Unmanned aerial vehicle inspection route planning method and system

Abstract

The invention relates to the technical field of routing inspection route planning, in particular to an unmanned aerial vehicle routing inspection route planning method and system. The method comprises the following steps: acquiring environment real-time data, including images and wind regime states, and performing time sequence synchronization processing to obtain environment state and wind regime synchronization data; thirdly, marking the obstacle form and density in the synchronous data, performing disordered trajectory prediction on the obstacle form and density, adjusting the obstacle avoidance path attitude of the unmanned aerial vehicle, and generating corresponding path attitude data; and finally, re-planning an inspection route according to the unmanned aerial vehicle obstacle avoidance path attitude data, and sending re-planning data to the unmanned aerial vehicle control terminal to execute a new inspection route. According to the invention, the routing inspection route planning technology is optimized, so that the routing inspection route planning technology is more perfect.

Description

Technical Field

[0001] This invention relates to the field of inspection route planning technology, and in particular to a method and system for planning inspection routes using unmanned aerial vehicles (UAVs). Background Technology

[0002] With the rapid development of drone technology, drones are being used more and more widely across various industries, especially in power line inspection. Compared to traditional manual inspection, drone inspection offers many advantages, such as high efficiency, flexibility, access to hard-to-reach areas, and reduced human risk. Previous drone inspection path planning methods often relied on pre-set paths or static environmental information. However, in dynamic environments, such as those affected by wind, climate change, and random obstacles, these traditional methods often struggle to cope with complex environmental changes. Particularly in areas with obstacles, drones may collide or deviate from their targets due to improper path planning, preventing the inspection mission from being completed successfully. Therefore, designing paths that can both avoid obstacles and ensure the successful completion of inspection tasks in dynamic and uncertain environments has become a major challenge in drone inspection technology.

[0003] In summary, traditional UAV inspection route planning methods suffer from inaccurate obstacle avoidance paths and attitude adjustments encountered during inspections, resulting in large errors in route planning. Summary of the Invention

[0004] Therefore, it is necessary to provide a method and system for planning unmanned aerial vehicle (UAV) inspection routes to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a method for planning unmanned aerial vehicle (UAV) inspection routes is provided, the method comprising the following steps: Step S1: Collect real-time environmental data during the inspection route, including real-time environmental images and real-time environmental wind conditions; perform time-series synchronization processing based on the real-time environmental wind conditions and real-time environmental images to obtain synchronized environmental status and wind conditions data. Step S2: Mark the shape and density of obstacles in the environmental status and wind condition synchronization data; predict the disordered trajectory of the obstacle shape and density, and then adjust the attitude of the UAV obstacle avoidance path to generate UAV obstacle avoidance path attitude data. Step S3: Perform inspection route replanning based on the UAV obstacle avoidance path attitude data to obtain inspection route replanning data; send the inspection route replanning data to the UAV control terminal to execute inspection route planning.

[0006] The present invention also provides a drone inspection route planning system for executing the drone inspection route planning method described above. The drone inspection route planning system includes: The environmental acquisition module is used to collect real-time environmental data during the inspection route. The real-time environmental data includes real-time environmental images and real-time environmental wind conditions. The module performs time-series synchronization processing based on the real-time environmental wind conditions and real-time environmental images to obtain synchronized environmental status and wind conditions data. The obstacle avoidance path attitude adjustment module is used to mark the shape and density of obstacles in the environmental status and wind condition synchronization data; perform disordered trajectory prediction on the shape and density of the obstacles, and then perform UAV obstacle avoidance path attitude adjustment to generate UAV obstacle avoidance path attitude data. The inspection route replanning module is used to replan the inspection route based on the UAV obstacle avoidance path attitude data to obtain the inspection route replanning data; the inspection route replanning data is then sent to the UAV control terminal to execute the inspection route planning.

[0007] The beneficial effects of this invention lie in the fact that, during the inspection process, real-time environmental data, including environmental images and wind condition information, is collected by sensors mounted on the UAV. This process ensures real-time environmental monitoring, providing accurate image data of the surrounding environment and wind condition information such as wind speed and direction. By performing time-series synchronization processing on this data, the timeliness and consistency of image data and wind condition data are ensured, providing real-time and accurate environmental condition data for subsequent path planning. This step is the foundation of the entire process, ensuring the accuracy of subsequent analysis. Based on step S1, the shape and density of obstacles in the synchronized environmental condition and wind condition data are marked, and disordered trajectory prediction is performed on these obstacles. This step, by predicting obstacles and their trajectories, can anticipate potential risk points in advance and provide a basis for obstacle avoidance path planning. The UAV adjusts its path attitude in real time according to changes in obstacle shape and density, ensuring smooth obstacle avoidance during flight. This step complements the aforementioned synchronized data, accurately assessing the safety of the flight path and preventing UAV collisions. Based on the UAV obstacle avoidance path attitude data obtained in step S2, the inspection route is replanned. This process ensures that the UAV can flexibly adjust its inspection route in complex environments to avoid obstacles and maintain the effective execution of the inspection task. The replanned path data is sent to the UAV control terminal, guiding the UAV to perform inspections according to the new planned route. This step enables the UAV to cope with dynamically changing environments and effectively avoid obstacles, improving the efficiency and safety of the inspection task. Step S3 is the execution phase of the entire system, combining the data and prediction results from the first two steps to ensure that the UAV can autonomously adjust and execute the optimal inspection route in real-time. Therefore, this invention is an optimization of a traditional UAV inspection route planning method, solving the problem that traditional UAV inspection route planning methods suffer from inaccurate obstacle avoidance paths and attitude adjustments during inspections, resulting in large errors in inspection route planning. It improves the accuracy of obstacle avoidance paths and attitude adjustments, and reduces the error in inspection route planning. Attached Figure Description

[0008] Figure 1 A flowchart illustrating the steps of a drone inspection route planning method; Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S2. Figure 3 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3. Detailed Implementation

[0009] Please see Figures 1 to 3 A method for planning inspection routes using unmanned aerial vehicles (UAVs), the method comprising the following steps: Step S1: Collect real-time environmental data during the inspection route, including real-time environmental images and real-time environmental wind conditions; perform time-series synchronization processing based on the real-time environmental wind conditions and real-time environmental images to obtain synchronized environmental status and wind conditions data. Step S2: Mark the shape and density of obstacles in the environmental status and wind condition synchronization data; predict the disordered trajectory of the obstacle shape and density, and then adjust the attitude of the UAV obstacle avoidance path to generate UAV obstacle avoidance path attitude data. Step S3: Perform inspection route replanning based on the UAV obstacle avoidance path attitude data to obtain inspection route replanning data; send the inspection route replanning data to the UAV control terminal to execute inspection route planning.

[0010] In this embodiment of the invention, reference is made to Figure 1 The above is a flowchart illustrating the steps of a UAV inspection route planning method according to the present invention. In this example, the UAV inspection route planning method includes the following steps: Step S1: Collect real-time environmental data during the inspection route, including real-time environmental images and real-time environmental wind conditions; perform time-series synchronization processing based on the real-time environmental wind conditions and real-time environmental images to obtain synchronized environmental status and wind conditions data. In this embodiment of the invention, a high-resolution camera module and a high-sensitivity miniature wind speed sensor module are simultaneously installed on the main structure of the UAV. While the UAV flies along a predetermined inspection route, the camera module continuously acquires environmental image data at a sampling rate of 60 frames per second, with an image resolution of no less than 1920×1080 pixels. Simultaneously, the miniature wind speed sensor module continuously measures three indicators—wind speed, wind direction, and wind energy fluctuation intensity—at millisecond intervals. Each sample is assigned an identifier code precisely corresponding to the image sampling timestamp. All raw data is cached in parallel by an onboard data cache chip and synchronized using a timing matching program on the main control chip. This synchronization program utilizes a dual-channel timestamp alignment algorithm to establish a one-to-one mapping between the image sampling frame sequence and the wind condition data sequence on the time axis. After matching, the time difference between the two types of data is controlled to within 10 milliseconds using a linear interpolation correction algorithm. Subsequently, the system employs a grayscale histogram equalization method to adaptively optimize the image brightness distribution to reduce interference from wind, dust, shadows, and backlighting, thereby generating environmentally optimized image data with enhanced clarity. For the time-synchronized wind data, the processing unit runs a noise reduction algorithm based on moving average and median filtering to smooth out abnormal peak values ​​of instantaneous wind speed, thereby eliminating local disturbance signals. Finally, the environmental optimization image and the preprocessed wind information are fused together again along a unified time dimension to form complete environmental status and wind condition synchronization data, providing a data foundation for subsequent obstacle detection and prediction.

[0011] Step S2: Mark the shape and density of obstacles in the environmental status and wind condition synchronization data; predict the disordered trajectory of the obstacle shape and density, and then adjust the attitude of the UAV obstacle avoidance path to generate UAV obstacle avoidance path attitude data. In this embodiment of the invention, after obtaining synchronized data on environmental status and wind conditions, an embedded image recognition algorithm is used to automatically label obstacle regions in the environmental optimization image. This algorithm, based on a convolutional neural network feature map structure, combines an edge gradient extraction operator with a shape segmentation module to identify the shape regions of obstacles such as branches, pebbles, and fallen leaves, and calculates the obstacle quantity density using region area and pixel density. Subsequently, the obstacle morphological parameters and quantity density information are input into the dynamic target motion estimation unit in the trajectory prediction module. This unit first calls a disordered motion path statistical model based on a Kalman prediction structure to dynamically analyze the obstacle's motion trend by calculating the probability distribution of the change in the centroid position of the obstacle in adjacent time frames. Then, combining wind speed and direction information, a vector superposition analysis method is used to calculate the rate of change of the obstacle's motion direction under wind force, obtaining the obstacle trajectory change sequence. Based on this, obstacle trajectory clustering analysis is performed, using Euclidean distance metric and a shortest path tree merging algorithm to classify a large number of local trajectory segments, aggregating obstacle motion trajectories with similar dynamic characteristics into clusters, and outputting disordered trajectory clustering features. Next, the path adjustment control module generates three-axis attitude control signals for the UAV in real time based on the characteristic parameters and the environmental wind distribution. The pitch angle, yaw angle and roll angle are dynamically and slightly adjusted by the angular rate adjustment output by the controller, thereby completing the UAV obstacle avoidance action adaptation and generating UAV obstacle avoidance path attitude data in real time.

[0012] Step S3: Perform inspection route replanning based on the UAV obstacle avoidance path attitude data to obtain inspection route replanning data; send the inspection route replanning data to the UAV control terminal to execute inspection route planning.

[0013] In this embodiment of the invention, the control system performs a replanning operation on the inspection route based on the UAV obstacle avoidance path attitude data. First, the main control module extracts the current spatial coordinate data of the UAV, including three sets of parameters: latitude, longitude, altitude, and attitude angle. This data is then fused with the corrected attitude data after obstacle avoidance to form a set of spatial coordinates after obstacle avoidance. After obtaining this set, the system calls the built-in inspection task database to read the originally set preset inspection path, which consists of several continuous coordinate points arranged in task order. Subsequently, the replanning algorithm module calculates the spatial deviation vector between the post-obstacle-avoidance spatial coordinates and the corresponding points on the preset path. By comparing the vector magnitudes, the positions of segments where the deviation exceeds a set threshold are determined, and these segments are divided into replanning intervals. For each replanning interval, a path topology model is established using node connectivity analysis. Furthermore, the shortest path solution combined with spatial connectivity verification is applied to evaluate the path length and stability of different feasible road segments, selecting the optimal path. The entire replanned path consists of multiple continuous attitude control points, with the intervals between points indexed by a fixed time step. After path calculation is completed, the system formats the data into a UAV route control command set, including attitude angles, displacement increments, velocity scalars, and direction vectors for each control cycle, and then transmits it to the UAV flight control terminal via a wireless communication link. Upon receiving the data, the flight control terminal parses and executes it sequentially, achieving dynamic replanning control of the inspection route and establishing a new flight path after obstacle avoidance to complete the mission.

[0014] This invention collects environmental data using a high-definition camera and a high-sensitivity wind speed sensor, and performs time-series synchronous processing to ensure accurate matching between images and wind condition data. Next, through image recognition and trajectory prediction, it marks the shape and density of obstacles, and combines this with wind speed data to predict obstacle movement trajectories, providing a dynamic control basis for obstacle avoidance. Finally, it replans the inspection route based on obstacle avoidance path data, ensuring that the drone can avoid obstacles and fly flexibly within the predetermined mission area. This series of interconnected technologies not only improves the inspection efficiency of the drone but also ensures stability and accuracy in complex environments.

[0015] Step S1 includes the following steps: Step S11: Collect real-time environmental data during the inspection route using the camera and miniature anemometer mounted on the drone. The real-time environmental data includes real-time environmental images and real-time wind conditions. Step S12: Perform data preprocessing on the real-time environmental wind speed status to obtain wind condition preprocessing data; Step S13: Enhance the contrast of the real-time environmental image to obtain an optimized real-time environmental image; Step S14: Perform time-series synchronization processing based on the preprocessed wind condition data and the real-time environmental optimization image to obtain synchronized environmental condition and wind condition data.

[0016] In this embodiment of the invention, a perception and recognition unit, including a high-resolution camera module and a three-dimensional ultrasonic anemometer, is fixedly installed on the main body of the UAV. The camera module acquires real-time environmental images of the space in front of and to both sides of the inspection channel at a rate of 60 frames per second, with an image resolution of 1920×1080 pixels. The three-dimensional ultrasonic anemometer acquires wind speed vector, wind direction, and turbulence intensity data of the current spatial point at a millisecond clock frequency, and automatically adds a timestamp number to each acquisition result. The flight control motherboard stores the images and wind speed information in a synchronization buffer, and establishes an initial correspondence between the two types of data through a timestamp index, forming a raw dataset of real-time environmental images and real-time environmental wind conditions.

[0017] After the wind condition data arrives at the processing unit, an internal noise preprocessing procedure is executed. First, median filtering is used to eliminate high-frequency spikes in the wind speed measurement, and the median value is taken for every 10 sets of sampled data to smooth out interference. Second, the moving average method is used to calculate the transient wind speed change trend, completing the stabilization of the wind speed fluctuation sequence. Then, Fourier transform is used to decompose the low-frequency principal components and remove the high-frequency random disturbances, thereby obtaining more continuous wind condition preprocessed data.

[0018] The grayscale histogram of the real-time environmental image is redistributed, and a local contrast enhancement algorithm is used to improve the brightness of dark areas. During enhancement, the image is divided into fixed-size regions, and the mean and variance of brightness for each region are calculated. Then, the grayscale distribution of each region is adjusted through mapping to improve overall image contrast while maintaining global brightness balance. The enhanced image output is a real-time optimized environmental image, providing clearer environmental feature boundaries for obstacle recognition.

[0019] The synchronization processing unit first reads the timestamp information of the preprocessed wind condition data and the real-time environmental optimization image, and then executes a linear interpolation synchronization algorithm using the time difference calculation module. For sampling points with misaligned time intervals, the interpolation algorithm estimates the interpolated data based on the time difference between adjacent frames, ensuring that the two types of data are aligned on the same time axis. Subsequently, a time series fusion strategy is adopted to encapsulate each image frame and the corresponding wind speed vector into a synchronization data packet, generating synchronized environmental condition and wind condition data, providing an accurate spatiotemporal correspondence basis for subsequent obstacle detection and path prediction.

[0020] Step S2 includes the following steps: Step S21: Mark the shape and density of obstacles in the environmental status and wind condition synchronization data; wherein the obstacles include tree branches, fallen leaves and gravel; Step S22: Based on the synchronized data of environmental conditions and wind conditions, perform disordered trajectory prediction on the shape and density of the obstacles to obtain disordered trajectory prediction data for the obstacles; Step S23: Perform cluster feature analysis on the disordered trajectory prediction data of obstacles and output the disordered trajectory cluster prediction features; Step S24: Based on the disordered trajectory clustering prediction features and the synchronized data of environmental status and wind conditions, adjust the UAV obstacle avoidance path attitude to generate UAV obstacle avoidance path attitude data.

[0021] As an example of the present invention, reference is made to Figure 2 As shown, in this example, step S2 includes: Step S21: Mark the shape and density of obstacles in the environmental status and wind condition synchronization data; wherein the obstacles include tree branches, fallen leaves and gravel; In this embodiment of the invention, after the environmental status and wind conditions are synchronized and input, the scene image is first divided into zones for detection using a spatial gridding method. Each grid cell is analyzed independently with a unit size of 50×50 pixels. The system uses a feature extraction algorithm based on edge gradient and texture direction analysis to identify obstacle regions. Abnormal geometric regions are determined by detecting gray-scale abrupt changes in edges, texture repetition cycles, and color deviations. In the detected regions, a contour correction algorithm based on morphological opening and closing operations is used to separate the contour boundaries of branches, fallen leaves, and gravel. The morphological judgment criteria mainly include the rate of change of boundary curvature, texture direction consistency, and pixel cluster coverage ratio. These three factors are combined using weighted coefficients to form an obstacle morphological judgment matrix. The obstacle number density is statistically analyzed using the pixel count method per unit area. The number of obstacle pixels in each zone is calculated as a percentage of the total number of pixels, and their center position in the image coordinates is recorded. The results are written into an obstacle labeling data table in the form of digital labels. The labeling data table shares a time index with the wind condition measurement time series data, realizing the synchronization of obstacle distribution data and real-time wind field parameters, providing input conditions for subsequent trajectory prediction algorithms.

[0022] Step S22: Based on the synchronized data of environmental conditions and wind conditions, perform disordered trajectory prediction on the shape and density of the obstacles to obtain disordered trajectory prediction data for the obstacles; In this embodiment of the invention, after the obstacle morphology and density data are generated, the disordered motion trajectory of the obstacles is calculated. The prediction process uses temporal difference analysis to calculate the geometric center drift direction of each obstacle in continuous time frames. The system simultaneously calls the wind speed and wind direction components from the wind condition synchronization data to logically map the force direction of each obstacle under different wind fields, obtaining a wind force influence direction distribution table. A wind force disturbance effect model is established using the wind speed change rate and the obstacle object area projection coefficient. The obstacle morphology feature vector is input into the disturbance model to generate a random direction step sequence. Each step state in the sequence reflects the displacement direction trend of the obstacle at a certain time interval. The system summarizes the direction vector trajectory of each sequence to form a trajectory scatter set through time recursion and performs density evaluation on the scatter distribution. The density evaluation results are used to determine the uncertainty distribution range of obstacle movement. After the calculation is completed, disordered obstacle trajectory prediction data is generated, which includes three core parameters: obstacle position sequence, direction change frequency, and spatial drift intensity.

[0023] Step S23: Perform cluster feature analysis on the disordered trajectory prediction data of obstacles and output the disordered trajectory cluster prediction features; In this embodiment of the invention, spatial feature vector standardization is first performed to normalize the three-dimensional coordinates, direction change frequency, and drift intensity of each obstacle trajectory point to the same numerical range. After normalization, a density-connected clustering algorithm is used to classify the trajectory samples. By calculating the Euclidean distance, direction angle difference, and drift amplitude difference between adjacent trajectory points, a distance matrix and a direction matrix are formed. The clustering algorithm scans local high-density regions in multi-dimensional space and automatically groups continuous trajectory points belonging to the same dynamic feature into the same class. To prevent erroneous clustering caused by small sample drift points, the system sets a minimum sample size threshold and a maximum acceptable direction difference threshold during the clustering process, and outputs unordered trajectory clustering prediction features under the constraints. These features include trajectory type classification labels, cluster center coordinates, and statistical information on the direction deviation of each cluster, providing a basic mathematical description for the dynamic adjustment of the UAV's flight attitude.

[0024] Step S24: Based on the disordered trajectory clustering prediction features and the synchronized data of environmental status and wind conditions, adjust the UAV obstacle avoidance path attitude to generate UAV obstacle avoidance path attitude data.

[0025] In this embodiment of the invention, after acquiring the disordered trajectory clustering prediction features and synchronized data on environmental conditions and wind conditions, the obstacle avoidance path of the UAV is dynamically corrected. First, the spatial distribution range and motion trend of each obstacle cluster in the clustering features are analyzed, and an obstacle threat sequence is established according to the spatial angle order from the current flight trajectory of the UAV. Then, wind speed vector and airflow shear data are used to calculate the direction of local air disturbances, and the lateral aerodynamic force distribution range experienced by the UAV is determined using the spatial vector superposition method. Based on the spatial order of the threat sequence, risk sections on the obstacle avoidance path are assessed, and corresponding angular velocity adjustment signals are generated using an attitude control algorithm. This algorithm adjusts the attitude deviation caused by aerodynamic disturbances to within a defined range by analyzing the angle ranges of three-axis pitch, roll, and yaw. Finally, combining the wind direction correction amount at each coordinate point on the obstacle avoidance path with the obstacle clustering distribution information, the UAV flight attitude path data is reconstructed to ensure that the dynamic response of the attitude angles accurately corresponds to the spatial curve of the obstacle avoidance path, ultimately outputting the UAV obstacle avoidance path attitude data.

[0026] This invention accurately predicts and adjusts the flight path of a drone by analyzing real-time wind data and obstacle features. First, the system analyzes the environmental conditions through spatial gridding, detecting and marking the shape and density of obstacles to provide foundational data for subsequent operations. Next, using the shape and density information of the obstacles, its possible trajectories are calculated, while wind disturbance analysis is performed on the trajectories. Then, through cluster analysis of these trajectory data, the dynamic features of the obstacles are extracted, providing focused information for obstacle avoidance path planning. Finally, based on the clustering features and synchronized wind data, the drone's flight attitude and path are dynamically adjusted to ensure precise coordination between flight attitude and obstacle avoidance path. Through this series of processes, the drone can respond to complex environmental changes in real time, achieving efficient and safe obstacle avoidance path adjustments, significantly improving flight stability and safety.

[0027] Step S22 includes the following steps: Step S221: Estimate the mass of the obstacle based on its shape and quantity density; extract wind direction and wind force from the synchronized environmental and wind condition data; Step S222: Based on the wind direction and wind force, predict the frequency of directional changes in the mass and quantity density of the obstacle to obtain the directional changes of the obstacle; Step S223: Based on the change in obstacle direction, the mass and quantity density of obstacle objects, perform sparse feature analysis of dwelling points to obtain the sparse features of obstacle dwelling points; wherein the sparse features of dwelling points refer to the degree of sparseness of the distribution of the short-term dwelling positions of the obstacle in different directions during the process of the obstacle changing direction under the action of wind force. Step S224: Calculate the orientation tendency shift probability based on the sparse characteristics of obstacle dwell points to obtain the orientation tendency shift probability; perform trajectory entropy evolution deduction based on the obstacle orientation change frequency and orientation tendency shift frequency to obtain trajectory entropy evolution data. Step S225: Based on the trajectory entropy evolution data and the orientation tendency transfer probability, perform disordered trajectory prediction to obtain disordered trajectory prediction data for obstacles.

[0028] In this embodiment of the invention, after obtaining obstacle morphology and quantity density data, the system first estimates the object mass using the obstacle's area distribution, edge thickness information, and grayscale intensity. Image data captured by the camera is converted into a grayscale matrix, where grayscale values ​​represent the obstacle's light reflectivity. The processing unit extracts the pixel percentage and corresponding height layering data of the obstacle using a block-based method, and calculates the obstacle's volume using a volume estimation method. A correspondence is established between the volume result and the quantity density parameters corresponding to the obstacle type; for example, branches, fallen leaves, and gravel are associated with different quantity densities. The system multiplies the two to obtain the obstacle object mass estimation matrix. After mass estimation, real-time wind direction and wind force are extracted from synchronized environmental and wind condition data. These two types of data are input in vector form, and a wind force vector group is constructed using wind direction angle and wind speed scalars. The system combines and stores the obstacle mass estimation matrix and the wind force vector group, preparing the input basis for subsequent direction change frequency prediction.

[0029] The system calculates the force response period for each obstacle group based on the wind speed change rate and wind direction shift angle. In the calculation process, the system establishes an action surface coefficient matrix based on the wind speed vector and the spatial distribution array of obstacles, and obtains the instantaneous force distribution through matrix multiplication and addition. For each time step, the force distribution data is weighted and merged to obtain the overall force direction update sequence for the obstacles. A sliding time window records the magnitude and duration of direction changes, thus forming obstacle direction change frequency data. This frequency data represents the time-series distribution of obstacle direction changes caused by disturbances under wind conditions. The system outputs the frequency data along with the obstacle number and wind direction data to form the obstacle direction change results, providing input conditions for the next step of sparse feature analysis of dwell points.

[0030] The sparse feature analysis module for stopping points first determines the directional inflection points of each obstacle in the time series based on the frequency data of obstacle direction changes. When an obstacle undergoes a significant direction change, its velocity briefly approaches zero; this instant is defined as a stopping point. The system performs local extremum detection on the direction sequences of all obstacles, recording the corresponding spatial coordinates at the detected extremum points. All stopping point data are aggregated according to time windows, and the set of stopping points within each window is projected onto a unified three-dimensional spatial distribution map. The spatial sparsity of stopping points is statistically determined by calculating the distance distribution density between different stopping points. This sparsity reflects the degree of scarcity of obstacle distribution during multiple directional changes. The system further compares and weights the obstacle mass and quantity density values ​​with the stopping point distribution density to extract the strength relationship of stopping point disturbance features for different obstacle categories under the same wind conditions. Finally, a sparse feature dataset of stopping points is output, which includes the spatial distribution density index, the number of stopping points, and the temporal distribution pattern for each type of obstacle, providing a basis for subsequent trajectory evolution and disordered motion prediction.

[0031] After acquiring the sparse features of obstacle dwell points, the system arranges the direction vectors of obstacles within continuous time segments according to time index based on the spatial coordinate sequence of these dwell points, forming a set of azimuth change paths. The system calculates the azimuth angle difference between adjacent dwell points and statistically analyzes the frequency distribution of each angle interval to construct an azimuth transfer statistics table. Subsequently, the system obtains a preliminary azimuth transfer ratio based on the ratio of the frequency of each azimuth angle change to the total number of transfers. To improve accuracy, the system further incorporates wind drift information from wind data, using the coupling degree between the wind direction change direction and the obstacle's movement direction as a weight to weight and correct the preliminary transfer ratio, reflecting the true tendency of obstacle azimuth changes under wind influence. The weighted ratios are normalized, and the sum of the values ​​is calibrated to 1, ultimately generating an azimuth tendency transfer probability matrix. Each row represents the current obstacle azimuth state, and each column represents the probability of it potentially transferring to a target azimuth state. The system then uses the obstacle azimuth change frequency data and the azimuth tendency transfer probability matrix as core inputs to perform trajectory entropy evolution and deduction. The implementation path of trajectory entropy evolution is as follows: First, the local information entropy value is calculated using the occurrence probability of each orientation state. Then, the overall uncertainty trend of the system is analyzed based on the time evolution rate of the direction change frequency. The system calculates the dynamic curve of entropy increase or decrease in the trajectory sequence through time accumulation. The change in the curve reflects the diffusion effect of the obstacle position and movement direction with the wind force. After all calculations are completed, trajectory entropy evolution data is output. The data includes three parameters: the entropy growth rate of each time interval, the state transition stability coefficient, and the direction diffusion interval, providing a quantitative basis for the final prediction of disordered trajectories.

[0032] After generating the trajectory entropy evolution data and the azimuth transition probability matrix, the system enters the disordered trajectory prediction stage. First, an obstacle state sequence set is established, and the azimuth transition results, sparse stop point data, direction change frequency, and trajectory entropy evolution curve of each obstacle from the previous stage are input into the comprehensive prediction module. Internally, the module iterates through and calculates the historical azimuth change probability of each obstacle, combining the entropy increase characteristics in the trajectory entropy evolution to identify the stable trend of the movement direction. When the local entropy value is increasing, the system diffuses the corresponding azimuth transition probability to adjacent azimuth intervals to simulate the disordered expansion caused by random disturbances; when the entropy value decreases, the transition probability is centralized, indicating that the obstacle's direction tends to stabilize. The system updates the azimuth vector sequence of all obstacles in a multi-step time-recursive manner, iteratively calculating the possible movement positions of obstacles at multiple future time nodes through time windows between continuous states. To control computational stability, the system sets a direction change threshold and a lower limit for the transition probability to prevent the spread of errors caused by data discrepancies. After integrating the direction and position output data of each obstacle at multiple time points, the system calculates the center prediction trajectory based on a weighted probability distribution and generates a spatial trajectory cloud map. The final output of disordered obstacle trajectory prediction data includes four indicators: obstacle prediction position coordinates, movement direction range, velocity change trend, and trajectory spread range, providing dynamic input reference for UAV obstacle avoidance path adjustment.

[0033] Specifically, the core logic of this process first estimates the mass of obstacles by analyzing their shape and density, and then extracts wind direction and force conditions from environmental data. Based on wind information, the system predicts the frequency of obstacle direction changes, thereby estimating their trend under wind influence. Next, by analyzing the sparse characteristics of obstacle dwell points, the system quantifies the sparseness of obstacle distribution in various directions during movement, revealing the characteristics of their motion patterns. Furthermore, combining the obstacle direction change frequency and wind conditions, the system analyzes the obstacle's direction change trend over a future period using an orientation shift probability model, providing a basis for subsequent trajectory evolution. Finally, the system uses this data to calculate the change in trajectory entropy, assessing the disordered movement trend of obstacles. Through the dynamic changes in entropy, the system can predict the possible positions and directions of movement of obstacles at different time points, ultimately generating accurate disordered trajectory prediction data. The entire process, through the combination of dynamic obstacle analysis and wind environment, achieves efficient prediction of obstacle movement in complex environments, ensuring the reliability and accuracy of path planning.

[0034] Step S24 includes the following steps: Step S241: Quantify the degree of wind shear fluctuation in the synchronous data of environmental conditions and wind conditions; calculate the wind speed shear rate increase index of the degree of wind shear fluctuation, that is, compare the growth rate of wind speed gradient in adjacent time periods. Step S242: Based on the wind shear rate increase index, perform a coupled analysis of wind direction shear angle increase on the degree of wind shear fluctuation to obtain coupled data of shear angle increase; Step S243: Measure the aerodynamic imbalance variance based on the wind speed shear rate increase index and shear angle increase coupling data; the aerodynamic imbalance variance refers to the degree of attitude fluctuation of the UAV under wind disturbance. Step S244: Adjust the UAV obstacle avoidance path attitude based on the aerodynamic imbalance variance and disordered trajectory clustering prediction features to generate UAV obstacle avoidance path attitude data.

[0035] In this embodiment of the invention, when processing synchronized data on environmental conditions and wind conditions, the system first reads all original sensor signal sequences containing wind speed, wind direction, and vertical airflow velocity. The wind shear fluctuation quantification module calculates the difference in wind speed between different height layers in the vertical direction layer by layer using a time-series difference calculation method. The wind speed difference between each pair of adjacent layers is processed using a mean smoothing algorithm to remove random noise, resulting in a wind speed gradient sequence. Subsequently, a time-averaged sliding analysis is performed on this gradient sequence to calculate the fluctuation amplitude of wind speed with height and time. After normalization adjustment, the fluctuation amplitude sequence forms a wind shear fluctuation index, the value of which reflects the stability of wind speed in the vertical direction. The calculation of the wind shear rate rise index is then carried out on this basis. By comparing the growth rate of the wind speed gradient in adjacent time periods, a first-order time difference accumulation algorithm is used to calculate the wind shear rate rise index. The generation of this index is completed by the linear fitting module, specifically by performing interval slope statistics on the changing trend of the wind speed gradient, encoding all results into a stable wind shear rise sequence to describe the degree of airflow disturbance enhancement, and providing dynamic parameters for subsequent angle coupling analysis.

[0036] After obtaining the wind shear rate increase index, the wind shear angle increase coupling analysis module begins operation. Based on the real-time wind direction angle sequence recorded in the environmental state data, the system calculates the wind direction angle difference between adjacent time periods and establishes a wind direction angle time-varying curve. The wind direction change data and the wind shear rate increase index form a coupled input sequence, and the correlation coefficient between the two over time is obtained through cross-correlation calculation. Subsequently, an angle offset superposition algorithm is used to multiply the correlation coefficient and the wind direction angle change amplitude item by item, forming a shear angle increase coupling matrix. This matrix reflects the amplification effect of the wind direction angle when the wind shear degree increases. The system further performs time sliding window analysis on the coupling matrix, extracts the peak angle response in local time periods, calculates the average and standard deviation of the coupling coefficient, and outputs stable shear angle increase coupling data. This data includes the extreme values ​​of wind direction angle offset, angle increase rate, and shear coupling degree that evolve over time, used to characterize the redistribution of airflow disturbances in spatial directions, providing reference input for subsequent aerodynamic imbalance variance measurement.

[0037] A statistical assessment of the stability of a UAV after being subjected to airflow disturbances is conducted. First, current flight status data of the UAV is collected, including four sets of dynamic parameters: speed, rate of climb, attitude angle, and angular velocity. The gradient of the airflow velocity vector change at each moment is calculated using the wind shear rate rise exponent, and the spatial distribution of the airflow direction is determined by combining the shear angle increase coupling data. The system performs vector multiplication of the wind speed vector with the UAV attitude angle matrix to obtain the lateral and rise components of the force, representing the disturbance forces on the roll and pitch axes, respectively. Subsequently, the fluctuation range of each component is analyzed through a statistical time window, and the variance of the torque change in each dimension is calculated. The calculation process uses a sliding statistical method to continuously update the calculation results, ensuring that it reflects the aerodynamic offset fluctuations generated by the aircraft under changes in wind shear intensity in real time. Finally, the aerodynamic imbalance variance is obtained through weighted synthesis. The data results include the variance of roll direction fluctuation, pitch direction fluctuation variance, and yaw direction coupling variance, providing quantitative parameters for the attitude adjustment control stage.

[0038] Based on aerodynamic imbalance variance and disordered trajectory clustering prediction features, the system adjusts the attitude of the UAV obstacle avoidance path. The spatial distribution center, direction of motion, and distribution radius information of obstacles are extracted from the disordered trajectory clustering prediction features and mapped uniformly to the spatial coordinate system corresponding to the flight path. The system then aligns the aerodynamic imbalance variance data with the path coordinates to determine the adjustment weight corresponding to the aerodynamic disturbance intensity in each attitude direction. Based on the weight allocation, the control law calculation module calculates the angular velocity adjustment values ​​of the UAV in pitch, roll, and yaw axes. The calculation process utilizes a three-axis coupled control algorithm, jointly inputting the lateral wind shear and obstacle orientation clustering results into the control law to determine the direction and magnitude of the attitude adjustment. The control signal is transmitted to the flight controller via the output interface, and the controller corrects the UAV's three-axis attitude in real time, ensuring the flight path bypasses the predicted obstacle area. After path adjustment, the system generates UAV obstacle avoidance path attitude data, including a path point sequence, angular velocity adjustment sequence, and attitude angle change trend, which guides the UAV in performing dynamic obstacle avoidance control.

[0039] Specifically, by quantifying the degree of wind shear fluctuation, the system calculates the changes in wind speed at different altitudes layer by layer, removes noise, and obtains a wind speed gradient sequence. Then, time-moving average analysis is used to determine the amplitude of wind speed fluctuations. Subsequently, the system combines the wind shear rate rise index to describe the degree of disturbance in the wind speed gradient over time, providing a dynamic parameter basis for subsequent analysis. Based on this, the system calculates the change in wind direction angle and its coupling relationship with the wind shear rate rise index, obtaining coupled data on the increase in shear angle, thus revealing the impact of airflow disturbance on wind direction changes. Next, combining the aerodynamic imbalance variance calculation process, the system evaluates the impact of aerodynamic disturbances caused by wind shear on flight stability through real-time analysis of the UAV's flight status. In this process, the combination of wind speed vector and attitude angle matrix allows for the precise quantification of aerodynamic imbalance fluctuations along different axes, providing a scientific basis for subsequent attitude adjustments. Finally, combining disordered trajectory clustering prediction features and aerodynamic imbalance variance, the system provides an optimized path for UAV attitude adjustment based on the spatial distribution of obstacles and the intensity of wind speed disturbances. The seamless integration of the entire process enables efficient linkage between wind speed disturbances and flight attitude adjustments, ensuring that the drone can fly stably in dynamic environments and successfully avoid obstacles.

[0040] The variance of the metering pneumatic imbalance includes: The current flight motion state of the UAV is obtained; based on the wind speed shear rate increase index and shear angle increase coupled data, the airflow disturbance nonlinear fitting of the spatial gradient vector field is performed to obtain the airflow spatial disturbance vector; Extract the elevator deflection angle and rudder deflection angle from the current flight motion state of the UAV; Based on the airflow space disturbance vector, the elevator deflection angle and rudder deflection angle are decomposed into a random process of spatial lateral force imbalance to obtain the UAV lateral force imbalance data. The torque nonlinear amplification effect is estimated by advancing the lateral force imbalance data of the UAV in the time dimension. The torque imbalance nonlinear amplification effect means that when the UAV is disturbed by the non-uniform wind field, its attitude force does not show a linear proportional response change, but rather a nonlinear enhancement state is generated under the combined effect of time advancement and spatial airflow gradient superposition. The aerodynamic imbalance variance is measured based on the nonlinear amplification effect of the torque imbalance.

[0041] In this embodiment of the invention, motion data such as flight acceleration, three-axis attitude angular velocity, pitch angle, roll angle, and altitude rate are collected in real time by the UAV flight control sensing unit, with a sampling period fixed at 10 milliseconds. The wind speed shear rate increase exponent and shear angle increase coupling data are synchronously input to the space airflow disturbance fitting module. The module establishes the gradient distribution of the airflow velocity field in a three-dimensional spatial coordinate system, using the wind speed change rate as the main input vector and the wind direction shear angle as the directional constraint vector. A piecewise polynomial fitting method is used to perform nonlinear interpolation on the space airflow gradient field. The interpolation process uses the wind speed change gradient as the independent variable and the airflow disturbance intensity as the dependent variable, calculating the nonlinear distribution value of the airflow disturbance at different spatial nodes according to a time series recursive method. The total effect of the airflow disturbance is obtained by gradient superposition within a continuous time window, outputting the airflow spatial disturbance vector. This vector contains three-axis vector components and their instantaneous change rates, reflecting the non-uniform intensity differences of the airflow in different directions, providing accurate input for subsequent control surface force decomposition.

[0042] The UAV flight control system synchronously outputs control surface deflection data during the speed measurement phase. Control surface sensors record the deflection angles and rates of change of the elevator and rudder in real time. The system calculates the relative azimuth angles of the control surfaces in the body coordinate system based on the sensor feedback data using a control surface geometric pose calculation model. To ensure data consistency, the system filters and smooths three consecutive time sampling points, eliminating high-frequency small-amplitude jitter caused by micro-vibrations, before inputting the corrected elevator and rudder deflection angles into the control surface dynamics analysis module. This module records the gradient of the control surface deflection angle at each time interval, providing a data foundation for subsequent spatial lateral force decomposition calculations of the control surface response.

[0043] Based on airflow spatial disturbance vector and control surface deflection angle data, the system calculates the force asymmetry of the UAV in a disturbed wind field through stochastic process decomposition of spatial lateral force imbalance. Using the body coordinate system as the calculation reference, the system performs triaxial decomposition of the airflow vector, calculating the inner product of the dynamic response of each axial aerodynamic disturbance component with the corresponding control surface deflection angle. The calculation result represents the actual force offset generated by the control surface after being excited by wind shear. The system continuously records multiple wind disturbance events within a time window, forming a lateral force variation sequence through a stochastic process superposition algorithm. Fluctuations in this sequence along the roll and yaw axes are considered lateral force imbalance information. After fitting the data with a probability distribution, UAV lateral force imbalance data is generated, including three types of parameters: average lateral force offset, peak range, and duration, recording the force imbalance characteristics of the UAV under the influence of the airflow gradient field.

[0044] When performing time-progressive analysis on lateral force imbalance data, the system runs a nonlinear torque propagation algorithm in a time-series coordinate system. First, the time correspondence between changes in UAV attitude angles and lateral force is calculated. The displacement vectors of forces on the roll and pitch axes are converted into torque signals using a spatial torque analysis module. The system performs exponential weighted cumulative calculations on each torque signal to obtain a torque growth curve over time. Due to the non-uniform spatial distribution of airflow disturbances, a dynamic amplification factor is set to reflect the nonlinear enhancement of torque in the high-speed shear region. This amplification calculation iteratively superimposes torque fluctuations in adjacent time slices to form a quantitative characteristic of torque amplification. All results constitute the nonlinear amplification effect data of torque imbalance, including three sets of quantitative results: the mean torque curve, fluctuation scale, and growth rate under time progression, used to describe the asymmetric torsional response of the UAV in continuous airflow shear. The nonlinear amplification effect of torque imbalance refers to the fact that when a UAV is disturbed by uneven wind fields, its attitude forces (especially the moments on the pitch, roll, and yaw axes) do not exhibit a linear proportional response. Instead, the analysis shows a nonlinear enhancement under the combined effects of time progression and the superposition of spatial airflow gradients. Specifically, when the wind speed shear intensity or wind direction angle change amplitude increases, the effect of local airflow disturbances on the control surfaces and airframe structure is amplified, causing the torque response curve to show an exponential or multiplicative trend. This effect reflects the complex coupling relationship between force and attitude in a dynamic aerodynamic environment, meaning that even small fluctuations in wind speed or angle can easily trigger significant attitude torque shifts and instability increases. Through time accumulation and weighted iterative analysis of lateral force imbalance data, this nonlinear enhancement phenomenon can be quantified, thus providing a precise basis for calculating aerodynamic imbalance variance and controlling flight attitude stability.

[0045] After generating data on the nonlinear amplification effect of torque imbalance, the system performs aerodynamic imbalance variance measurement. This step uses a statistical analysis module to perform aggregate variance calculations on torque fluctuation samples from multiple time periods, thereby quantifying the impact of aerodynamic disturbances on overall flight attitude stability. The system groups the torque variation curves along three axes, calculates the mean and deviation for each group, and then weights and synthesizes the variances in all directions, with weights defined based on the inertial distribution constant of the UAV's airframe structure. The calculation results form aerodynamic imbalance variance data, including roll variance, pitch variance, yaw variance, and overall aerodynamic variance index. This data characterizes the degree of attitude imbalance caused by airflow disturbances during flight, providing quantitative input for the obstacle avoidance path attitude adjustment module and enabling real-time and accurate correction of UAV attitude control.

[0046] Adjusting the attitude of the drone's obstacle avoidance path includes: Based on the disordered trajectory clustering prediction features, obstacle avoidance paths are selected in the local space to obtain local space obstacle avoidance paths. The obstacle avoidance speed requirement is obtained by solving the obstacle avoidance speed requirement in the local space of the disordered trajectory clustering prediction features. The dynamic adjustment range of the three-axis attitude angles of the UAV is evaluated based on the aerodynamic imbalance variance; wherein the three-axis attitude angles include pitch angle, roll angle and yaw angle. The dynamic adjustment range of the three-axis attitude angle of the UAV and the aerodynamic imbalance variance are used to couple the three-axis angular velocity adjustment to obtain the three-axis angular velocity adjustment data; The UAV obstacle avoidance path attitude is adjusted based on the local spatial obstacle avoidance path, obstacle avoidance required speed, the dynamic adjustment range, and the three-axis angular velocity adjustment data to generate UAV obstacle avoidance path attitude data.

[0047] In this embodiment of the invention, spatial analysis is performed on the clustering prediction feature data of disordered trajectories. The system establishes a local spatial grid centered on the current position of the UAV in a global coordinate system, with a fixed grid density of one node per cubic meter. The coordinates of the obstacle cluster centers and the motion direction vectors in the disordered trajectory clustering prediction features are projected point-by-point onto the grid node set. After projection, the Euclidean distance from each node to the obstacle center is calculated. The system uses a minimum avoidance threshold to filter out all nodes with distances less than the safety limit, and forms a passable path chain among the remaining nodes using a connectivity search algorithm. The path chain is generated by weighting nodes according to the principle of minimizing node cost. The cost weight comes from the superposition coefficient of the obstacle movement trend and the wind field disturbance direction, and the wind field superposition coefficient is jointly determined by the environmental wind speed vector and the obstacle offset direction. During the iteration process, the algorithm dynamically calculates node priorities and expands layer by layer, finally selecting the connected path with the minimum comprehensive weight as the local spatial obstacle avoidance path output. This path consists of a series of spatial coordinate points, each containing a position vector and local orientation angle information, used to guide the attitude tracking and adjustment of the UAV.

[0048] After determining the obstacle avoidance path in the local space, the system calculates the obstacle avoidance speed requirement based on the clustering prediction features of disordered trajectories. First, it extracts the obstacle's motion speed, direction change frequency interval, and local wind speed vector along the path, inputting these three types of data into the time series analysis unit. The analysis unit uses a sliding time window cumulative averaging method to calculate the obstacle speed change trend and corrects this trend for the wind speed change rate, ensuring the result reflects the impact of actual spatial flow resistance on the UAV's speed requirement. The system sets a safe reaction distance between the UAV and obstacles and calculates the minimum speed increment required for local obstacle avoidance using the distance-to-response-time ratio. Subsequently, it performs linear interpolation on the speed requirement at each coordinate point along the path, forming a continuous obstacle avoidance speed distribution curve. The obstacle avoidance speed not only reflects the sequential acceleration changes between path segments but also includes speed suppression sections in different obstacle groups to maintain stable navigation. The final output obstacle avoidance speed requirement data is indexed by the path point sequence, with each path point storing the instantaneous target speed value and speed gradient information for subsequent dynamic matching in the attitude coupling stage.

[0049] Based on aerodynamic imbalance variance data, the dynamic adjustment range of the three-axis attitude angles of a UAV is evaluated. The torque fluctuation values ​​in the roll, pitch, and yaw directions from the aerodynamic imbalance variance are read, and the torque fluctuations are mapped to attitude angle disturbance amplitudes using the three-axis inertial matrix. The system calculates the disturbance peak value in each axis through a time sliding window to obtain the real-time fluctuation range of the three-axis attitude angles. Subsequently, combined with the current steady-state angle measurement data of the UAV, vector superposition analysis is used to determine the upper and lower limits of the attitude angles, forming the dynamic range boundaries of the pitch, roll, and yaw angles. This range boundary is defined as the upper limit of the UAV's adjustable parameters in the form of deviation amplitude and serves as an attitude correction constraint condition when executing center trajectory control. By establishing this range, the system can maintain attitude angle changes within a safe range during peak airflow disturbance periods, ensuring continuity and physical constraints for subsequent angular velocity control, and providing an amplitude limitation basis for three-axis coupled control.

[0050] The upper and lower limits of the three-axis attitude angles are converted into standardized coordinates, and a multi-axis proportional allocation rule is applied to calculate the angular velocity allocation ratio for each axis based on its variance weight. Then, the wind field disturbance direction is superimposed with the obstacle clustering orientation to generate a comprehensive disturbance vector, which is used to correct the direction coefficient of the angular velocity increment for each axis. The system executes a stepwise update algorithm over time, correcting the three-axis angular velocities in each calculation cycle based on the difference between the previous angular velocity output and the current disturbance change. After multiple iterations, the three-axis angular velocity adjustment data is output, recording the instantaneous angular velocities, rates of change, and coupling coefficients of the UAV's pitch, roll, and yaw directions, used to coordinate the balance of dynamic attitude response.

[0051] After generating the angular velocity adjustment data, the path attitude adjustment module integrates the local spatial obstacle avoidance path, obstacle avoidance required velocity, dynamic range of attitude angles, and three-axis angular velocity adjustment data to perform the final attitude adjustment. The system maps each spatial node in the path to the flight control output table in chronological order. For each node, the deviation between its expected displacement vector and the current attitude angle is calculated, and the angular velocity output is corrected using a proportional-integral control law to ensure the UAV's attitude remains stable along the corrected path. The obstacle avoidance required velocity serves as the node evaluation input to guide the propulsion rate, while the three-axis angular velocity data serves as the control feedback input for real-time closed-loop adjustment. The system continuously monitors the error between the actual attitude angle value and the set target through a synchronous sampling process. When the error exceeds the boundary of the dynamic adjustment range, the pitch, roll, and yaw speed command values ​​are immediately redistributed according to the angular velocity coupling coefficient. After the full-cycle attitude control process, the final UAV obstacle avoidance path attitude data is output. This data records the complete path attitude parameter sequence, angular velocity change sequence, and timing propulsion index, which are used by the UAV flight control system to perform real-time obstacle avoidance flight.

[0052] Step S3 includes the following steps: Step S31: Update the spatial coordinate point range of the UAV obstacle avoidance path attitude data to obtain the spatial coordinates of the UAV after obstacle avoidance; Step S32: Obtain the preset route for the entire inspection process; Step S33: Based on the spatial coordinates of the UAV after obstacle avoidance, replan the inspection route of the entire inspection process to obtain the inspection route replanning data. Step S34: Send the inspection route replanning data to the UAV control terminal to execute the inspection route planning.

[0053] As an example of the present invention, reference is made to Figure 3 As shown, step S3 in this example includes: Step S31: Update the spatial coordinate point range of the UAV obstacle avoidance path attitude data to obtain the spatial coordinates of the UAV after obstacle avoidance; In this embodiment of the invention, after the obstacle avoidance path attitude data is generated, the system first performs a spatial coordinate point range update operation. Based on the pose parameters of each node in the local obstacle avoidance path, a three-dimensional coordinate sequence and its corresponding time index are extracted. By calculating the spatial displacement between consecutive nodes, a set of track points actually traversed by the UAV during obstacle avoidance is generated. The set is arranged in chronological order and then interpolated for correction. The interpolation method uses a uniform time step difference method to eliminate trajectory discontinuities caused by non-constant sensor sampling intervals. The corrected track points are used to construct a smooth path curve through a spatial fitting algorithm. Subsequently, the boundary range of the path curve is calculated to obtain the spatial coordinates of the UAV after obstacle avoidance. The boundary range is represented in the form of a minimum envelope. By generating a convex hull region containing all actual flight points of the UAV in the three-dimensional coordinate field, the outer boundary of the UAV's feasible space is determined. The final output spatial coordinate data after obstacle avoidance is stored in full time series form. Each coordinate element contains a position vector, attitude angle, and velocity identifier, which are used as input data for subsequent route replanning.

[0054] Step S32: Obtain the preset route for the entire inspection process; In this embodiment of the invention, the system accesses the task database to read a preset inspection route. The preset route is predefined by the planning module and represents the UAV's flight trajectory under normal inspection conditions through a continuous sequence of geographic coordinate points. When reading this sequence, the system simultaneously retrieves the corresponding geographic projection parameters and task positioning reference point information to ensure that the coordinate system is consistent with the spatial coordinates after obstacle avoidance. Subsequently, the system performs data standardization on the preset route, converting the coordinate axis units, coordinate origin references, and attitude storage formats recorded at different flight stages into a unified format, eliminating representation differences between multi-source task data. The standardized route is stored as a set of paths segmented by flight stage. Each stage node contains a three-dimensional position point, desired speed, and task type identifier, serving as the basic reference input for obstacle avoidance replanning.

[0055] Step S33: Based on the spatial coordinates of the UAV after obstacle avoidance, replan the inspection route of the entire inspection process to obtain the inspection route replanning data. In this embodiment of the invention, the spatial coordinates after obstacle avoidance are paired and matched with the preset route throughout the entire process. The spatial offset distance under the same time index is compared and the deviation vector is calculated. For areas where the deviation exceeds the tolerance threshold, a set of areas to be replanned is generated. The replanning module constructs a local topology map based on these areas and searches for connected paths in the topology map with the current position of the UAV and the original target node as endpoints. The weight of each path edge is determined by the weighted sum of three indicators: path length, wind field disturbance direction, and obstacle clustering density. The system uses a dynamic programming algorithm to traverse different path combinations layer by layer, gradually calculates the path cost function, and selects the feasible path with the minimum weight. To ensure path continuity, the system performs curvature constraint analysis on the connecting segments of different paths and eliminates angle abrupt changes through an arc smoothing algorithm. The final output inspection route replanning data includes a three-dimensional path point sequence, an attitude control reference sequence, and a path distance index, providing a basis for subsequent control command issuance.

[0056] Step S34: Send the inspection route replanning data to the UAV control terminal to execute the inspection route planning.

[0057] In this embodiment of the invention, after the replanning data is generated, the system executes the inspection path data transmission operation. First, the replanning data is encoded into a flight control command format, with each data packet containing a path point number, target attitude angle, target velocity, and time index. After encoding, the data is encapsulated by the command scheduling module and segmented using a unified transmission protocol to ensure the timeliness and data integrity of the communication link. After receiving the data, the flight control terminal parses it according to the command sequence, reads the corresponding path point information in each control cycle, and calculates and outputs control signals based on the target attitude angle and aerodynamic state feedback. Simultaneously, the system initiates a continuity detection program to monitor command reception receipts, ensuring that the replanning commands are executed completely synchronously. The UAV control terminal executes the updated inspection route according to the new path, realizing the replanning of the flight trajectory and synchronous updating of attitude control, completing the actual flight task of the inspection route after obstacle avoidance.

[0058] Step S33 includes the following steps: Step S331: Calculate the deviation vector of the preset route for the entire inspection process based on the spatial coordinates of the UAV after obstacle avoidance to obtain the path deviation vector; Step S332: Based on the path deviation vector, perform optional path segmentation on the preset route of the entire inspection process to obtain multiple segments of paths to be replanned; Step S333: Perform connectivity analysis on the multiple paths to be replanned and solve for the shortest path to obtain the shortest path for the multiple paths to be replanned. Step S334: Perform inspection route replanning based on the shortest path to be replanned for multiple segments, and obtain inspection route replanning data.

[0059] In this embodiment of the invention, during the deviation vector calculation process, the spatial coordinate data of the UAV after obstacle avoidance and the coordinate data of the original inspection route are read. Both sets of data consist of a sequence of three-dimensional position points. The data alignment module matches the two sets of path points with a time index as a reference, calculates the difference between the actual coordinates at the same time step after obstacle avoidance and the preset coordinates, and obtains the position offset in the X, Y, and Z axes. The system synthesizes the offset at each moment to generate a directional offset vector. To improve accuracy, a moving average smoothing process is used on continuous offset vectors to eliminate the influence of instantaneous disturbances. Subsequently, the calculation module statistically analyzes the amplitude changes and directional distribution of the offset vectors to form a path deviation vector set. In addition to recording the offset direction and distance, this set also includes the rate of change of deviation within adjacent time intervals, used to quantify the dynamic degree of track deviation. After all calculations are completed, the path deviation vector is output and stored in the form of a three-dimensional numerical array as input data for subsequent optional path segmentation processing.

[0060] The system analyzes the time-series distribution of the deviation vector and defines intervals where the deviation amplitude is continuously above a threshold as abnormal path segments. Each abnormal segment represents a spatial segment where the UAV's flight maneuvers significantly deviate from the preset route. The system divides the entire preset route according to these segment boundaries, forming multiple non-overlapping path segments. For the boundary nodes of each path segment, a spatial continuity index is calculated, and the accessibility between path segments is determined through boundary point connectivity analysis. If there are significant height or direction differences between path segments, the system generates transition nodes in the adjacent space. These transition nodes are determined using cubic spline smoothing interpolation for subsequent connection analysis. After processing, multiple paths to be replanned are output, each with segment start and end spatial indices and offset strength labels for weighted calculation during shortest path solving.

[0061] When performing connectivity analysis on multiple replanning paths, the system constructs a topology network using path segment boundary nodes as vertices and accessible adjacent path segments as edges. Each edge is assigned a path cost, which includes weights for path length, wind disturbance direction deviation angle, and obstacle distribution density. The system records direct reachability relationships between segments using an adjacency matrix and employs a depth-first traversal algorithm to determine the connectivity of different path segments. For disconnected node pairs, the system establishes virtual connections by interpolating local curves to maintain topological integrity. After the connectivity structure stabilizes, the shortest path solution module calculates the minimum weight path from the starting node to the target node using a layer-by-layer relaxation method based on the cost matrix. To prevent local optima during calculation, the system uses a multi-path candidate combination sorting method to progressively filter the set of continuous path segments with the minimum cost. The output shortest path for multiple replanning segments is composed of node sequence, spatial length weight, and wind disturbance additional value as core parameters.

[0062] After the shortest path is solved, the system replans the inspection route based on multiple shortest path data segments. First, the selected path segments are geometrically reconstructed, rearranging all path node sequences in a 3D coordinate system using spatial curve interpolation to ensure continuous and smooth tangent directions connecting different segments. Then, the system extracts the spatial normal vector of each node to generate a path attitude reference curve. This attitude reference curve is combined with aerodynamic imbalance variance data to further correct attitude angle offsets at nodes, ensuring the path curvature meets stable flight constraints. The replanning module reallocates path point indices according to the time step, ensuring the UAV executes flight control with a constant control cycle. After all reconstruction is complete, the system outputs the replanned inspection route data, including a 3D coordinate sequence of path points, an attitude angle change sequence, a velocity reference table, and a path weight index, used for subsequent path command encapsulation and the UAV's final execution of trajectory update control.

[0063] The present invention also provides a drone inspection route planning system for executing the drone inspection route planning method described above. The drone inspection route planning system includes: The environmental acquisition module is used to collect real-time environmental data during the inspection route. The real-time environmental data includes real-time environmental images and real-time environmental wind conditions. The module performs time-series synchronization processing based on the real-time environmental wind conditions and real-time environmental images to obtain synchronized environmental status and wind conditions data. The obstacle avoidance path attitude adjustment module is used to mark the shape and density of obstacles in the environmental status and wind condition synchronization data; perform disordered trajectory prediction on the shape and density of the obstacles, and then perform UAV obstacle avoidance path attitude adjustment to generate UAV obstacle avoidance path attitude data. The inspection route replanning module is used to replan the inspection route based on the UAV obstacle avoidance path attitude data to obtain the inspection route replanning data; the inspection route replanning data is then sent to the UAV control terminal to execute the inspection route planning.

[0064] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for planning inspection routes using unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: Step S1: Collect real-time environmental data during the inspection route, including real-time environmental images and real-time environmental wind conditions; perform time-series synchronization processing based on the real-time environmental wind conditions and real-time environmental images to obtain synchronized environmental status and wind conditions data. Step S2: Mark the shape and density of obstacles in the environmental status and wind condition synchronization data; The obstacle shape and density are used to predict disordered trajectories, and then the UAV obstacle avoidance path attitude is adjusted to generate UAV obstacle avoidance path attitude data; wherein, step S2 includes: Step S21: Mark the shape and density of obstacles in the environmental status and wind condition synchronization data; wherein the obstacles include tree branches, fallen leaves and gravel; Step S22: Based on the synchronized data of environmental conditions and wind conditions, perform disordered trajectory prediction on the shape and density of the obstacles to obtain disordered trajectory prediction data for the obstacles; Step S23: Perform cluster feature analysis on the disordered trajectory prediction data of obstacles and output the disordered trajectory cluster prediction features; Step S24: Based on the disordered trajectory clustering prediction features and synchronized environmental and wind conditions data, adjust the UAV obstacle avoidance path attitude to generate UAV obstacle avoidance path attitude data; wherein, step S24 includes: Step S241: Quantify the degree of wind shear fluctuation in the synchronous data of environmental conditions and wind conditions; calculate the wind speed shear rate increase index of the degree of wind shear fluctuation, that is, compare the growth rate of wind speed gradient in adjacent time periods. Step S242: Based on the wind shear rate increase index, perform a coupled analysis of wind direction shear angle increase on the degree of wind shear fluctuation to obtain coupled data of shear angle increase; Step S243: Measure the aerodynamic imbalance variance based on the wind speed shear rate increase index and shear angle increase coupling data; the aerodynamic imbalance variance refers to the degree of attitude fluctuation of the UAV under wind disturbance. Step S244: Adjust the UAV obstacle avoidance path attitude based on the aerodynamic imbalance variance and disordered trajectory clustering prediction features to generate UAV obstacle avoidance path attitude data. Step S3: Perform inspection route replanning based on the UAV obstacle avoidance path attitude data to obtain inspection route replanning data; send the inspection route replanning data to the UAV control terminal to execute inspection route planning.

2. The UAV inspection route planning method according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Collect real-time environmental data during the inspection route using the camera and miniature anemometer mounted on the drone. The real-time environmental data includes real-time environmental images and real-time wind conditions. Step S12: Perform data preprocessing on the real-time environmental wind speed status to obtain wind condition preprocessing data; Step S13: Enhance the contrast of the real-time environmental image to obtain an optimized real-time environmental image; Step S14: Perform time-series synchronization processing based on the preprocessed wind condition data and the real-time environmental optimization image to obtain synchronized environmental condition and wind condition data.

3. The UAV inspection route planning method according to claim 1, characterized in that, Step S22 includes the following steps: Step S221: Estimate the mass of the obstacle based on its shape and quantity density; extract wind direction and wind force from the synchronized environmental and wind condition data; Step S222: Based on the wind direction and wind force, predict the frequency of directional changes in the mass and quantity density of the obstacle to obtain the directional changes of the obstacle; Step S223: Based on the change in obstacle direction, the mass and quantity density of obstacle objects, perform sparse feature analysis of dwelling points to obtain the sparse features of obstacle dwelling points; wherein the sparse features of dwelling points refer to the degree of sparseness of the distribution of the short-term dwelling positions of the obstacle in different directions during the process of the obstacle changing direction under the action of wind force. Step S224: Calculate the orientation tendency shift probability based on the sparse characteristics of obstacle dwell points to obtain the orientation tendency shift probability; perform trajectory entropy evolution deduction based on the obstacle orientation change frequency and orientation tendency shift frequency to obtain trajectory entropy evolution data. Step S225: Based on the trajectory entropy evolution data and the orientation tendency transfer probability, perform disordered trajectory prediction to obtain disordered trajectory prediction data for obstacles.

4. The UAV inspection route planning method according to claim 1, characterized in that, The variance of the metering pneumatic imbalance includes: The current flight motion state of the UAV is obtained; based on the wind speed shear rate increase index and shear angle increase coupled data, the airflow disturbance nonlinear fitting of the spatial gradient vector field is performed to obtain the airflow spatial disturbance vector; Extract the elevator deflection angle and rudder deflection angle from the current flight motion state of the UAV; Based on the airflow space disturbance vector, the elevator deflection angle and rudder deflection angle are decomposed into a random process of spatial lateral force imbalance to obtain the UAV lateral force imbalance data. The torque nonlinear amplification effect is estimated by advancing the lateral force imbalance data of the UAV in the time dimension. The torque imbalance nonlinear amplification effect means that when the UAV is disturbed by the non-uniform wind field, its attitude force does not show a linear proportional response change, but rather a nonlinear enhancement state is generated under the combined effect of time advancement and spatial airflow gradient superposition. The aerodynamic imbalance variance is measured based on the nonlinear amplification effect of the torque imbalance.

5. The UAV inspection route planning method according to claim 1, characterized in that, Adjusting the attitude of the drone's obstacle avoidance path includes: Based on the disordered trajectory clustering prediction features, obstacle avoidance paths are selected in the local space to obtain local space obstacle avoidance paths. The obstacle avoidance speed requirement is obtained by solving the obstacle avoidance speed requirement in the local space of the disordered trajectory clustering prediction features. The dynamic adjustment range of the three-axis attitude angles of the UAV is evaluated based on the aerodynamic imbalance variance; wherein the three-axis attitude angles include pitch angle, roll angle and yaw angle. The dynamic adjustment range of the three-axis attitude angle of the UAV and the aerodynamic imbalance variance are used to couple the three-axis angular velocity adjustment to obtain the three-axis angular velocity adjustment data; The UAV obstacle avoidance path attitude is adjusted based on the local spatial obstacle avoidance path, obstacle avoidance required speed, the dynamic adjustment range, and the three-axis angular velocity adjustment data to generate UAV obstacle avoidance path attitude data.

6. The UAV inspection route planning method according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Update the spatial coordinate point range of the UAV obstacle avoidance path attitude data to obtain the spatial coordinates of the UAV after obstacle avoidance; Step S32: Obtain the preset route for the entire inspection process; Step S33: Based on the spatial coordinates of the UAV after obstacle avoidance, replan the inspection route of the entire inspection process to obtain the inspection route replanning data. Step S34: Send the inspection route replanning data to the UAV control terminal to execute the inspection route planning.

7. A drone inspection route planning system, characterized in that, For executing the UAV inspection route planning method as described in claim 1, the UAV inspection route planning system includes: The environmental acquisition module is used to collect real-time environmental data during the inspection route. The real-time environmental data includes real-time environmental images and real-time environmental wind conditions. The module performs time-series synchronization processing based on the real-time environmental wind conditions and real-time environmental images to obtain synchronized environmental status and wind conditions data. The obstacle avoidance path attitude adjustment module is used to mark the shape and density of obstacles in the environmental status and wind condition synchronization data; perform disordered trajectory prediction on the shape and density of the obstacles, and then perform UAV obstacle avoidance path attitude adjustment to generate UAV obstacle avoidance path attitude data. The inspection route replanning module is used to replan the inspection route based on the UAV obstacle avoidance path attitude data to obtain the inspection route replanning data; the inspection route replanning data is then sent to the UAV control terminal to execute the inspection route planning.

Images