55. The UAV performs inspection tasks according to the optimized task allocation and flight trajectory, and responds to changes in crop density distribution information in real time during the inspection process.

Images