A method for data fusion and quality control of unmanned aerial vehicle rain enhancement for karst complex terrain

Abstract

This invention discloses a quality control method for UAV rain enhancement data fusion in complex karst terrain, belonging to the field of electronic digital data processing technology. This method preprocesses multi-source data to form a standardized dataset, extracts landform feature points to construct a terrain-constrained spatiotemporal benchmark, divides the terrain gradient into zones for spatiotemporal registration, applies a terrain-meteorological coupled weighting algorithm to weight and fuse the data, and utilizes a three-dimensional quality control threshold adaptive algorithm to perform point-by-point quality control and anomaly processing on the fused data, finally outputting the fused data. This invention achieves precise spatiotemporal alignment by constructing a terrain-constrained benchmark and gradient partitioning, integrates landform feature recognition and adaptive quality control thresholds, eliminates dependence on satellite signals, improves the spatial consistency and temporal continuity of the fused data, effectively solves the problems of spatiotemporal misalignment and fusion deviation caused by terrain, provides high-quality data support for UAV rain enhancement operations, and enhances the scientific rigor and effectiveness of rain enhancement operations.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology, specifically to a quality control method for UAV rain enhancement data fusion in complex karst terrain. Background Technology

[0002] Karst regions, due to their unique geological structures, exhibit complex and diverse landforms such as peak clusters, caves, and depressions. The terrain is highly undulating with complex surface cover types. Meteorological conditions in these areas are variable, and precipitation distribution is uneven. Artificial rain enhancement has become an important means to improve water resources and regulate the ecological environment. Drone-based rain enhancement, with its advantages of high flexibility, wide operating range, and adaptability to complex terrain, is increasingly widely used in karst regions. However, the effectiveness of drone-based rain enhancement operations highly depends on the support of multi-source meteorological and topographic data. Effective integration of multi-source data, including drone-borne data, ground-based radar data, satellite remote sensing data, ground rain gauge data, and topographic data, is crucial for improving the accuracy of rain enhancement operations. Currently, technologies for meteorological data fusion and quality control have made some progress. Multi-source data fusion has become an important research direction in the meteorological field, with various fusion algorithms and quality control methods constantly emerging, providing multiple approaches to meteorological data processing. At the same time, the accurate acquisition and application of topographic data is gradually becoming a point of convergence between meteorology and geography, laying a technical foundation for meteorological data processing in complex terrain.

[0003] Traditional UAV rain enhancement data fusion quality control methods are mostly designed based on general terrain scenarios, failing to fully consider the special characteristics of karst terrain. This leads to numerous adaptability issues when applied in this region. Traditional methods often rely on satellite positioning signals to establish registration benchmarks, but karst areas are characterized by high mountains and deep valleys, making satellite signals easily blocked. This results in a significant decrease in spatial registration accuracy, making it impossible to achieve precise spatiotemporal alignment of multi-source data. Furthermore, traditional data fusion methods often use fixed weights for calculation, failing to dynamically adjust based on the actual matching degree between terrain features and meteorological data. This makes it difficult to highlight the observational advantages of different data sources in different karst terrain areas, resulting in insufficient continuity and completeness of the fused data. In addition, traditional quality control often uses globally uniform threshold standards, failing to consider the spatiotemporal fluctuation characteristics of data under different karst terrain units, easily leading to misjudgments or omissions in quality control. Abnormal data processing methods are also relatively simplistic, unable to provide targeted repair based on the type and scope of anomalies. Consequently, the reliability of the processed data fails to meet the actual needs of UAV rain enhancement operations in karst areas. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a quality control method for UAV rain enhancement data fusion in complex karst terrain. This method includes data preprocessing, terrain feature point extraction and benchmark construction, terrain gradient partitioning and spatiotemporal registration, weighted fusion based on terrain-meteorological coupling weight algorithm, and point-by-point quality control based on three-dimensional quality control threshold adaptive algorithm. By constructing a terrain-constrained spatiotemporal benchmark that does not rely on satellite signals, and combining partitioned fusion and adaptive quality control strategies, it achieves accurate alignment, efficient fusion and intelligent quality control of multi-source data in karst environments.

[0005] To solve the above-mentioned technical problems, this invention provides the following technical solution: a quality control method for UAV rain enhancement data fusion in complex karst terrain, the specific steps of which are as follows:

[0006] S100. Data Preprocessing: Collect UAV airborne data, ground-based radar data, satellite remote sensing data, ground rain gauge data, and topographic data in the karst area. Perform format unification, missing value imputation, and outlier removal on the above multi-source raw data to form a standardized dataset, which provides data support for the topography-meteorology coupled weight generation algorithm and the three-dimensional quality control threshold adaptive algorithm.

[0007] S200, Benchmark Construction: Identify and extract karst landform feature points from the standardized dataset. Based on the spatial coordinates and elevation information of the landform feature points, establish a terrain-constrained spatiotemporal registration benchmark that does not rely on satellite positioning signals, providing a unified reference for the spatial alignment of multi-source data.

[0008] S300, Spatiotemporal Registration: Based on the terrain slope, elevation and land cover type, the terrain gradient is divided into zones. Within each zone, time synchronization and coordinate normalization are performed on multi-source data based on the terrain-constrained spatiotemporal registration benchmark, so that data from different sources are unified in the same spatiotemporal framework.

[0009] S400, Data Fusion: The terrain-meteorological coupled weight generation algorithm is applied to calculate the comprehensive spatiotemporal weight of each terrain at the corresponding time. According to the comprehensive spatiotemporal weight, the multi-source data within the unified spatiotemporal framework is weighted and fused to form continuous and complete fused data.

[0010] S500 Quality Control: The three-dimensional quality control threshold adaptive algorithm is applied to determine the adaptive quality control threshold for the corresponding terrain and data type. Based on the adaptive quality control threshold, point-by-point judgment and anomaly handling are performed on the fused data, and the fused quality control data that meets the requirements of spatiotemporal consistency and reliability is output.

[0011] Furthermore, the multi-source raw data specifically includes UAV-borne data, ground-based radar data, satellite remote sensing data, ground rain gauge data, and topographic data. Specifically, UAV-borne data is collected through multiple sensors mounted on the UAV, covering relevant meteorological and operational trajectory data for the karst region; ground-based radar data is collected through fixed ground-based radar stations deployed within the karst region, acquiring precipitation-related data; satellite remote sensing data is collected collaboratively by polar-orbiting and geostationary satellites, acquiring cloud system and image-related data for the region; ground rain gauge data is collected through automatic rain gauges deployed in different topographic areas within the study area, acquiring precipitation-related data for the region; and topographic data mainly consists of high-precision DEM topographic data, acquired through a combination of UAV-borne LiDAR sensor scanning and downloading publicly available topographic data. After all data is collected, it is uniformly aggregated, and then standardized datasets are formed through format standardization, missing value imputation, and preliminary outlier removal.

[0012] Furthermore, the karst landform feature points are landmark points in the karst region that have unique topographic features, stable spatial locations, and are easy to identify. Specifically, they include the center point of the solution funnel, the inflection point of the upper and lower boundaries of the steep cliff, the outline corner point of the cave entrance, the center point of the bottom of the depression, and the feature point at the top of the rock column.

[0013] The specific operations for identifying and extracting geomorphic feature points are as follows: Based on a standardized dataset, high-precision DEM topographic data and UAV aerial image data are extracted, and the two are combined for collaborative analysis; through a combination of topographic curvature analysis, elevation change detection, and image texture feature recognition, various geomorphic feature points are automatically identified. For karst funnels, the center point is determined by identifying the topographic depression area and locating its geometric center; for steep cliffs, the turning points of the upper and lower boundaries are extracted as inflection points by detecting elevation change areas; for cave entrances, the cave entrance outline is identified by image texture differences, and key corner points of the outline are extracted; for depressions and rock pillars, the lowest point at the bottom and the highest point at the top are located as feature points, respectively. After extraction, the spatial coordinates and elevation information of each feature point are verified, abnormal points are eliminated, and finally, the set of geomorphic feature points used to construct the spatiotemporal registration benchmark is determined.

[0014] Furthermore, the terrain-constrained spatiotemporal registration benchmark is constructed with extracted karst landform feature points as its core, without relying on satellite positioning signals. Its core components include a spatial benchmark and a temporal benchmark. The spatial benchmark establishes a region-specific three-dimensional spatial coordinate system based on the spatial coordinates and elevation information of all verified landform feature points. Each landform feature point serves as a spatial control node, determining the relative spatial positions between nodes and forming a spatial reference framework covering the entire karst study area. The temporal benchmark uses the standard timestamps of each landform feature point's data collection as its basis, uniformly calibrating the time information of all relevant data for all landform feature points. This eliminates clock deviations from different acquisition devices, achieving unified alignment in the time dimension. This spatiotemporal registration benchmark clarifies the reference standard for spatial alignment of multi-source data and can be directly used as a unified reference for spatial position calibration and coordinate matching of multi-source data such as UAV airborne data, ground-based radar data, and satellite remote sensing data, ensuring accurate alignment of various types of data within a unified spatiotemporal framework.

[0015] Furthermore, the topographic gradient zoning is based on the topographic data within the standardized dataset, combined with the actual surface conditions of the study area. It comprehensively divides the entire karst study area into different topographic gradient zones based on topographic slope, ground elevation, and surface cover type. During the division, the slope is used to distinguish between gentle, sloping, and steep areas: slope < 15° is considered gentle, 15° ≤ slope < 35° is considered sloping, and slope ≥ 35° is considered steep. Elevation is used to distinguish between low, medium, and high altitude areas: elevation < 800m is considered low altitude, 800m ≤ elevation ≤ 1500m is considered medium altitude, and elevation > 1500m is considered high altitude. Surface cover type is used to distinguish between bare rock areas, vegetation-covered areas, valley areas, and peak-cluster areas. Through the overlay analysis and spatial clustering of these three types of topographic factors, several topographic gradient zones with clear boundaries and consistent attributes are formed. The topographic features and surface conditions within each zone are relatively uniform, serving as the basic processing unit for spatiotemporal registration of multi-source data.

[0016] Furthermore, the mathematical expression for the terrain-meteorological coupled weight generation algorithm is:

[0017] ;

[0018] in, The combined spatiotemporal weight of the terrain unit in the i-th row and j-th column at time t; The topographic stability factor is determined by the number and distribution of characteristic points such as dissolution funnels, steep cliffs, and cave entrances within the unit, and its value ranges from [0,1]. The meteorological matching factor reflects the consistency of UAV, ground-based radar, and satellite remote sensing data within the unit at time t, and its value ranges from [0,1]. The data reliability factor is determined by factors such as equipment calibration status and ambient noise level, and its value ranges from [0,1]. These are the weight coefficients corresponding to the terrain stability factor, meteorological matching factor, and data reliability factor, respectively, satisfying... It can be dynamically adjusted according to different terrain types such as peak clusters, canyons, and exposed rock areas within the karst region;

[0019] In the data fusion phase, based on the aforementioned comprehensive spatiotemporal weights, weighted fusion is performed on multi-source data within the unified spatiotemporal framework, mathematically expressed as:

[0020] ;

[0021] in, This represents the fused data of the terrain unit in the i-th row and j-th column at time t; represents the observation data of the k-th data source at time t in the i-th row and j-th column of the terrain unit; n represents the total number of data sources participating in the fusion.

[0022] Furthermore, the mathematical expression of the three-dimensional quality control threshold adaptive algorithm is:

[0023] ;

[0024] in, The adaptive thresholds are defined for the k-th type of terrain unit, the l-th type of data type, and the m-th quality control indicator; k is the terrain type index after the study area is divided, used to identify different terrain gradient partitions; l is the data type index, used to identify multi-source data types; and m is the quality control indicator index, used to identify specific quality control indicators. This is the global benchmark threshold for this quality control indicator, and the initial threshold is uniform across the entire region. is the measured fluctuation coefficient of the l-type data within the k-th type of terrain unit, reflecting the spatiotemporal fluctuation of the data within that terrain unit; is the average fluctuation coefficient of the l-th type of data in the whole region, and is the statistical mean of the fluctuation degree of this type of data in the whole region, which serves as the benchmark for terrain fluctuation correction; is the historical misjudgment rate of the m-th quality control indicator within the k-th terrain unit, reflecting the historical performance of the quality control indicator within that terrain unit; Let be the average misjudgment rate of the m-th quality control indicator in the whole region, and let be the statistical mean of the historical judgment deviation of the quality control indicator in the whole region, which serves as the benchmark for correcting the quality control effect. This is an adjustment coefficient used to balance the influence weight of terrain fluctuations and quality control effects. It can be optimized according to operational needs, and the value range can be set according to the actual scenario, with a value range of [0,1].

[0025] Furthermore, the main contents of the point-by-point determination and anomaly handling are as follows:

[0026] Point-by-point judgment: Traverse each spatiotemporal data point in the fused data, and compare the measured value of the point with the adaptive quality control threshold under the corresponding terrain type, data type and quality control index. When the measured value is within the reasonable range of the adaptive quality control threshold, it is judged as valid data; when it exceeds the range, it is marked as abnormal data.

[0027] Anomaly Handling: A tiered processing strategy is adopted for marked anomalous data: For isolated single-point anomalies, the neighborhood spatiotemporal interpolation method is used to perform numerical repair by utilizing the spatiotemporal correlation of valid data around the point; for continuous anomalous areas, the anomalous range is first located by using terrain-constrained spatiotemporal registration benchmarks, and then the multi-source data cross-validation method is used to correct the anomalies by combining multi-source data; for severely anomalous data that cannot be repaired, they are removed, and the cause of the anomaly and the handling method are marked in the data record.

[0028] Furthermore, the neighborhood spatiotemporal interpolation method uses the spatial location and timestamp of the data point marked as an anomaly as a benchmark, selects eight spatially adjacent terrain units and two temporally adjacent valid data points as interpolation samples, and generates interpolation correction values ​​through a weighted average method to replace the measured values ​​of the anomaly points, ensuring the continuity and consistency of the data in the spatiotemporal dimension. The multi-source data cross-validation method extracts corresponding observation values ​​from different sources such as UAV airborne data, ground-based radar data, and satellite remote sensing data for the same terrain unit and the same time. By comparing the degree of deviation between the data from each source, the reliability of the fused data is verified a second time. The preset range of the deviation of the multi-source data is ±15% of the mean of the data from each source. When the deviation between the single source data and the mean of the multi-source data exceeds this range, the fusion result of that point is reviewed and corrected, further improving the overall credibility of the fused quality control data.

[0029] Compared with existing technologies, this UAV rain enhancement data fusion quality control method for complex karst terrain has the following advantages:

[0030] I. This invention designs a multi-source data fusion quality control process tailored to the complex terrain characteristics of karst areas. First, it completes standardized preprocessing of the original multi-source data. Then, it constructs a terrain-constrained spatiotemporal registration benchmark independent of satellite positioning signals. Combined with terrain gradient zoning, it achieves accurate spatiotemporal alignment of the multi-source data. Relying on a terrain-meteorological coupling weight generation algorithm, it assigns comprehensive spatiotemporal weights to different terrain units and different times, making the weighted fusion of multi-source data more closely aligned with the terrain and meteorological characteristics of the karst region. Finally, a three-dimensional quality control threshold adaptive algorithm matches specific adaptive thresholds for different terrains, data types, and quality control indicators, achieving precise quality control of the fused data. The entire process revolves around the full-process adaptation design of karst landform characteristics, effectively solving the problems of spatiotemporal misalignment and large fusion deviations caused by the complex terrain of the region, and significantly improving the spatial consistency and temporal continuity of the fused data.

[0031] Second, this invention incorporates the unique attributes of karst topography into each stage of data processing. In the geomorphic feature point extraction stage, it combines high-precision topographic data with aerial imagery to achieve accurate feature point identification, laying a solid foundation for spatiotemporal registration. In the quality control stage, a hierarchical anomaly handling strategy is adopted, using differentiated processing methods such as interpolation repair, cross-validation correction, or elimination for different types of abnormal data. This approach balances the rationality of data repair with the rigor of quality control. Compared with traditional general-purpose data fusion quality control methods, this method eliminates the dependence on satellite positioning signals, adapts to the topographic gradient differences in karst regions, and makes data fusion and quality control more targeted. The output fused quality control data can accurately reflect the actual meteorological and precipitation conditions in karst regions, providing high-quality data support for UAV rain enhancement operations and improving the scientific nature and effectiveness of rain enhancement operations.

[0032] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0034] Figure 1 A flowchart of a quality control method for UAV rain enhancement data fusion in complex karst terrain;

[0035] Figure 2A schematic diagram of data transmission for a UAV-based rain enhancement data fusion quality control method for complex karst terrain;

[0036] Figure 3 This is a schematic diagram illustrating the data transmission construction based on the spatiotemporal reference of topographic feature points according to the present invention. Detailed Implementation

[0037] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0038] Example 1: Implementation of quality control for UAV rain enhancement data fusion in karst peak-cluster depression areas;

[0039] This embodiment is applied to summer drone-based rain enhancement operations in a typical karst peak-cluster depression region in Southwest China. This region features a typical topography of towering peaks and numerous depressions, with dramatic terrain undulations and satellite signals easily blocked by mountains. Conventional data processing methods are insufficient to meet the high-precision data requirements of rain enhancement operations. This invention utilizes a drone-based rain enhancement data fusion quality control method for complex karst terrain to conduct full-process multi-source data fusion quality control processing, accurately adapting to regional terrain characteristics. This provides high-quality data support for the planning, implementation, and effect evaluation of drone-based rain enhancement operations. The entire process strictly follows five core steps, with each step deeply integrated with regional terrain characteristics to ensure the accuracy and adaptability of data processing. Figure 1 As shown.

[0040] Comprehensive data collection was conducted within the peak-cluster depression area, including airborne data from drones, ground-based radar data, satellite remote sensing data, ground rain gauge data, and topographic data. The drone-borne data, collected synchronously by rain-enhancing drones equipped with multiple sensors along their operational flight paths, covers key meteorological elements such as temperature, humidity, cloud droplet spectrum, and air pressure, as well as precise drone operational trajectory data, providing a direct reflection of near-surface meteorological conditions in the operational area. Ground-based radar data, continuously acquired from fixed ground-based radar stations within the area, provides precipitation-related data such as precipitation echo intensity and cloud movement speed, enabling a comprehensive assessment of regional precipitation conditions. Remote monitoring; satellite remote sensing data is collected through collaborative observation of polar-orbiting and geostationary satellites to obtain images and meteorological data such as the distribution range of regional cloud systems, cloud thickness, and cloud top height, achieving full coverage monitoring of meteorological conditions in large-scale areas; ground rain gauge data is collected by automatic rain gauges evenly deployed at different terrain locations such as the top of peak clusters, the bottom of depressions, and both sides of valleys, which can obtain accurate actual ground precipitation data; topographic data is obtained by combining high-precision DEM topographic data scanned by UAV-borne LiDAR sensors with publicly available authoritative topographic data, which can accurately reconstruct the regional topographic undulations and geomorphological features. The aforementioned multi-source raw data are uniformly summarized and organized to eliminate format differences caused by different equipment and channels during data collection. Format unification, missing value imputation, and preliminary outlier removal are then performed sequentially. Data of different formats and dimensions are converted into a unified and standardized format. Missing data caused by equipment failure or environmental interference are scientifically imputed, and outlier data that significantly deviates from the normal observation range are initially screened and removed. Finally, a standardized dataset is formed, providing complete, standardized, and reliable data support for the efficient operation of subsequent terrain-meteorological coupled weight generation algorithms and three-dimensional quality control threshold adaptive algorithms, ensuring the effectiveness of subsequent processing steps from the data source.

[0041] High-precision DEM topographic data and UAV aerial imagery data were accurately extracted from a standardized dataset. These two types of data were then collaboratively analyzed and processed, fully leveraging the elevation accuracy of the DEM data and the landform visualization of the aerial imagery data to achieve accurate identification and extraction of karst landform feature points. Extracted feature points included the center points of solution funnels, the inflection points of steep cliff boundaries, the outline corners of cave entrances, the center points of depression bottoms, and the feature points at the tops of rock pillars. These feature points are all unique, spatially stable, and easily identifiable landmarks within the region, serving as core references for regional spatiotemporal registration. All extracted landform feature points underwent rigorous verification of their spatial coordinates and elevation information, eliminating outliers caused by observation errors or terrain obstruction, and selecting a set of landform feature points with accurate spatial locations and reliable elevation information. Based on the spatial coordinates and elevation information of this set of feature points, a terrain-constrained spatiotemporal registration benchmark is constructed, independent of satellite positioning signals. This completely eliminates the impact of satellite signal obstruction on data registration in peak-cluster depression areas. The spatial benchmark is based on the spatial coordinates and elevation information of all verified geomorphic feature points, establishing a region-specific three-dimensional spatial coordinate system covering the entire peak-cluster depression study area. Each geomorphic feature point serves as a spatial control node, clarifying the relative spatial relationships between nodes and forming a precise spatial reference framework. The temporal benchmark is based on the standard timestamps of data collection for each geomorphic feature point, uniformly calibrating the time information of all geomorphic feature point-related data to eliminate time errors caused by clock deviations from different acquisition devices, achieving unified alignment in the time dimension. This terrain-constrained spatiotemporal registration benchmark provides a unified and accurate reference for the spatial alignment of multi-source data, ensuring accurate spatiotemporal matching of various data in subsequent processing, laying a solid spatiotemporal foundation for multi-source data fusion. Figure 3 As shown.

[0042] Based on high-precision topographic data within a standardized dataset, and combined with the actual surface conditions of the peak-cluster depression region, topographic gradient zoning was conducted by comprehensively considering three core topographic factors: topographic slope, ground elevation, and surface cover type. The regions were divided into three categories: gentle slope (<15°), sloping slope (15°≤<35°), and steep slope (≥35°); low elevation (<800m), medium elevation (800m≤≤1500m), and high elevation (>1500m); and bare rock, vegetated, valley, and peak-cluster regions were distinguished by surface cover type. Through overlay analysis and spatial clustering of the three types of topographic factors, several topographic gradient zones with clear boundaries and relatively uniform internal topographic features and surface conditions were identified. This ensures consistency in topographic features within each zone and avoids a decrease in data registration accuracy due to excessive topographic differences. Each terrain gradient partition is used as the basic processing unit for spatiotemporal registration of multi-source data. Within each partition, time synchronization and coordinate normalization are performed on multi-source data such as UAV airborne data, ground-based radar data, satellite remote sensing data, and ground rain gauge data. Time synchronization calibrates the timestamps of each data source, allowing observation data from different devices at the same time to form a correspondence. Coordinate normalization unifies the coordinate system of all data to the three-dimensional spatial coordinate system of the terrain-constrained spatiotemporal registration benchmark, completely eliminating spatial deviations caused by complex terrain and different observation perspectives of devices. This unifies multi-source data from different sources and observation methods within the same spatiotemporal framework, achieving accurate spatiotemporal alignment of multi-source data. This allows subsequent data fusion to be carried out on the basis of spatiotemporal consistency, improving the rationality and accuracy of data fusion.

[0043] Based on accurate spatiotemporal registration of multi-source data, a topography-meteorology coupled weight generation algorithm is applied to carry out data fusion processing. The mathematical expression of the topography-meteorology coupled weight generation algorithm is:

[0044] ;

[0045] in, The combined spatiotemporal weight of the terrain unit in the i-th row and j-th column at time t; For terrain stability factor; Meteorological matching factor; For data reliability factor; These are the weighting coefficients for the terrain stability factor, meteorological matching factor, and data reliability factor, respectively. This algorithm enables a deep integration of terrain features and meteorological data, making the data fusion results more consistent with the actual terrain and meteorological conditions of peak-cluster depression areas. Based on the algorithm requirements, the terrain stability factor is determined by combining the number and distribution of feature points such as dissolution funnels, steep cliffs, and cave entrances within each terrain gradient zone. This factor reflects the stability of terrain units and provides a terrain dimension reference for weight calculation. The meteorological matching factor is determined by combining the observation consistency of multi-source data such as UAVs, ground-based radar, and satellite remote sensing at corresponding times. This factor reflects the fit of multi-source meteorological data and provides a meteorological dimension reference for weight calculation. The data reliability factor is determined by combining factors such as equipment calibration status, observation environment noise level, and equipment operational stability. This factor reflects the observation reliability of each data source and provides a data quality dimension reference for weight calculation. Simultaneously, the weight coefficients of terrain stability factor, meteorological matching factor, and data reliability factor are dynamically adjusted according to different terrain types in peak-cluster depression areas, such as peak clusters, canyons, and exposed rock areas, ensuring that the sum of each coefficient is 1. This allows the weight allocation to accurately adapt to the characteristics of different terrain units. Subsequently, the comprehensive spatiotemporal weight of each terrain unit in the region at each moment of the UAV rain enhancement operation is calculated. Based on the calculated comprehensive spatiotemporal weight, a weighted fusion operation is performed on multi-source data within the same spatiotemporal framework. This fully integrates the observation advantages of each data source, compensating for the observation limitations of a single data source in peak-cluster depression areas caused by terrain obstruction and observation range limitations. For example, the near-ground accuracy of UAV-borne data, the large-scale coverage of satellite remote sensing data, and the professional precipitation monitoring of ground-based radar data are utilized. This ensures that the fused data can balance accuracy and comprehensiveness, ultimately forming continuous and complete fused data that accurately reflects the meteorological and precipitation characteristics of different terrain units in peak-cluster depression areas, providing basic data for subsequent quality control.

[0046] A three-dimensional quality control threshold adaptive algorithm is applied to perform quality control processing on fused data. The mathematical expression of the three-dimensional quality control threshold adaptive algorithm is:

[0047] ;

[0048] in, The adaptive threshold is defined for the k-th type of terrain unit, the l-th type of data type, and the m-th quality control indicator. This serves as the global benchmark threshold for quality control indicators; is the measured fluctuation coefficient of the l-th type of data within the k-th type of terrain unit; The average fluctuation coefficient of the l-th type of data across the entire region; The historical misjudgment rate of the m-th quality control indicator within the k-th terrain unit; Let m be the average misjudgment rate of the m-th quality control indicator across the entire region; To adjust the coefficients, this algorithm enables dynamic adjustment of the quality control thresholds, allowing quality control standards to accurately adapt to the characteristics of different terrains and data types in peak-cluster depression areas. Based on the algorithm requirements, and considering different terrain unit types after the area is divided, different data types such as UAV-borne data, ground-based radar data, and satellite remote sensing data, as well as different quality control indicators such as precipitation intensity, cloud parameters, and elevation matching degree, corresponding adaptive quality control thresholds are determined. During threshold determination, the algorithm dynamically adjusts the thresholds by combining the measured fluctuation coefficients of various data in different terrain units with the average fluctuation coefficient for the entire region, and the historical misjudgment rate of each quality control indicator with the average misjudgment rate for the entire region. Furthermore, by adjusting the coefficients, the algorithm balances the influence weights of terrain fluctuations and quality control effectiveness, ensuring that the quality control thresholds align with the data fluctuation characteristics and quality control requirements of different terrain units, avoiding misjudgments or omissions caused by using a globally uniform threshold. Based on the established adaptive quality control threshold, point-by-point judgment is performed on the fused data. Each spatiotemporal data point in the fused data is traversed, and the measured value of that point is precisely compared with the adaptive quality control threshold under the corresponding terrain type, data type, and quality control indicators. If the measured value is within a reasonable range of the adaptive quality control threshold, it is considered valid data; otherwise, it is marked as anomalous data, achieving comprehensive and accurate screening of the fused data. A hierarchical processing strategy is adopted for marked anomalous data. For isolated single-point anomalous data, a neighborhood spatiotemporal interpolation method is used to perform numerical repair using the spatiotemporal correlation of valid data around the point, ensuring the continuity of data in the spatiotemporal dimension. For continuous anomalous areas, the anomalous range is first accurately located using terrain-constrained spatiotemporal registration benchmarks, and then a multi-source data cross-validation method is used to perform secondary verification and anomalous correction by combining observations from different data sources, improving the accuracy of anomalous data repair. Severely anomalous data that cannot be repaired is removed, and the cause of the anomalous data and the processing method are marked in detail in the data record, achieving standardized processing of anomalous data. Through full-process quality control, the final output is fused quality control data that meets the requirements of spatiotemporal consistency and reliability. This data can accurately and comprehensively reflect the meteorological, precipitation and topographic characteristics of the peak-cluster depression area, providing high-precision and high-reliability data support for the flight path planning, operation timing selection, operation parameter adjustment and operation effect evaluation of UAV rain enhancement operations in the area, thereby improving the scientificity and accuracy of UAV rain enhancement operations.

[0049] This embodiment addresses the topographical characteristics of the karst peak-cluster depression region in Southwest China and the needs of summer rain enhancement operations. It utilizes the method of this invention to complete the entire process of multi-source data fusion quality control. Starting with the standardized processing of multi-source raw data, a terrain-constrained spatiotemporal registration benchmark independent of satellite signals is constructed. Accurate spatiotemporal registration is achieved by combining topographic factors. Weighted fusion of multi-source data is realized through a topographic-meteorological coupled weight generation algorithm, followed by hierarchical quality control using a three-dimensional quality control threshold adaptive algorithm. Each step is tailored to the geomorphological characteristics of the peak-cluster depression region, effectively solving the data registration and fusion challenges caused by satellite signal obstruction and large topographic undulations. The final output fused quality control data is both accurate and comprehensive, directly supporting the entire process decision-making for UAV rain enhancement operations in this region, significantly improving the scientific rigor and accuracy of the operations.

[0050] Example 2: Implementation of quality control for UAV-based rain enhancement data fusion in karst canyon steep cliff areas;

[0051] This embodiment is applied to spring drone-based rain enhancement operations in a region with a concentrated distribution of karst canyons and steep cliffs in South China. This region features deep canyons, widespread steep cliffs, steep slopes, and a surface cover dominated by bare rock and sparse vegetation. Meteorological data observations are significantly affected by topographical obstruction and complex airflow, making multi-source observation data prone to bias and anomalies. Conventional data fusion and quality control methods are difficult to adapt to the complex terrain characteristics of this region. This invention utilizes a drone-based rain enhancement data fusion and quality control method tailored to complex karst terrain to process multi-source rain enhancement data, specifically addressing the data processing challenges in canyon and steep cliff areas. This provides high-quality fused quality control data for drone-based rain enhancement operations in this region. Each step of the operation is closely integrated with the regional topography and meteorological observation characteristics, progressively achieving data standardization, registration, fusion, and quality control. Figure 2 As shown.

[0052] The system systematically collects airborne data from drones, ground-based radar data, satellite remote sensing data, ground rain gauge data, and topographic data in the steep cliff area of ​​this karst canyon. The drone-borne data is collected by rain-enhancing drones equipped with multiple sensors, including temperature and humidity sensors, barometric pressure sensors, cloud particle detectors, and GPS positioning devices. During flight along the canyon route, meteorological data at different altitudes within the canyon and precise drone flight trajectory data are simultaneously acquired, accurately capturing the complex near-surface meteorological characteristics within the canyon. Ground-based radar data is collected by fixed ground-based radar stations deployed in open areas around the canyon, continuously monitoring the formation, development, movement, and intensity of precipitation clouds, achieving long-distance, unobstructed monitoring of precipitation in the canyon area. Measurement: Satellite remote sensing data is collected through collaborative observation of polar-orbiting and geostationary satellites to obtain remote sensing data such as cloud movement paths, cloud height, cloud thickness, and surface temperature in the canyon area, presenting regional meteorological and surface conditions from a macroscopic perspective; Ground rain gauge data is collected by automatic rain gauges strategically placed at different terrain locations such as both sides of the canyon, below steep cliffs, and at the confluence of valleys, to obtain real-time actual ground precipitation data, providing accurate ground references for precipitation monitoring; Topographic data is obtained by finely scanning the steep cliff area of ​​the canyon using an airborne LiDAR sensor on a drone to acquire high-precision DEM topographic data, which is supplemented and improved by combining with publicly available karst regional topographic data to accurately restore the depth characteristics of the canyon, the elevation changes of the steep cliffs, and the overall topographic undulations of the region. All collected multi-source raw data were uniformly summarized, and the data collection time, collection location, data type, and data format were sorted out. Then, format unification processing was carried out to convert heterogeneous data collected from different devices and channels into a unified format standard, solving the problem of data format incompatibility. Missing data caused by temporary equipment failure, signal interruption, or extreme terrain obstruction were scientifically and reasonably imputed to ensure data integrity. Outliers that deviated significantly from the normal observation range due to environmental interference or equipment misjudgment were initially removed to reduce the impact of invalid data on subsequent processing. Finally, a standardized dataset was formed, providing a standardized, complete, and reliable data foundation for the smooth application of the terrain-meteorological coupled weight generation algorithm and the three-dimensional quality control threshold adaptive algorithm, ensuring that subsequent algorithms can carry out calculations and processing based on high-quality raw data.

[0053] High-precision DEM topographic data and high-resolution aerial imagery of canyons and cliffs from a standardized dataset were accurately extracted. These two types of data underwent in-depth collaborative analysis, fully leveraging the elevation accuracy of the DEM data and the detailed geomorphic features of the aerial imagery. Through a combination of topographic curvature analysis, elevation change detection, and image texture feature recognition, karst geomorphic feature points in the region were automatically identified. Key features extracted included the center points of solution funnels, the inflection points of the upper and lower boundaries of cliffs, the outline corners of cave entrances, the center points of depression bottoms, and the tops of rock pillars. These feature points are spatially stable and exhibit significant geomorphic characteristics in the canyon and cliff area, unaffected by satellite signal obstruction, and can serve as core markers for regional spatiotemporal registration. All extracted geomorphic feature points underwent rigorous verification of their spatial coordinates and elevation information. Multiple verification methods were used to eliminate abnormal points caused by scanning errors, image blurring, and complex terrain, resulting in a set of geomorphic feature points with accurate spatial coordinates, reliable elevation information, and clear feature identification. Using this set of feature points as the core, a terrain-constrained spatiotemporal registration benchmark independent of satellite positioning signals is constructed. This effectively solves the problem of low registration accuracy caused by weak satellite signals and severe obstruction in canyon and steep cliff areas. The benchmark comprises two parts: a spatial benchmark and a temporal benchmark. The spatial benchmark is based on the spatial coordinates and elevation information of all verified geomorphic feature points, establishing a region-specific three-dimensional spatial coordinate system covering the entire canyon and steep cliff research area. Each geomorphic feature point serves as a spatial control node, clarifying the relative spatial relationships between nodes and forming a spatial reference framework that conforms to the actual terrain of the region. The temporal benchmark is based on the standard timestamps of the data collection at each geomorphic feature point, uniformly calibrating the time information of all geomorphic feature point-related data, eliminating clock deviations from different acquisition devices such as UAVs, radar stations, satellites, and rain gauges, and achieving unified alignment of all data in the temporal dimension. This terrain-constrained spatiotemporal registration benchmark provides a unified and accurate reference for the spatial alignment and temporal synchronization of multi-source data, ensuring accurate spatiotemporal matching of various data in subsequent processing. This fundamentally solves the problem of spatiotemporal misalignment of multi-source data in canyon and steep cliff areas, laying a solid spatiotemporal foundation for multi-source data fusion.

[0054] Based on high-precision topographic data within a standardized dataset, and combined with the actual surface conditions of canyon and steep cliff areas, three core factors—topographic slope, ground elevation, and land cover type—were comprehensively selected for topographic gradient zoning. Slope was divided into gentle, sloping, and steep regions according to actual distribution characteristics, accurately matching the slope variations of canyons and steep cliffs. Elevation was divided into high-altitude, mid-altitude, and low-altitude regions, reflecting the depth of the canyon and the elevation characteristics of the steep cliffs. Land cover type was divided into bare rock, vegetated, gully, and peak-cluster regions, adapting to the actual land cover conditions of the region. Through overlay analysis of the three types of topographic factors and spatial clustering algorithms, multiple topographic gradient zones with clear boundaries and relatively uniform internal topographic features and surface conditions were created. This ensures consistency in topographic, elevation, and land cover characteristics within each zone, avoiding data registration deviations caused by excessive differences in features within the region. Each topographic gradient zone serves as the basic processing unit for spatiotemporal registration of multi-source data, improving the accuracy and relevance of the registration. Within each terrain gradient zone, time synchronization and coordinate normalization are performed on multi-source data, including UAV airborne data, ground-based radar data, satellite remote sensing data, and ground rain gauge data. The time synchronization process uses a unified timestamp calibration to ensure that observation data from different devices at the same time form a precise time correspondence, eliminating time deviations. The coordinate normalization process transforms the coordinate system of all data into a three-dimensional spatial coordinate system of the terrain-constrained spatiotemporal registration benchmark, eliminating spatial position deviations caused by device observation perspectives and terrain undulations. This ensures that multi-source data from different sources and observation methods are completely unified within the same spatiotemporal framework, achieving precise spatiotemporal alignment of multi-source data. This allows subsequent weighted fusion to be carried out under the premise of spatiotemporal consistency, ensuring the scientific and rational nature of data fusion.

[0055] Based on multi-source data with accurate spatiotemporal registration, a topographic-meteorological coupled weight generation algorithm is applied for weighted fusion processing. This algorithm can achieve deep coupling of topographic features, meteorological matching degree, and data reliability, making the weight allocation and fusion results more consistent with the actual situation of canyon and steep cliff areas. For each topographic unit in the canyon and steep cliff area, the topographic stability factor, which reflects the stability of the topographic unit, is calculated based on the number and distribution characteristics of karst landform feature points such as internal solution funnels, steep cliffs, and cave entrances. Based on the consistency of observation values ​​of multi-source data such as UAV airborne data, ground-based radar data, and satellite remote sensing data at corresponding times, a meteorological matching degree factor, which reflects the meteorological data fit, is calculated. Based on factors such as the calibration status of each observation device, operational stability, noise level of the observation environment, and the influence of topographic obstruction, a data reliability factor, which reflects the reliability of data observation, is calculated. Simultaneously, based on typical terrain types in canyon and steep cliff areas such as canyons, cliffs, bare rock areas, and gullies, the weight coefficients corresponding to terrain stability factors, meteorological matching factors, and data reliability factors are dynamically adjusted to ensure that the sum of each coefficient is 1. This allows the weight allocation to accurately adapt to the observation characteristics and data quality of different terrain units. Subsequently, the comprehensive spatiotemporal weight of each terrain unit at each moment of the UAV rain enhancement operation is accurately calculated. Based on the calculated comprehensive spatiotemporal weight, a weighted fusion operation is performed on all multi-source observation data within the same spatiotemporal framework. This fully integrates the observation advantages of each data source, compensating for the observation blind spots and data biases of single data sources in canyon and steep cliff areas. For example, UAV-borne data can accurately reflect the near-surface meteorological characteristics inside the canyon, ground-based radar data can accurately monitor precipitation cloud systems, and satellite remote sensing data can achieve large-scale regional coverage. Through weighted fusion, the advantages of each data source are complemented, allowing the fused data to accurately and comprehensively reflect the meteorological, precipitation, and terrain characteristics of different terrain units and at different times in the canyon and steep cliff area. Ultimately, continuous and complete fused data is formed, providing high-quality basic data for subsequent quality control and ensuring the smooth implementation of the quality control process.

[0056] To address the terrain and data characteristics of steep canyon areas, a three-dimensional adaptive quality control threshold algorithm was applied to conduct quality control of fused data. This algorithm enables dynamic adaptive adjustment of the quality control threshold, completely resolving the issues of misjudgment and missed judgment in complex terrain caused by traditional fixed thresholds. Based on the algorithm requirements, and considering different terrain unit types after the area was divided, as well as different data types such as UAV airborne data, ground-based radar data, satellite remote sensing data, and ground rain gauge data, and different quality control indicators such as precipitation intensity, cloud particle concentration, elevation matching degree, and data continuity, corresponding adaptive quality control thresholds were determined. During threshold determination, the measured fluctuation coefficients of various data in different terrain units were dynamically calculated and adjusted, along with the historical misjudgment rate of each quality control indicator and the regional average misjudgment rate. Adjustment coefficients were used to reasonably balance the impact of terrain fluctuations on the data and the actual requirements of quality control effectiveness, ensuring that the quality control thresholds accurately reflect the data fluctuation characteristics and quality control standards of different terrain units in the steep canyon area, thus improving the accuracy of quality judgment. Based on the determined adaptive quality control threshold, point-by-point judgment is performed on the fused data. Every spatiotemporal data point in the fused data is comprehensively traversed. The measured value of the point is accurately compared with the adaptive quality control threshold under the corresponding terrain type, data type and quality control index. Valid data and abnormal data are strictly distinguished. Measured values ​​within the reasonable range are judged as valid data, and measured values ​​outside the reasonable range are marked as abnormal data, so as to achieve a comprehensive and thorough screening of the fused data. A scientific hierarchical processing strategy is adopted for marked anomalous data. For isolated single-point anomalous data, the neighborhood spatiotemporal interpolation method is used, selecting spatially adjacent terrain units and temporally adjacent effective data points as interpolation samples. The interpolation correction value is generated by weighted averaging and replaces the measured value of the anomalous point to ensure the continuity and consistency of the data in the spatiotemporal dimension. For continuous anomalous data in areas such as deep canyons and around steep cliffs, the anomalous range is first accurately located by using terrain-constrained spatiotemporal registration benchmarks. Then, the multi-source data cross-validation method is used to extract the corresponding observation values ​​from different data sources at the same time for the same terrain unit. The degree of deviation between the data sources is compared for secondary verification, and the fused data is accurately corrected. Severe anomalous data that cannot be corrected after repair and verification is strictly removed, and the cause, location and processing method of the anomalous data are marked in detail in the data record to achieve standardized and refined processing of anomalous data. Through comprehensive and high-standard quality control, the final output is fused quality control data that meets the requirements of spatiotemporal consistency and reliability. This data can accurately reflect the complex meteorological and precipitation characteristics of the canyon and steep cliff area, providing high-precision and high-reliability data support for all aspects of the drone rain enhancement operation in the area, such as flight path planning, operation altitude adjustment, catalyst timing selection, and operation effect evaluation. This effectively improves the scientific nature, accuracy, and effectiveness of drone rain enhancement operations in the canyon and steep cliff area.

[0057] This embodiment addresses the terrain and spring rain enhancement operations characteristics of the karst canyon and steep cliff region in South China, strictly adhering to the method of this invention for multi-source rain enhancement data fusion quality control. Standardization processes unify the heterogeneous multi-source data, a dedicated terrain-constrained spatiotemporal registration benchmark is constructed to address weak satellite signals, and refined spatiotemporal registration is achieved by combining three major terrain factors. A terrain-meteorological coupled weight generation algorithm enables complementary weighted fusion, and a three-dimensional quality control threshold adaptive algorithm completes dynamic and hierarchical quality control. The entire process is adapted to the terrain and observation characteristics of the canyon and steep cliff region, effectively resolving data deviation and anomalies. The final output fused quality control data meets spatiotemporal consistency and reliability requirements, providing high-precision data support for the planning and implementation of UAV rain enhancement operations in this region and improving the overall operational effectiveness.

[0058] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A quality control method for UAV rain enhancement data fusion in complex karst terrain, characterized in that, The specific steps of this method are as follows: S100. Data preprocessing: Collect multi-source raw data in the karst region and perform format standardization, missing value imputation and outlier preliminary removal to form a standardized dataset; S200, Benchmark Construction: Identify and extract karst landform feature points from the standardized dataset. Based on the spatial coordinates and elevation information of the landform feature points, establish a terrain-constrained spatiotemporal registration benchmark that does not rely on satellite positioning signals. The karst landform feature points are landmark locations in the karst region that have unique topographic features, stable spatial locations, and are easily identifiable. Specifically, they include the center point of the solution funnel, the inflection points of the upper and lower boundaries of the steep cliff, the outline corner points of the cave entrance, the center point of the bottom of the depression, and the feature points at the top of the rock column. S300, Spatiotemporal Registration: Based on the terrain slope, elevation and land cover type, the terrain gradient is divided into zones. Within each zone, time synchronization and coordinate normalization are performed on multi-source data based on the terrain-constrained spatiotemporal registration benchmark, so that data from different sources are unified in the same spatiotemporal framework. S400, Data Fusion: A terrain-meteorological coupled weight generation algorithm is applied to calculate the comprehensive spatiotemporal weight of each terrain at the corresponding time. Based on this comprehensive spatiotemporal weight, multi-source data within a unified spatiotemporal framework are weighted and fused to form continuous and complete fused data. The mathematical expression of the terrain-meteorological coupled weight generation algorithm is: ； in, The combined spatiotemporal weight of the terrain unit in the i-th row and j-th column at time t; For terrain stability factor; Meteorological matching factor; For data reliability factor; , , These are the weighting coefficients for the terrain stability factor, meteorological matching factor, and data reliability factor, respectively. S500 Quality Control: The three-dimensional quality control threshold adaptive algorithm is applied to determine the adaptive quality control threshold for the corresponding terrain and data type. Based on the adaptive quality control threshold, point-by-point judgment and anomaly handling are performed on the fused data, and the fused quality control data is output.

2. The method for quality control of UAV rain enhancement data fusion for complex karst terrain as described in claim 1, characterized in that, In step S100, data preprocessing, the multi-source raw data specifically includes UAV-borne data, ground-based radar data, satellite remote sensing data, ground rain gauge data, and terrain data. Specifically, the UAV-borne data is collected through multiple sensors mounted on the UAV, covering relevant meteorological and operational trajectory data for the karst region; the ground-based radar data is collected through fixed ground-based radar stations deployed within the karst region, acquiring precipitation-related data for the region; the satellite remote sensing data is collected collaboratively by polar-orbiting and geostationary satellites, acquiring cloud system and image-related data for the region; the ground rain gauge data is collected through automatic rain gauges deployed in different terrain areas within the study area, acquiring precipitation-related data for the region; and the terrain data is mainly high-precision DEM terrain data, acquired through a combination of scanning with UAV-borne LiDAR sensors and downloading publicly available terrain data.

3. The method for quality control of UAV rain enhancement data fusion for complex karst terrain as described in claim 1, characterized in that, In step S200, the benchmark construction, the terrain-constrained spatiotemporal registration benchmark is constructed with the extracted karst landform feature points as the core, without relying on satellite positioning signals. The core content includes two parts: a spatial benchmark and a temporal benchmark. The spatial benchmark is based on the spatial coordinates and elevation information of all verified landform feature points to establish a region-specific three-dimensional spatial coordinate system. Each landform feature point is used as a spatial control node to determine the relative spatial position relationship between each node, forming a spatial reference frame covering the entire karst study area. The temporal benchmark is based on the standard timestamps of the data collection of each landform feature point to uniformly calibrate the time information of all landform feature point related data and eliminate clock deviations of different acquisition devices.

4. The method for quality control of UAV rain enhancement data fusion for complex karst terrain as described in claim 1, characterized in that, In step S300, spatiotemporal registration, the topographic gradient partitioning is based on the topographic data in the standardized dataset, combined with the actual surface conditions of the study area, and comprehensively divided according to topographic slope, ground elevation and surface cover type. The entire karst study area is divided into different topographic gradient partitions. When dividing, the slope is used to distinguish gentle areas, sloping areas and steep areas; the elevation value is used to distinguish high-altitude areas, medium-altitude areas and low-altitude areas; and the surface cover type is used to distinguish bare rock areas, vegetation-covered areas, valley areas and peak cluster areas.

5. The method for quality control of UAV rain enhancement data fusion for complex karst terrain as described in claim 1, characterized in that, In step S500, during quality control, the mathematical expression of the three-dimensional quality control threshold adaptive algorithm is: ； in, The adaptive threshold is defined for the k-th type of terrain unit, the l-th type of data type, and the m-th quality control indicator. This serves as the global benchmark threshold for quality control indicators; is the measured fluctuation coefficient of the l-th type of data within the k-th type of terrain unit; This represents the average fluctuation coefficient of the l-th type of data across the entire region. The historical misjudgment rate of the m-th quality control indicator within the k-th terrain unit; Let m be the average misjudgment rate of the m-th quality control indicator across the entire region; , This is the adjustment coefficient.

6. The method for quality control of UAV rain enhancement data fusion for complex karst terrain as described in claim 1, characterized in that, In step S500, during quality control, the main contents of point-by-point judgment and anomaly handling are as follows: Point-by-point judgment: Traverse each spatiotemporal data point in the fused data, compare the measured value with the adaptive quality control threshold under the corresponding terrain type, data type and quality control index. When the measured value is within the reasonable range of the adaptive quality control threshold, it is judged as valid data; when it exceeds the reasonable range, it is marked as abnormal data. Anomaly Handling: A tiered processing strategy is adopted for marked anomalous data: For isolated single-point anomalies, the neighborhood spatiotemporal interpolation method is used to perform numerical repair by utilizing the spatiotemporal correlation of valid data around the anomaly point; for continuous anomalous areas, the anomaly range is first located by using terrain-constrained spatiotemporal registration benchmarks, and then the multi-source data cross-validation method is used to correct the anomalies by combining multi-source data; for severely anomalous data that cannot be repaired, they are removed, and the cause of the anomaly and the handling method are marked in the data record.

7. A method for quality control of UAV rain enhancement data fusion in complex karst terrain as described in claim 6, characterized in that, The neighborhood spatiotemporal interpolation method uses the spatial location and timestamp of the data points marked as anomalies as a benchmark, selects eight spatially adjacent terrain units and two temporally adjacent valid data points as interpolation samples, and generates interpolation correction values ​​through a weighted average method to replace the measured values ​​of the anomalies. The multi-source data cross-validation method extracts corresponding observation values ​​from different sources for multi-source fused data of the same terrain unit and the same time. By comparing the degree of deviation between the data from each source, the reliability of the fused data is verified a second time. The preset range of the deviation of the multi-source data is ±15% of the mean of the data from each source. When the deviation between the single source data and the mean of the multi-source data exceeds the preset range, the fusion result is reviewed and corrected.

Images