An unmanned aerial vehicle flight parameter adaptive optimization water bird population investigation method

Abstract

The application relates to the technical field of wild animal monitoring and ecological protection, and discloses a kind of water bird population survey method of adaptive optimization of unmanned aerial vehicle flight parameter, which comprises the following steps: first, a multi-dimensional aerial image interpretation sign library is constructed, a flight parameter and water bird disturbance response correlation test is carried out, a quantitative behavior response spectrum is established, then, a flight distance probability model is constructed according to the response spectrum, the minimum safe monitoring distance threshold is calculated, and the best monitoring flight parameter combination is determined in combination with statistical accuracy and environmental factors, and then the aerial photography task is executed to obtain population data.In addition, the method uses unmanned aerial vehicle remote sensing data to extract habitat factors, constructs a habitat suitability evaluation system, identifies the dominant limiting factor to generate repair countermeasures, and delimits dynamic protection partition.The application effectively solves the contradiction between image resolution and wild animal safety in unmanned aerial vehicle monitoring, and realizes precise decision-making for water bird protection and habitat restoration.

Description

Technical Field

[0001] This invention relates to the field of wildlife monitoring and ecological protection technology, specifically a method for surveying waterbird populations with adaptive optimization of UAV flight parameters. Background Technology

[0002] Waterbirds, as indicator species of wetland ecosystems, are crucial for assessing wetland ecological quality through their population size and distribution. Traditional waterbird surveys primarily rely on manual ground observation, using optical instruments such as telescopes for counting. However, wetland environments are typically characterized by complex topography, with large areas of reed beds, marshes, or open water. Ground observations are easily affected by vegetation obstruction, distance limitations, and blind spots, leading to significant errors in population statistics and low operational efficiency, making it difficult to comprehensively cover large-scale monitoring areas.

[0003] With the rapid development of low-altitude remote sensing technology, drones, due to their maneuverability, wide field of view, and high-resolution imaging capabilities, have gradually become an important tool for waterbird surveys. However, in practical applications, drone monitoring faces an inherent contradiction between image resolution and wildlife safety. To obtain high-definition images for accurate species identification, drones typically need to maintain a low flight altitude, which often generates high-intensity noise and visual stimulation, inducing sensitive waterbirds to become alert, flee, or even take flight in alarm. This human disturbance not only directly affects the normal habitat and breeding of waterbirds but also causes the survey subjects to leave the monitoring area before the data collection is completed, resulting in distorted monitoring data. Currently, the selection of parameters such as flight altitude, speed, and approach method in drone survey operations largely relies on the operator's subjective experience, lacking a quantitative model based on waterbird behavioral response mechanisms. This makes it impossible to accurately calculate the minimum safe distance threshold for different species while ensuring data accuracy.

[0004] Furthermore, existing waterbird monitoring efforts often stop at population statistics, failing to deeply couple population data with habitat environmental quality analysis. Traditional habitat assessments are mostly based on two-dimensional satellite remote sensing imagery, making it difficult to accurately extract three-dimensional structural information such as vegetation height, which is crucial for waterbird survival, resulting in rather crude suitability assessment results. Simultaneously, due to the lack of quantitative diagnosis of the dominant limiting factors leading to habitat quality decline, subsequent ecological restoration measures often lack clear spatial direction and strategic focus, making it difficult to form a complete technical closed loop from population surveys and problem diagnosis to restoration decision-making, thus limiting the improvement of refined wetland management. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for adaptive optimization of UAV flight parameters for waterbird population surveys. This method resolves the conflict between image resolution and wildlife safety in UAV waterbird surveys and overcomes the lack of three-dimensional habitat structure diagnosis and precise restoration decision-making loops in traditional methods.

[0006] To achieve the above objectives, the present invention provides a method for adaptive optimization of UAV flight parameters for waterbird population surveys, which mainly includes the following technical processes: First, a baseline survey of the target waterbird species within the target area was conducted. Multi-angle aerial images were collected using drones, and a multi-dimensional aerial image interpretation marker library was constructed through feature extraction. This marker library covers geometric morphology, spectral color, texture, and shadow features. In particular, a shape index was obtained by calculating the quotient of the square of the perimeter of the target waterbird's edge contour multiplied by the product of the projected pixel area and pi. This shape index serves as a quantitative benchmark for distinguishing between waterfowl and wading birds, providing a comparative basis for subsequent image interpretation.

[0007] Secondly, correlation experiments were conducted between UAV flight parameters and waterbird disturbance responses. A multi-factor experimental matrix was constructed, incorporating different flight altitudes, speeds, and approach methods, and coordinate data was acquired in real time to calculate the spatial slant distance between the UAV and the target waterbird. During this process, a quantified four-level behavioral response spectrum was established, recording the target waterbird's response level as the slant distance changed. The four-level behavioral response spectrum categorized the waterbird's state into unresponsive, alert, avoidant, and startled states, achieving a standardized classification of the disturbance level.

[0008] Next, a probability model for the startled flight distance is constructed based on the startled response data. Using a binomial logistic regression model, the startled flight state is defined as "startled flight has occurred," and all other states are defined as "not startled flight has occurred." The minimum safe monitoring distance threshold for the target waterbird is calculated by dividing the difference between one and a preset acceptable startled probability threshold by the threshold itself, taking the natural logarithm of this ratio, and combining this with the intercept and slope parameters of the regression model.

[0009] Subsequently, the optimal combination of monitoring flight parameters was determined. This process required simultaneous consideration of safety, statistical accuracy, and environmental adaptability. On one hand, based on the population statistical accuracy verification results, the maximum permissible flight altitude and maximum permissible flight speed that meet high accuracy requirements were determined. On the other hand, based on environmental meteorological factors, the effective range of solar altitude angle and wind speed threshold that maximizes image background contrast and minimizes reflectivity were determined. The final selected combination of flight parameters must ensure that the combined vector magnitude of flight altitude and horizontal distance is greater than or equal to the minimum safe monitoring distance threshold, and simultaneously fall within both the high accuracy range and the environmentally suitable range.

[0010] Finally, the drone aerial photography mission was carried out using the determined optimal combination of monitoring flight parameters, and the acquired images were interpreted using the aforementioned interpretation mark library to obtain high-precision data on the population size and distribution of waterbirds.

[0011] Furthermore, this invention also applies the data obtained by the above-mentioned survey methods to habitat environmental factor extraction and suitability assessment: In terms of environmental factor extraction, the structure-reconstruction-motion algorithm was used to process UAV imagery and construct a 3D dense point cloud model to generate a digital surface model and a digital elevation model. By calculating the difference between the two, a normalized digital surface model was obtained, thereby accurately extracting the vegetation height factor. At the same time, the visible light atmospheric impedance vegetation index or the supergreen index was used to extract the vegetation cover factor, and combined with ground survey data on water depth, food, and disturbance sources, the spatial quantification of environmental factors was completed.

[0012] In terms of suitability assessment, a habitat suitability assessment system was constructed. A single-factor suitability index assessment model was built for each environmental factor, and the weights were determined using the analytic hierarchy process (AHP). The weighted geometric mean method was used to calculate the comprehensive habitat suitability index, which involves exponentially multiplying the single-factor suitability indices by the environmental factor weights and then multiplying all the exponentiation results together.

[0013] Furthermore, this invention generates targeted remediation strategies and protection zones based on the evaluation results: For areas with a low overall habitat suitability index, a limiting factor diagnosis is performed. The single-factor suitability index with the lowest value is identified as the dominant limiting factor, and remediation strategies are matched accordingly: when the dominant limiting factor is unsuitable water depth, micro-topography modification or water level regulation is adopted; when the vegetation structure is unsuitable, mechanical mowing or replanting of emergent plants is adopted; when food scarcity is the cause, biological manipulation or habitat enhancement is adopted; when excessive human disturbance is the cause, a physical isolation zone is established based on the minimum safe monitoring distance threshold calculated above.

[0014] Simultaneously, dynamic protection zones are delineated based on the numerical distribution characteristics of the comprehensive habitat suitability index: highly suitable areas are designated as core protected areas, where low-altitude flights are prohibited; moderately suitable areas with good connectivity are designated as ecological restoration areas, serving as target areas for implementing restoration strategies; and low-suitable areas located on corridors are designated as buffer zones. After implementing restoration strategies, the aforementioned survey and evaluation process is repeated, and a closed-loop assessment of ecological performance is achieved by comparing the changes in the index before and after restoration.

[0015] This invention provides a method for adaptively optimizing UAV flight parameters to conduct waterbird population surveys. It offers the following advantages: 1. This invention constructs a waterbird behavioral response spectrum and a binomial logistic regression startled flight probability model to calculate the minimum safe monitoring distance threshold. It also comprehensively considers the population statistical accuracy verification results and environmental meteorological factors to determine the optimal combination of monitoring flight parameters. Technically, this method quantifies the nonlinear relationship between UAV flight parameters and waterbird disturbance response, effectively resolving the contradiction between image resolution and wildlife safety in near-ground remote sensing monitoring. While ensuring high accuracy of waterbird population statistics, it significantly reduces human disturbance to sensitive species during survey operations.

[0016] 2. This invention utilizes the UAV motion recovery structure algorithm to construct a three-dimensional dense point cloud, generate a normalized digital surface model to extract vegetation height factors, and combine the visible light vegetation index and weighted geometric mean method to construct a habitat suitability evaluation system. Compared with traditional monitoring methods based on two-dimensional satellite imagery, this technical solution can accurately capture the vertical structural characteristics of wetland habitats and reflect the limiting bottleneck effect of habitat factors through the geometric mean algorithm, thereby more objectively and realistically reflecting the quality status and spatial heterogeneity of waterbird habitats.

[0017] 3. This invention identifies dominant limiting factors based on a single-factor suitability index, and generates targeted restoration strategies including water level regulation, vegetation management, and physical isolation zones based on safe monitoring distance thresholds. It also delineates dynamic protection zones. This decision generation mechanism based on quantitative diagnosis establishes a technical closed loop from population surveys and habitat problem diagnosis to precise restoration and ecological performance evaluation, providing a spatially clear and specific operational basis for the refined management of wetlands. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram illustrating the verification of population statistical accuracy and the determination of the optimal monitoring window period in this invention; Figure 3 This is a schematic diagram of the process for acquiring environmental factors of multi-source habitats and standardizing data processing according to the present invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] See attached document Figure 1This invention provides a method for adaptive optimization of UAV flight parameters for surveying waterbird populations. The method includes step S1: conducting a baseline survey of the target species and constructing a multi-dimensional aerial image interpretation marker library.

[0021] Step S1 specifically includes conducting a baseline survey of the biological and ecological characteristics of five target waterbird species in the study area: Great Bustard, Great Egret, Grey Heron, Ruddy Shelduck, and Common Coot. Through a combination of ground-based fixed-point observations and literature data retrieval, morphological characteristics, lifestyle habits, diurnal activity rhythms, seasonal migration patterns, and reproductive status data of the target waterbirds within the study area are obtained. Morphological characteristics data include the species' body length, wingspan, plumage, and flight posture; lifestyle habits data include flocking patterns, startled flight distance thresholds, and sensitivity to environmental disturbances; reproductive status data includes the start time and duration of the breeding season and nest site preferences. Based on the above data, the optimal time window and key coverage areas for UAV monitoring are determined. For example, for the Great Bustard during the breeding season, the focus is on covering the meadow or farmland habitats where its courtship grounds and nesting sites are located.

[0022] After completing the baseline survey, multi-rotor drones equipped with high-resolution visible light cameras were used to acquire image data of target waterbirds to construct an aerial image interpretation tag library. Drone flights were conducted under meteorological conditions with sufficient sunlight and wind speeds less than 5 m / s. For different habitats including water bodies, grasslands, bare land, and reed beds, multi-angle aerial images of target waterbirds in various behavioral states—stationary, foraging, and flying—were collected. During the acquisition process, the shooting height, shooting angle, solar altitude angle, and environmental background type for each image were recorded to ensure the diversity and representativeness of the sample data.

[0023] In constructing a multi-dimensional aerial image interpretation marker library, the acquired raw images undergo preprocessing, including geometric and radiometric correction, followed by manual annotation and feature extraction of target waterbirds. Feature extraction covers four dimensions: geometric morphology, spectral color, texture, and shadow features.

[0024] Based on geometric morphological characteristics, the projected area, perimeter, aspect ratio, and shape index of the target waterbirds were extracted from a top-down view. The projected area was used to distinguish between the Great Bustard (large size) and the Ruddy Shelduck (medium size) or the Coot (small size); the shape index was used to distinguish between rounded waterfowl (such as the Coot) and slender wading birds (such as the Grey Heron). Shape index... The calculation formula is as follows: ; in, This represents the shape index, which is a dimensionless number. This indicates the perimeter of the edge outline of the target waterbird in the image; This represents the area of ​​the projected pixels of the target waterbird on the image. Pi is the mathematical constant of a circle.

[0025] When the target projection is close to a circle The value is close to 1; when the target projection is elongated or irregular in shape, The value is significantly greater than 1.

[0026] To mitigate the impact of illumination variations on color recognition, the image was converted from the RGB color space to the HSV (Hue, Saturation, Luminance) color space based on its spectral color characteristics. The mean hue of the target waterbird region was then extracted. Mean saturation and average brightness .

[0027] For Ruddy Shelduck, extract its characteristic orange-yellow hue range; for Great Egret, extract its high-brightness feature values ​​to segment it from dark backgrounds (such as water or vegetation); for Coot, extract its low-brightness and low-saturation black feather feature values.

[0028] For texture features, the gray-level co-occurrence matrix (GLCM) is used to extract the energy, contrast, correlation, and entropy of the target surface. For the Great Bustard, due to the camouflage pattern on its back feathers, its texture entropy value is significantly higher than that of the Great Egret with its uniform plumage. Texture features can be used to effectively distinguish Great Bustard individuals in complex grassland backgrounds.

[0029] Based on shadow features, the length and shape of shadows in the images are used to assist in interpreting the three-dimensional structure of target waterbirds. In the early morning or late afternoon when the sun's altitude angle is low, the long necks of grey herons and great egrets will cast S-shaped or long strip-shaped shadows on the ground or water surface. These shadow features are extracted and stored in the interpretation mark library as an auxiliary criterion for distinguishing wading birds from waterfowl.

[0030] Finally, the extracted geometric morphology, spectral color, texture, and shadow features are structured and stored to create multi-dimensional feature vectors for each target waterbird under different habitat backgrounds, forming a standardized aerial image interpretation tag library. This tag library includes interpretation datasets for Great Bustards, Great Egrets, Grey Herons, Ruddy Shelducks, and Coots. Each dataset contains corresponding typical image slices and their quantified feature parameter ranges, providing comparison benchmarks and training samples for subsequent UAV automatic identification and population statistics.

[0031] The method of the present invention includes the following steps in step S2: conducting a correlation experiment between UAV flight parameters and waterbird disturbance response based on multivariate gradients, and establishing a quantitative four-level behavioral response spectrum.

[0032] Step S2 specifically involves constructing a multi-factor experimental matrix consisting of flight altitude, flight speed, and approach method. The experimental variables are designed as follows: Flight altitude variable Five discrete altitude gradients were set at 30 meters, 50 meters, 80 meters, 100 meters, and 120 meters. This altitude gradient covered a range from low-altitude areas that strongly intimidate waterbirds to high-altitude areas typically considered for covert surveillance. Flight speed variable. The approach speed is set to three discrete gradient levels: 3 m / s, 5 m / s, and 8 m / s. The approach mode is set to horizontal approach mode, meaning the UAV maintains a horizontal attitude at the predetermined altitude and flies in a straight line from the periphery of the target waterbird directly above it.

[0033] During the experiment, the UAV's onboard high-precision Global Positioning System (GPS) module and Inertial Measurement Unit (IMU) recorded flight trajectory data at a frequency of 1 Hz, including timestamps, longitude, latitude, absolute altitude, and flight speed. Simultaneously, ground monitoring personnel used a laser rangefinder to determine the initial latitude, longitude, and altitude of the target waterbird. Based on the UAV's real-time coordinates... Coordinates of the target waterbird Calculate the Euclidean distance (slant distance) between the UAV and the target waterbird at each moment. The calculation formula is as follows: ; in, This represents the straight-line slant distance between the drone and the target waterbird; The three-dimensional coordinates of the UAV in a Cartesian coordinate system; This represents the three-dimensional coordinates of the target waterbird in a Cartesian coordinate system. The slant distance... As the core physical quantity for determining the threshold of waterbird disturbance, it is directly related to the subsequent construction of the safety threshold model.

[0034] To quantify the degree of disturbance experienced by Great Bustards, Great Egrets, Grey Herons, Ruddy Shelducks, and Coots under different combinations of flight parameters, this embodiment defines a standardized four-level Behavior Spectrum. This behavioral spectrum divides the waterbirds' responses to external stimuli, from indifference to flight, into four discrete levels. Monitoring personnel simultaneously determine the instantaneous response level of individual targets using both ground-based high-powered telescopes and UAV-based downward-looking images.

[0035] Level 0 Response (No Response): The target waterbird maintains its natural behavioral patterns, including foraging, resting, preening, or normal social interaction. For waterfowl (such as ruddy shelducks and coots), this manifests as swimming steadily on the surface or diving headfirst to forage, without changing direction or speed. For wading birds (such as great egrets and herons) and landfowl (such as great bustards), this manifests as maintaining a standing posture or walking slowly, with its head not pointing towards the drone. In this case, the drone is determined to be beyond the undetectable distance.

[0036] Level 1 response (Alert state): The target waterbird interrupts its current foraging or resting behavior, showing attention to the drone. Specific characteristics include: rapidly raising its head and holding it still (neck extended posture), following the drone's movement with its gaze, or frequently turning its head left and right for a vigilant scan. For Great Bustards, this manifests as a significantly straightened neck and cessation of movement; for Ruddy Shelducks and Coots, it manifests as ceasing paddling, holding their necks upright, and shortening the distance between groups. The slant distance between the drone and the target when this state occurs is recorded and defined as the Alert Distance (AD).

[0037] Level 2 Response (Avoidance): The target waterbird exhibits obvious displacement avoidance behavior but has not yet taken off. Specific characteristics include: waterfowl (Ruddy Shelduck, Common Coot) increase swimming speed, change swimming paths to move away from the drone's projection point, or quickly swim towards reeds or other concealed areas; wading birds (Great Egret, Grey Heron) and landfowl (Great Bustard) exhibit rapid running or jumping behavior, with body postures showing preparatory movements for takeoff (such as leaning forward and slightly spreading wings). This stage indicates that the drone has entered the area that would substantially interfere with the target.

[0038] Level 3 Response (Flush): The target waterbird exhibits a violent escape response. Specifically, it flaps its wings to completely detach from the ground or water surface, initiating powered flight and flying away from its current habitat patch. The slant distance between the UAV and the target at the instant this occurs is recorded and defined as the Flight Initiation Distance (FID). This FID value is a key parameter for determining the minimum safe distance for UAV ecological monitoring.

[0039] During the experiment, for each target species (Great Bustard, Great Egret, Grey Heron, Ruddy Shelduck, and Common Coot), different [measures were taken]. and Repeat the effective approach test at least 30 times under the combined conditions, and record the reaction time of the target waterbird as it gradually changes from level 0 to level 3 in each test. Numerical data are used to create a structured dataset that includes species category, environmental context, flight parameters, and behavioral response thresholds. This standardized experimental procedure and quantitative grading definition eliminate the subjective ambiguity of human descriptions, ensuring the objectivity and repeatability of the acquired disturbance threshold data.

[0040] The method of the present invention further includes, in step S2: constructing a flight-in-the-flight (FID) probability model based on the aforementioned behavioral response spectrum data and calculating the minimum safe monitoring distance threshold.

[0041] Step S2 specifically involves binary processing of the structured experimental data obtained in step S2.2. This involves classifying the behavioral response levels of the target waterbirds. Convert to a binary response variable .when (i.e., when a startled flight occurs), the definition is... ;when (That is, when there is no startled flight, but only unresponsiveness, alertness, or avoidance) is defined as follows: Record the straight-line slant distance between the UAV and the target for each valid test. As an independent variable A binomial logistic regression model was constructed to quantitatively describe the nonlinear relationship between the probability of startled reaction and the approach distance of the drone.

[0042] Independent regression models were established for each of the five target species: Great Bustard, Great Egret, Grey Heron, Ruddy Shelduck, and Common Coot. (Probability of startled flight) The mathematical expression is as follows: ; in, Indicates a specific slope distance The probability of a target waterbird taking flight is given below, and its value ranges from [value missing]. ; is the base of the natural logarithm; The regression intercept parameter reflects the translation position of the model curve on the horizontal axis and represents the basic sensitivity baseline of the species. The regression slope parameter reflects the rate at which the probability of startled flight changes with distance. Typically, A negative value indicates that as the slant distance between the drone and the target increases... As the number of [something] increases, the probability of being startled decreases in an S-shaped curve.

[0043] Model parameters and The maximum likelihood estimation method is used to solve the problem. An iterative algorithm is employed to construct the likelihood function and solve for the maximum value of the log-likelihood function, thereby obtaining the optimal parameter estimates for a specific species under a specific habitat. This method effectively handles the discreteness and noise present in field experimental data and accurately fits the trend curve of startled flight probability as a function of distance.

[0044] After constructing the startled flight probability model, a standardized acceptable startle probability threshold is set. In this embodiment, to ensure the non-destructive nature of ecological monitoring, The value is set to 0.05, meaning that a maximum of 5% of individuals are allowed to take flight in alarm, while the remaining 95% remain within the habitat patch. Based on this probability threshold, the minimum safe distance threshold for drone operations is calculated by performing an inverse operation on the aforementioned Logistic equation. The calculation formula is as follows: ; in, This is the minimum safe slant distance for the target species. This threshold has a clear physical meaning: when the straight-line spatial distance between the drone and the target waterbird is greater than... At that time, the probability of waterbirds taking flight and escaping is lower than the set safety limit. .

[0045] Furthermore, in order to convert the calculated straight-line slant range threshold into flight path parameters executable by the UAV flight control system, this embodiment will... Decomposed into flight altitude Horizontal distance The constraints are as follows. In actual monitoring operation planning, the following inequality constraints must be satisfied: ; When a drone performs an orthophoto acquisition task (i.e., it is located directly above the target at a horizontal distance) Minimum safe flight altitude Must meet Therefore, for species such as the Great Bustard that are highly sensitive to disturbance, the system automatically outputs their corresponding... Values ​​(e.g., 120 meters); for more tolerant species such as Coot, the system outputs lower values. Values ​​(e.g., 50 meters). Through this calculation process, the ecologically significant animal behavior thresholds are transformed into rigid flight parameter constraints for UAV remote sensing operations, forming specific "UAV Ecological Monitoring Safety Operation Specifications" for different species.

[0046] See attached document Figure 2The method of the present invention further includes in step S2: verifying the statistical accuracy of the UAV monitoring data through ground synchronous observation, and comprehensively considering environmental meteorological factors to determine the optimal monitoring window period and flight parameter combination that meet the dual constraints of high accuracy and low disturbance.

[0047] Step S2 specifically includes establishing a synchronized air-to-ground verification mechanism. Within the same time window as the UAV's aerial photography mission, ground surveyors use high-powered monoculars (20-60x magnification) at pre-set concealed observation points to manually count and identify the target waterbirds within the experimental plots. The ground observation points are located at least 300 meters away from the target waterbird colony, upwind or crosswind, to ensure that ground personnel do not disturb the birds and guarantee the reliability of the ground data. The population numbers are then extracted after interpreting the UAV images. Population size recorded in synchronous observations with the ground Comparisons were performed to calculate the population statistics accuracy under different combinations of flight parameters. The calculation formula is as follows: ; in, This represents the absolute error between the UAV monitoring data and the ground reference value. This indicator is used to quantitatively evaluate the error at a specific flight altitude. With flight speed The effectiveness of unmanned aerial vehicle (UAV) remote sensing systems in identifying target species.

[0048] Based on the acquired accuracy data, the impact mechanism of flight parameters on image quality and recognition accuracy is analyzed. Regarding flight altitude... There exists a trade-off range between resolution and coverage: as The reduction in ground sampling distance (GSD) and the improvement in image spatial resolution are beneficial for the feature identification of small waterbirds (such as coots) and juveniles, but the coverage area of ​​a single image is reduced and may approach the disturbance threshold. along with As the speed increases, the image resolution decreases, exacerbating pixel mixing and reducing recognition accuracy. Regarding flight speed... The main consideration is the motion blur it causes. When When the image size is too large and exceeds the camera shutter speed limit, image blurring occurs along the flight direction, resulting in loss of texture features. Correlation analysis was used to determine when the following conditions were met... The maximum permissible flight altitude under this preset accuracy standard With maximum permissible flight speed .

[0049] After determining the physical constraints of the flight parameters, environmental meteorological variables are further introduced to determine the optimal monitoring window. The main focus is on the solar altitude angle. With wind speed The coupled impact of two key environmental variables on monitoring effectiveness.

[0050] Regarding the solar altitude angle The effects of this on water surface reflection and shadow occlusion were analyzed. At excessively high angles (nearly 90 degrees of direct sunlight), strong specular reflection occurs on the water surface, causing oversaturation of the background brightness and obscuring the spectral characteristics of the waterbirds; when When the angle is too low (less than 30 degrees), the shadows cast by ground features are too long, which may cause waterbirds to be obscured by the shadows of reeds or embankments, resulting in missed detection. This embodiment determines the effective range of solar altitude angles that maximizes image background contrast and minimizes reflectivity by comparing the image interpretation results at different time periods. .

[0051] Regarding wind speed This study analyzes the impact of wind speed on the attitude stability of UAVs and the distribution of waterbirds. High wind speeds increase the tilt angle of multi-rotor UAVs, causing geometric distortion in orthophotos. Simultaneously, wind and waves alter water surface textures, increasing background noise during interpretation. Furthermore, waterbirds tend to seek shelter on leeward slopes or deep in reed beds under high wind speeds, reducing visibility. This embodiment sets a wind speed threshold. (e.g., 5 m / s) can be used as a workable boundary condition.

[0052] Finally, by finding the intersection of the safe threshold set and the high-precision threshold set, the optimal monitoring parameter space for a specific species is determined. This parameter space must simultaneously satisfy the following three constraints: Safety constraints: Drone flight altitude Horizontal distance The combination satisfies ; Precision constraints: Flight altitude And flight speed To ensure that the image resolution and clarity meet the interpretation requirements; Environmental constraints: Solar altitude angle at the time of operation and instantaneous wind speed .

[0053] The system outputs the parameter combinations that meet all the constraints as a standardized "Target Waterbird UAV Monitoring Operation Specification Table", which clearly indicates the best operation time (such as 09:00-11:00 in the morning or 15:00-17:00 in the afternoon) and the corresponding optimal flight altitude and speed settings for different species such as Great Bustard and Ruddy Shelduck in different seasons.

[0054] See attached document Figure 3 The method of the present invention includes step S3: combining UAV remote sensing technology with ground field survey methods to obtain multi-source environmental factor data covering four dimensions: hydrology, vegetation, food and disturbance, and performing spatialization and standardization processing on the heterogeneous data.

[0055] Step S3 specifically includes extracting microhabitat structure factors from high-precision 3D data generated based on UAV imagery. This fully utilizes the high overlap sequence of UAV images acquired in steps S1 and S2 (heading overlap). The Structure from Motion (SfM) algorithm and Multi-View Stereo (MVS) technology are used to construct a 3D dense point cloud model of the survey area. Based on the point cloud model, a Digital Surface Model (DSM) and a Digital Orthophoto Map (DOM) with a spatial resolution better than 0.1 meters are generated.

[0056] Vegetation height factors are extracted using DSM data. First, ground points and non-ground points in the point cloud are separated using a filtering algorithm. A Digital Elevation Model (DEM) representing the bare terrain surface is then generated using ground point interpolation. Subsequently, a normalized digital surface model (nDSM) is calculated through raster algebra operations. The pixel values ​​of this nDSM represent the absolute height of the ground cover. For the meadow habitat of the Great Bustard breeding grounds, the average vegetation height is extracted. This is used to assess their concealment; for wading bird habitats such as great egrets, the height distribution of emergent plants (such as reeds) is identified to determine whether they obstruct the line of sight for takeoff.

[0057] Extracting vegetation cover factors from DOM data Vegetation indices are calculated based on the visible light band. For conventional UAV cameras lacking a near-infrared band, the Visible Atmospheric Impedance Vegetation Index (VDVI) or the Supergreen Index (ExG) are used to enhance vegetation features. A dynamic threshold is set for binarization segmentation, dividing the image into vegetation pixels and non-vegetation pixels. Within an evaluation cell (e.g., a 10m x 10m grid), the proportion of vegetation pixels to the total number of pixels is calculated as the vegetation cover value for that cell.

[0058] Step S3 also includes obtaining hydrological, food, and anthropogenic disturbance factors through ground-based synchronous surveys and GIS analysis. Since the visible light sensors of UAVs cannot directly penetrate turbid water to measure water depth, this embodiment uses RTK-GPS in conjunction with a depth sounding rod for on-site point measurement. A grid of sampling points is deployed in the waterbird activity area, and the water depth value at each point is measured. The data includes the type of substrate (muddy, sandy, rocky). Using Kriging or inverse distance weighted (IDW) interpolation, discrete water depth sampling data are transformed into a continuous spatial grid surface, generating a comprehensive water depth distribution map. This factor is crucial for wading birds; for example, great egrets can only inhabit shallow waters where the depth is less than the height of their tibia and tarsus joint (approximately 30-40 cm).

[0059] Targeting food richness factors A specific survey was conducted based on the diet (herbivorous, carnivorous, or omnivorous) of the target species. For ruddy shelducks that feed on benthic organisms, bottom sediment samples were collected to determine benthic animal biomass; for great bustards that feed on tender grasses, the aboveground biomass dry weight of edible plants was measured. The biomass data were normalized and assigned to the corresponding habitat patches.

[0060] Targeting human interference factors The study area's roads, settlements, high-voltage power line towers, and farmland boundaries were vectorized using a Geographic Information System (GIS). Euclidean distance analysis was performed to calculate the straight-line distance from any point within the study area to the nearest source of human disturbance. A larger distance value indicates less human disturbance to the habitat and higher ecological security.

[0061] Step S3 further includes performing unified gridded resampling and standardization on the aforementioned multi-source heterogeneous data. All environmental factor layers are uniformly projected onto the same planar coordinate system and resampled into a raster data matrix with the same spatial resolution (e.g., 2m × 2m). Since the dimensions and physical meanings of each environmental factor are different (e.g., water depth in centimeters, cover in percentage, distance in meters), and their contributions to habitat suitability differ, a range standardization method is used to map them to a dimensionless interval of [0,1]. For positive indicators (i.e., factors whose larger values ​​are more conducive to waterbird survival, such as distance from the disturbance source)... Food abundance Standardized formula as follows: ; For negative indicators (i.e., factors whose larger values ​​are more detrimental to the survival of waterbirds, such as the intensity of human activities), the standardized formula... as follows: ; in, These are the measured values ​​of environmental factors within the grid cell; and These are the maximum and minimum values ​​of the factor across the entire study area, respectively.

[0062] For intermediate indicators with an optimal suitable range (such as water depth) (Both too deep and too shallow are unsuitable); the optimal range should be set according to the ecological habits of the target species. .when When the suitability is 1, the value is 1; when or When the degree of deviation decreases, the suitability decreases linearly or non-linearly. After the above processing, a standardized environmental factor dataset is formed, consisting of a water depth suitability layer, a vegetation height suitability layer, a vegetation cover suitability layer, a food richness layer, and a disturbance distance suitability layer, which serves as the input variables for the subsequent construction of the HS model.

[0063] The method of the present invention includes the following steps in step S4: constructing a single-factor suitability curve for the target species based on the principle of ecological niche, determining the weights of environmental factors using multi-criteria decision analysis, and establishing a comprehensive habitat suitability index model to generate a spatialized suitability distribution map.

[0064] Step S4 specifically involves establishing a single-factor suitability index (SI) evaluation model for each target species (Great Bustard, Great Egret, Grey Heron, Ruddy Shelduck, and Common Coot) based on their specific ecological needs. The SI model maps the standardized environmental factor values ​​from step S3 into a dimensionless index characterizing biological suitability, with values ​​ranging from 0 to 1, where 0 represents completely unsuitable and 1 represents optimally suitable.

[0065] For different types of environmental factors, Sl functions with different mathematical forms are constructed. For intermediate factors with a clearly defined optimal ecological amplitude (such as water depth)... The SI curve is constructed using a Gaussian distribution function. Taking the Great Egret as an example, its optimal foraging water depth is related to the length of its tarsometatarsal bones. An optimal mean water depth is set. (e.g., 30 cm) and standard deviation The formula for calculating the Sl value is as follows: ; in, Indicates the first Suitability index of each environmental factor; This is the measured value of the factor; This represents the ecological optimum value for this species on this factor. The parameter used to control the width of the suitability curve. When the measured water depth... Equal to the optimal value hour, =1; with Deviation , The value decays in a normal distribution.

[0066] For limiting factors that exhibit a monotonically increasing or decreasing trend (such as distance from the source of human interference) The SI curve is constructed using either logistic or piecewise linear form. A minimum tolerance distance is set. With a completely safe distance .when hour, ;when hour, ;when When the distance is between the two, the Sl value increases monotonically with increasing distance.

[0067] Step S4 also includes determining the weights of each environmental factor in the comprehensive evaluation system. An analytic hierarchy process (AHP) is used to construct a judgment matrix. Ornithologists are invited or existing literature data is used to conduct pairwise comparisons and scoring of the relative importance of factors such as water depth, vegetation cover, food richness, and disturbance distance (using a 1-9 scale). A comparison matrix is ​​then constructed. And solve for its largest eigenvalue. The corresponding eigenvectors. After normalizing the eigenvectors, the weight vectors of each factor are obtained. ,in To ensure the logical rationality of the weight allocation, a consistency check needs to be performed on the judgment matrix, and the consistency ratio needs to be calculated. Only when If the weight allocation is valid, then the pairwise comparison judgment matrix is ​​determined to be valid; otherwise, the pairwise comparison judgment matrix needs to be readjusted.

[0068] Step S4 further includes constructing a mathematical model for the Integrated Habitat Suitability Index (HSI). To reflect the weakest link effect among ecological factors (i.e., Liebig's law of minimum factors), meaning that the absence of any key factor (such as lack of water or strong disturbance) will render the habitat completely unusable, this embodiment uses a weighted geometric mean method as the integrated evaluation model, rather than a simple weighted arithmetic mean method. The formula for calculating HSI is as follows: ; in, This is the comprehensive suitability index for grid cells; Indicates from arrive The product (Product) is the product of the weighted similarity indices of all samples. The total number of environmental factors participating in the evaluation; For the first Single-factor suitability index for each factor; For the first The model assigns weights to each factor. This model ensures that when any one of the key limiting factors... When it approaches 0, the overall The value also approaches 0, accurately reflecting the limiting characteristics of the ecosystem.

[0069] Step S4 concludes with spatial calculations and hierarchical mapping based on a GIS platform. Using the raster calculator function of the GIS, standardized environmental factor layers are substituted into the aforementioned HSI model for pixel-by-pixel calculation, generating a raster map of the HSI value distribution across the entire study area. Based on the HSI value distribution characteristics, habitat suitability is divided into four levels using either the natural breakpoint method or the fixed threshold method: unsuitable habitat (…). ), low suitable habitat ( Suitable habitats ( ) ) and highly suitable habitats The final output is a "Spatial Distribution Map of Habitat Suitability" for each target waterbird, which visually shows the potential core area, marginal area, and ecological gap area of ​​the species in the study area, providing a spatial positioning basis for subsequent habitat restoration.

[0070] The method of the present invention includes the following steps in step S5: based on the HSI spatial distribution map and each single-factor suitability layer generated in step S4, using a limiting factor diagnostic algorithm to identify the dominant factors leading to habitat quality decline, and accordingly formulating precise restoration and protection strategies based on hierarchical zoning. Omitted. Step S5 specifically includes constructing a grid-scale diagnostic model for habitat limiting factors. For any habitat within the study area that is determined to be unsuitable or poorly suitable (i.e., For a given raster cell, the system automatically retrieves all single-factor suitability index values ​​corresponding to that cell. According to Liebig's law of minimum factors, the single factor with the smallest value is determined as the absolute constraint factor of that raster cell. If weighted effects exist, the weighted contribution loss value of each factor is calculated. Select The factor corresponding to the maximum value is used as the dominant limiting factor.

[0071] Based on the diagnosed dominant limiting factor type, targeted habitat restoration engineering solutions are generated: when the dominant limiting factor is diagnosed as unsuitable water depth, micro-topographic modification and water level control strategies are implemented. For the foraging needs of wading birds such as Great Egrets and Grey Herons, if the water depth is diagnosed as excessive (>50 cm), the solution is designed to construct shallow waters or implement bottom elevation projects to reshape the water depth to a shallow habitat of 10-30 cm. For the activity needs of waterfowl such as Ruddy Shelducks and Coots, if the water area is diagnosed as dry or insufficiently deep, the solution is designed to implement ecological water replenishment or dredging projects to connect waterways. Simultaneously, a composite topographic structure of ecological islands in deep-water and shallow-water areas is constructed. Using bulldozers and other machinery, ecological islands are built in open water areas, providing waterbirds with roosting and breeding grounds away from terrestrial predators.

[0072] When the dominant limiting factor is diagnosed as food scarcity, biomanipulation and habitat enhancement strategies are implemented. If drone monitoring data shows that the height of herbaceous vegetation exceeds 50 cm or the shrub cover is too high, causing the great bustard's vision to be obstructed or the mating grounds to be abandoned, the plan is to implement mechanical grass cutting or shrub removal measures to control the vegetation height to 15-30 cm and maintain an open field of vision. In response to the nesting needs of coots and ruddy shelducks, if the vegetation cover on the waterfront is insufficient, the plan is to plant emergent plants such as reeds and cattails on the shoreline to construct a ring-shaped concealment zone with a width of not less than 10 meters to provide shelter for nest sites.

[0073] When the dominant limiting factor is diagnosed as food scarcity, biomanipulation and habitat enhancement strategies are implemented. For piscivorous birds (grey herons, great egrets), native fish fry and benthic snails are introduced into the water to construct a multi-trophic food web; for herbivorous or omnivorous birds (great bustards, ruddy shelducks), some unharvested crops (such as winter wheat and corn) are retained around the habitat as supplementary food sources, or legume forage is artificially sown.

[0074] When the dominant constraint factor is diagnosed as excessive human interference, a spatial isolation and buffer control strategy is implemented. The minimum safe distance calculated in step S2 is used. Centered on highly suitable habitat patches, establish outwards a radius greater than [missing information]. Physical barriers should be established. Physical fences or ditches should be erected along the boundaries of these barriers to prevent vehicles and livestock from entering. Simultaneously, based on acoustic propagation models, ecological sound barriers should be installed or tall tree belts planted in areas adjacent to highways or high-voltage lines to reduce noise and visual disturbance.

[0075] Step S5 also includes dynamic protection zoning based on the HSI spatial distribution map. The area has been designated as a core protected area, within which all unauthorized human entry and low-altitude drone flights are prohibited, with only long-range monitoring using telephoto equipment permitted. Furthermore, areas with good connectivity are designated as ecological restoration zones, which are the key target areas for the aforementioned restoration projects; However, areas located on key ecological corridors are designated as buffer zones, where low-intensity land use control is implemented.

[0076] Finally, using the drone monitoring technology from step S1, data is collected again at specific time intervals after the restoration measures are implemented (such as after a breeding season), and the HSI value is recalculated. By comparing the changes in HSI before and after restoration, the ecological performance of the restoration project is quantitatively assessed, forming a closed-loop adaptive management system from monitoring, evaluation, diagnosis, restoration to remonitoring.

Claims

1. A method for adaptively optimizing UAV flight parameters to survey waterbird populations, characterized in that, Includes the following steps: Step 1: Conduct a baseline survey of the target waterbird species within the target area, collect multi-angle aerial images using drones, and construct a multi-dimensional aerial image interpretation marker library through feature extraction, which will be used as a comparison benchmark for subsequent image interpretation. Step 2: Conduct a correlation test between UAV flight parameters and waterbird disturbance response, construct a multi-factor test matrix composed of flight altitude, flight speed and approach method, and calculate the slant distance between the UAV and the target waterbird by acquiring coordinate data in real time, and obtain the disturbance response data of the target waterbird under different flight parameters; The approach method is set to horizontal approach mode, that is, the drone maintains a horizontal attitude at a predetermined altitude and flies in a straight line from the periphery of the target waterbird toward the target directly above it; Step 3: Based on the disturbance response data, establish a quantitative four-level behavioral response spectrum and record the response level of the target waterbird as the slant distance changes; Step 4: Construct a startled flight distance probability model based on the four-level behavioral response spectrum, calculate the minimum safe monitoring distance threshold of the target waterbird by analyzing the startled flight distance probability model, and determine the optimal combination of monitoring flight parameters that simultaneously meets the safety and accuracy constraints by combining the population statistical accuracy verification results and environmental meteorological factors. Step 5: Perform UAV aerial photography mission using the optimal combination of monitoring flight parameters, and interpret the acquired images using the multi-dimensional aerial image interpretation tag library to obtain waterbird population size and distribution data; The waterbird population survey method specifically includes the following steps for extracting habitat environmental factors from UAV remote sensing data: A 3D dense point cloud model is constructed using the structure-reconstruction-motion algorithm, generating a digital surface model and a digital elevation model; By calculating the difference between the digital surface model and the digital elevation model, a normalized digital surface model is obtained, and then the vegetation height factor is extracted. Vegetation coverage factors were extracted using the visible light atmospheric impedance vegetation index or the supergreen index. Based on ground survey data, spatial interpolation and range standardization were performed on water depth, food abundance, and distance from human disturbance sources.

2. The method for adaptive optimization of UAV flight parameters for waterbird population surveys according to claim 1, characterized in that, In step 1, the construction of the multi-dimensional aerial image interpretation marker library specifically includes extracting the geometric morphological features, spectral color features, texture features, and shadow features of the target waterbird; The geometric morphological features include a shape index, which is calculated as follows: by calculating the square of the perimeter of the edge outline of the target waterbird on the image, and by calculating the quotient of the square value multiplied by the product of the projected pixel area of ​​the target waterbird on the image and four times pi, the shape index is obtained; the shape index is used to distinguish between waterfowl and wading birds.

3. The method for adaptive optimization of UAV flight parameters for waterbird population surveys according to claim 1, characterized in that, In step 2, the multi-factor test matrix includes five discrete gradients of flight altitude from 30 meters to 120 meters and three discrete gradients of flight speed from 3 meters per second to 8 meters per second. The slant distance is calculated as follows: The three-dimensional coordinates of the UAV and the target waterbird in the spatial rectangular coordinate system are obtained. The slant distance is obtained by calculating the sum of the squares of the coordinate differences between the two in the three coordinate axes and by taking the square root of the sum of squares.

4. The method for adaptive optimization of UAV flight parameters for waterbird population surveys according to claim 1, characterized in that, In step 3, the fourth-order behavioral response spectrum is specifically defined as follows: Level 0 response is a no-response state; the target waterbird continues its original foraging or resting behavior, and its head does not point towards the drone. Level 1 response is an alert state; the target waterbird quickly raises its head and remains still, its gaze follows the drone's movement, or it stops paddling. Level 2 response is an avoidance state. The target waterbirds exhibit displacement avoidance behavior, waterfowl change their swimming path and move away from the drone, and wading birds or land birds show quick running or leaning forward in preparation for takeoff. Level 3 reaction is a startled take-off state, in which the target waterbird flaps its wings to completely leave the ground or water surface and engages in powered flight.

5. The method for adaptive optimization of UAV flight parameters for waterbird population surveys according to claim 4, characterized in that, In step 4, the construction of the startled flight distance probability model is based on the binomial logistic regression model, which sets the level 3 response in the four-level behavioral response spectrum as startled flight and sets the level 0 to level 2 responses as no startled flight. The minimum safe monitoring distance threshold is calculated as follows: The minimum safe monitoring distance threshold is obtained by calculating the ratio of the difference obtained by subtracting a preset acceptable disturbance probability threshold from the acceptable disturbance probability threshold, taking the natural logarithm of the ratio, subtracting the regression intercept parameter from the natural logarithm, and then taking the quotient of the result of the natural logarithm minus the regression intercept parameter and the regression slope parameter. The square root of the sum of the squares of the flight altitude and the horizontal distance in the optimal combination of monitoring flight parameters must be greater than or equal to the minimum safe monitoring distance threshold.

6. The method for adaptive optimization of UAV flight parameters for waterbird population surveys according to claim 1, characterized in that, In step 4, determining the optimal combination of monitoring flight parameters further includes the following steps: The statistical accuracy of UAV monitoring data was verified by ground-based synchronous observation, and the accuracy of population statistics was calculated by comparative analysis. Determine the maximum permissible flight altitude and maximum permissible flight speed that meet the population statistics accuracy rate of 90% or higher. Determine the effective range of solar elevation angle that maximizes image background contrast and minimizes reflectivity, and the wind speed threshold. The optimal combination of monitoring flight parameters is selected such that it falls within the safe range determined by the minimum safe monitoring distance threshold, the high-precision range determined by the maximum allowable flight altitude and the maximum allowable flight speed, and the suitable environmental range determined by the effective range of the solar altitude angle and the wind speed threshold.

7. The method for adaptive optimization of UAV flight parameters for waterbird population surveys according to claim 1, characterized in that, The specific steps involved in constructing a habitat suitability evaluation system using the waterbird population survey method include: A single-factor suitability index evaluation model is constructed for each of the aforementioned environmental factors, and the weight of each environmental factor is determined by the analytic hierarchy process. The weighted geometric mean method is used to calculate the comprehensive habitat suitability index, specifically as follows: By using the weights of the environmental factors as indices, the single-factor suitability index of each environmental factor is exponentially calculated, and the comprehensive habitat suitability index is obtained by multiplying the exponential results of all environmental factors. Based on the numerical distribution characteristics of the comprehensive habitat suitability index, habitats are divided into four levels: unsuitable, low-suitable, moderately suitable, and highly suitable.

8. The method for adaptive optimization of UAV flight parameters for waterbird population surveys according to claim 7, characterized in that, The waterbird population survey method generates habitat restoration strategies based on limiting factor diagnosis in the following ways: For areas where the comprehensive habitat suitability index is less than 0.5, the single-factor suitability index with the smallest value is identified as the dominant limiting factor. When the dominant limiting factor is unsuitable water depth, generate micro-geomorphic modification or water level regulation strategies. When the dominant limiting factor is unsuitable vegetation structure, generate strategies for mechanical mowing or replanting emergent plants. When the dominant limiting factor is food scarcity, strategies for biological manipulation or habitat enhancement are generated. When the dominant limiting factor is excessive human interference, a physical isolation zone is established based on the minimum safety monitoring distance threshold.

9. A method for adaptive optimization of UAV flight parameters for waterbird population surveys according to claim 8, characterized in that, The specific steps for delineating dynamic conservation zones using the waterbird population survey method include: Areas with a comprehensive habitat suitability index greater than or equal to 0.8 will be designated as core protected areas, and low-altitude drone flights will be prohibited. Areas with a comprehensive habitat suitability index greater than or equal to 0.5 and less than 0.8, and whose connectivity meets a preset threshold, are designated as ecological restoration zones and serve as target areas for implementing the restoration strategies. Areas with a comprehensive habitat suitability index of less than 0.5 and located on ecological corridors are designated as buffer zones. After implementing the restoration measures, steps 1 to 5 are performed again to assess ecological performance by comparing the changes in the comprehensive habitat suitability index before and after restoration.

Images