Surveying and mapping data processing method and system based on cloud computing

Abstract

The invention relates to the technical field of surveying and mapping data processing, and discloses a surveying and mapping data processing method based on cloud computing, and the method comprises the steps: obtaining topographic elevation, topographic slope and environmental parameters, carrying out the multi-source fusion and feature extraction based on the data, obtaining a preliminary feature set, carrying out the dynamic correction and integrity verification of the preliminary feature set, and obtaining a final feature set; generating a correction feature set; the method comprises the following steps: obtaining an optimized feature set through residual analysis, dividing high-frequency feature cache, constructing a partition feature set, carrying out synchronization verification on partition data, obtaining a consistency feature set, executing multi-thread integration, forming a global integration feature set, completing abnormal region reconstruction and optimization, generating a high-precision feature set, and finally, carrying out parallel task decomposition and accelerated integration to obtain a high-precision feature set. And outputting high-precision surveying and mapping data. The method solves the problem that the scheduling precision is limited under multi-source complex data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of surveying and mapping data processing, and particularly relates to a surveying and mapping data processing method and system based on cloud computing. BACKGROUND

[0002] With the rapid advancement of digital projects such as smart cities, land space planning, and disaster warning systems, the demand for high-precision topographic surveying and mapping data is increasing. Surveying and mapping data processing, as an important branch of geographic information science and engineering technology, not only provides support for basic geographic information platforms, but also plays a key role in many scenarios such as mineral exploration, urban renewal, and ecological monitoring. Especially in complex terrain areas or large-scale scenarios, how to achieve centimeter-level, even sub-centimeter-level surveying and mapping results has become one of the core goals in the process of digital transformation of surveying and mapping engineering.

[0003] In existing surveying and mapping data processing methods, a common implementation route is to grid the original elevation data in grid units, and extract key factors such as slope, roughness, and vegetation coverage through morphological analysis or classification algorithms to establish a topographic type discrimination model. Taking a random forest classification model as an example, multi-dimensional features are input into the trained discrimination model to complete the identification and block label division of the landform morphology. However, this method assumes that the spatial structure of the data is uniform, and a fixed analysis scale is divided to perform feature extraction and classification, which causes accuracy to decline in areas with high elevation mutations, complex features, or discontinuous boundaries. In addition, most of the processes still rely on centralized computing architecture, which is difficult to fully adapt to parallel processing frameworks in cloud environments, resulting in insufficient processing efficiency and limited data scale.

[0004] In summary, the existing technology has the problem of limited scheduling accuracy under multi-source complex data. SUMMARY

[0005] The present application provides a surveying and mapping data processing method and system based on cloud computing to solve the problem of limited scheduling accuracy under multi-source complex data.

[0006] In a first aspect, to solve the above technical problems, the present application provides a surveying and mapping data processing method based on cloud computing, comprising: obtaining terrain elevation, terrain slope, and environmental parameters; performing multi-source data fusion and feature extraction according to the terrain elevation, terrain slope, and environmental parameters to obtain a preliminary feature set; performing dynamic correction and integrity verification according to the preliminary feature set to obtain a corrected feature set; performing terrain residual analysis according to the corrected feature set to obtain an optimized feature set; According to the optimization feature set, high-frequency feature cache partitioning is performed to obtain a partition feature set; According to the partition feature set, synchronicity verification is performed to obtain a consistency feature set; According to the consistency feature set, multi-thread feature integration processing is performed to obtain a global integrated feature set; According to the global integrated feature set, abnormal area feature reconstruction and high-precision optimization are performed to obtain a high-precision feature set; According to the high-precision feature set, parallel task decomposition and accelerated feature integration are performed to obtain high-precision surveying and mapping data.

[0007] Preferably, according to the terrain elevation, the terrain slope and the environmental parameters, multi-source data fusion and feature extraction are performed to obtain a preliminary feature set, including: According to the terrain elevation, the terrain slope and the environmental parameters, rasterization mapping and resampling processing are performed to obtain a spatially consistent multi-source surveying and mapping feature matrix; Based on the random forest algorithm, feature importance evaluation and terrain type classification are performed on the spatially consistent multi-source surveying and mapping feature matrix to obtain a fusion feature set of different regional terrain forms; Based on the support vector machine algorithm, integrity judgment is performed on the fusion feature set of different regional terrain forms. If the judgment result is that the feature dimension is not complete, the regional feature is excluded. If the judgment result is complete, the regional feature is weighted and fused with the environmental factors to obtain a preliminary feature set.

[0008] Preferably, according to the preliminary feature set, dynamic correction is performed to obtain a corrected feature set, including: Obtain a terrain slope correction coefficient; According to the preliminary feature set, regional slope value extraction is performed to obtain a regional slope value feature set; According to the regional slope value feature set and the terrain slope correction coefficient, correction is performed to obtain a corrected feature set.

[0009] Preferably, according to the corrected feature set, terrain residual error analysis is performed to obtain an optimization feature set, including: According to the corrected feature set, residual error distribution extraction and spatial analysis are performed to obtain residual error spatial feature values; If the residual error spatial feature values are greater than or equal to a preset dynamic threshold, the corrected feature set is optimized based on the random forest algorithm to obtain an optimization feature set.

[0010] Preferably, according to the optimization feature set, high-frequency feature cache partitioning is performed to obtain a partition feature set, including: The optimization feature set is divided into a terrain grid according to a preset spatial scale to obtain initial partition feature data; According to the initial partition feature data, high-frequency region identification and cache optimization are performed to obtain an updated partition feature structure; Based on the XGBoost algorithm, the updated partition feature structure is subjected to optimal configuration reasoning and partition reconstruction to obtain a partition feature set.

[0011] Preferably, the synchronization verification according to the partition feature set obtains a consistent feature set, including: According to the partition feature set, a partition task structured configuration is performed to obtain a structured task unit set; According to the structured task unit set, boundary continuity synchronization verification is performed to obtain a partition boundary region that needs to be optimized; Based on the CatBoost algorithm, the partition boundary region that needs to be optimized is subjected to consistency scoring and feature screening to obtain a consistent feature set.

[0012] Preferably, the multi-thread feature integration processing according to the consistent feature set obtains a global integrated feature set, including: According to the consistent feature set, multi-thread concurrent integration processing is performed to obtain a preliminary integrated feature set; According to the preliminary integrated feature set, boundary continuity optimization processing is performed to obtain a smooth optimized integrated feature set; According to the smooth optimized integrated feature set, load balancing and global integration output are performed to obtain a global integrated feature set.

[0013] Preferably, the abnormal region feature reconstruction and high-precision optimization according to the global integrated feature set obtain a high-precision feature set, including: According to the global integrated feature set, abnormal region detection and extraction are performed to obtain a preliminary high-precision feature set; According to the preliminary high-precision feature set, consistency verification and feature reconstruction processing are performed to obtain an optimized high-precision feature set; According to the optimized high-precision feature set, boundary redrawing and precision verification optimization are performed to obtain a high-precision feature set.

[0014] Preferably, the parallel task decomposition and accelerated feature integration according to the high-precision feature set obtain high-precision surveying and mapping data, including: According to the high-precision feature set, task decomposition and parallel integration processing are performed to obtain preliminary surveying and mapping data; According to the preliminary surveying and mapping data, consistency verification is performed to obtain an optimized feature integration result; According to the optimization feature integration result, task decomposition and boundary reorganization operation are performed to obtain final high-precision surveying and mapping data.

[0015] In a second aspect, the present application provides a cloud computing-based surveying and mapping data processing system, comprising: A data acquisition module is configured to acquire terrain elevation, terrain slope and environmental parameters. A preliminary feature set module is configured to perform multi-source data fusion and feature extraction according to the terrain elevation, terrain slope and environmental parameters to obtain a preliminary feature set. A corrected feature set module is configured to perform dynamic correction and integrity verification according to the preliminary feature set to obtain a corrected feature set. An optimized feature set module is configured to perform terrain residual analysis according to the corrected feature set to obtain an optimized feature set. A partitioned feature set module is configured to perform high-frequency feature cache partitioning according to the optimized feature set to obtain a partitioned feature set. A consistent feature set module is configured to perform synchronization verification according to the partitioned feature set to obtain a consistent feature set. A globally integrated feature set module is configured to perform multi-thread feature integration processing according to the consistent feature set to obtain a globally integrated feature set. A high-precision feature set module is configured to perform abnormal area feature reconstruction and high-precision optimization according to the globally integrated feature set to obtain a high-precision feature set. A high-precision surveying and mapping data module is configured to perform parallel task decomposition and accelerated feature integration according to the high-precision feature set to obtain high-precision surveying and mapping data.

[0016] In a third aspect, the present application further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the cloud computing-based surveying and mapping data processing method of any one of the above.

[0017] In a fourth aspect, the present application further provides a computer readable storage medium comprising a stored computer program, wherein the computer program controls the device where the computer readable storage medium is located to execute the cloud computing-based surveying and mapping data processing method of any one of the above when the computer program is running.

[0018] Compared with the prior art, the present application has the following beneficial effects: (1) In the initial feature construction stage, based on the obtained terrain elevation, terrain slope and environmental parameters, a spatially consistent multi-source surveying and mapping feature matrix is constructed, the consistency and alignability of the data are improved through rasterization mapping and resampling processing, then the importance of the fused feature matrix is evaluated and the terrain type is classified by introducing the random forest algorithm, the integrity is judged by combining the support vector machine algorithm, and the qualified regional features and environmental factors are weighted and fused to form the preliminary feature set. This step effectively improves the fusion accuracy and structural robustness of surveying and mapping data under different regional scales, and can adapt to the feature difference expression demand under complex terrain conditions.

[0019] (2) In the dynamic correction stage, according to the regional slope information in the preliminary feature set, the slope feature value of each space unit is extracted, and the corresponding terrain slope correction coefficient is matched to dynamically adjust the regional slope and form the corrected feature set. This processing process provides differential compensation for regions with steep slope changes or edge transitions, effectively improving the continuity and consistency of feature data in spatial structure, and providing a stable foundation for subsequent residual detection and feature optimization.

[0020] (3) In the space division and cache optimization stage, the optimized feature set is divided into multiple initial grid regions by a preset spatial scale, high-frequency distribution features are extracted combined with regional access frequency, and an optimized mapping relationship between access frequency and computing resource allocation is constructed based on the XGBoost algorithm to generate updated partition feature structure. This step can dynamically adjust the data cache structure according to the regional heat, realize the coordination between fine allocation of cache resources and parallel task division, effectively alleviate the access delay of hot regions, and improve the concurrent processing capability of distributed computing.

[0021] (4) In the feature integration and output stage, task decomposition and boundary reorganization operations are performed according to the optimized feature integration result, and for the integration time exceeding the preset threshold or the boundary feature jumping region, the task granularity and boundary structure are dynamically adjusted, and the calculation unit and regional connection mode are redivided. This mechanism can ensure that the feature integration result realizes balance between data continuity and processing timeliness, and finally outputs high-precision surveying and mapping data that meets the requirements of spatial consistency and precision, providing a stable and reliable data foundation for subsequent applications.

[0022] In summary, the present application constructs a high-precision surveying and mapping data processing flow for cloud computing environment from feature extraction, parameter correction, task division to integration output. This flow can adapt to complex requirements such as multi-source data fusion, multi-thread execution and spatial heterogeneity processing, significantly improving the structural adaptation ability, computing efficiency and output precision of surveying and mapping tasks. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1is a cloud computing-based surveying and mapping data processing method flow diagram provided by the first embodiment of the present application; Figure 2 is a cloud computing-based surveying and mapping data processing system structure diagram provided by the second embodiment of the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0025] Referring to Figure 1 The first embodiment of the present application provides a cloud computing-based surveying and mapping data processing method, comprising the following steps: S11, acquiring terrain elevation, terrain slope and environmental parameters; S12, performing multi-source data fusion and feature extraction according to the terrain elevation, terrain slope and environmental parameters to obtain a preliminary feature set; S13, performing dynamic correction and integrity verification according to the preliminary feature set to obtain a corrected feature set; S14, performing terrain residual analysis according to the corrected feature set to obtain an optimized feature set; S15, performing high-frequency feature cache division according to the optimized feature set to obtain a partitioned feature set; S16, performing synchronicity verification according to the partitioned feature set to obtain a consistent feature set; S17, performing multi-thread feature integration processing according to the consistent feature set to obtain a global integrated feature set; S18, performing abnormal area feature reconstruction and high-precision optimization according to the global integrated feature set to obtain a high-precision feature set; S19, performing parallel task decomposition and accelerated feature integration according to the high-precision feature set to obtain high-precision surveying and mapping data.

[0026] In step S11, terrain elevation, terrain slope and environmental parameters are acquired; It is worth noting that in step S11, obtaining the terrain elevation, terrain slope and environmental parameters is the basic link in the survey data processing process, and its accuracy is directly related to the reliability and effectiveness of subsequent feature extraction and modeling. In order to ensure the specific feasibility and engineering landing of this step, the present application is based on a multi-source sensor integration platform, and through the cooperative operation of unmanned aerial vehicles, laser radars, satellite images and ground observation equipment, the terrain and environmental parameters of the target area are comprehensively perceived and data is collected.

[0027] First, in terms of terrain elevation acquisition, the present application uses a laser radar (LiDAR) system mounted on an unmanned aerial vehicle platform to scan the ground and generate high-density point cloud data. Each point in the point cloud data records its spatial three-dimensional coordinate value, and then a terrain surface model is constructed through the elevation dimension. In order to ensure the spatial consistency of the elevation information, the present application performs spatial gridding processing on the point cloud data, divides the survey area into a plurality of 10m by 10m grid cells, calculates the average elevation value in each grid cell, and outputs it as a representative value to form a terrain elevation grid map. Taking a certain hilly area as an example, the elevation recorded by a point in the original point cloud is 198 meters, the adjacent point is 204 meters, and the spatial distance between the two is 12 meters. Through grid aggregation processing, the final output grid elevation is the average value 201 meters.

[0028] The terrain slope is calculated and processed by the central difference method. For each target grid cell, the system selects the elevation values of the adjacent two grids in the east-west direction and the north-south direction, respectively, calculates the slope components in the two directions, and synthesizes the slope angle according to the above. Taking a 10m resolution terrain grid as an example, if the target point has an elevation of 210m to the east and 200m to the west, and no elevation difference in the north-south direction, then the slope in the east-west direction is 10m difference divided by 20m distance, resulting in a slope of 0.5, which is converted to an angle of 26.57 degrees, and the north-south direction is zero. The final slope angle of this point is 26.57 degrees. In addition, in order to improve the stability of the slope calculation, the system introduces a 3x3 neighborhood sliding window for each target grid, and uses a local mean filtering mechanism to calculate the average slope value in the neighborhood, In terms of obtaining environmental parameters, the application adopts a method of fusing multiple remote sensing and ground observation data for processing. Main environmental parameters include but are not limited to vegetation coverage, soil moisture, surface temperature, wind speed and direction, precipitation intensity and solar radiation intensity. Specifically, the vegetation coverage is obtained by analyzing the visible light image acquired by the optical camera through the NDVI vegetation index, the soil moisture is collected in real time by the wireless sensing node installed on the ground, the surface temperature is provided by the thermal infrared sensor, the wind speed and direction are collected by the laser wind radar or the ground weather station, and the precipitation intensity and solar radiation are obtained from the existing meteorological satellite database. Taking a typical plain area as an example, in a target area of 10 hectares, the NDVI calculation shows that the average vegetation coverage is 65%, and the ground sensing node shows that the average humidity is 75%. Such information is fused through time and space coordinates to form a standardized environmental factor matrix, which is input into the subsequent feature modeling process.

[0029] It should be noted that the application does not rely on a single sensing platform when obtaining terrain and environmental parameters, but builds a multi-source data fusion system. The data of different platforms are projected to a unified spatial reference system through time synchronization, coordinate transformation and data calibration process, and the Kalman filter and weight fusion algorithm are used to integrate the observation values of each sensor for the same target point to obtain the most reliable parameter output. Specifically, the Kalman filter is used to predict and update the observation data in continuous time series and eliminate the influence of random noise. Weighted fusion assigns weights according to the historical stability, accuracy level and observation confidence of each data source, and performs weighted averaging on multiple observation values for the same position to output the final value. For example, in a certain mountainous area, the laser radar measures the height as 520 meters, and the satellite remote sensing inversion result is 515 meters. The system evaluates that the observation accuracy of the radar is higher than that of the remote sensing, and assigns weights of 0.7 and 0.3 respectively. The final fused output height is 520 times 0.7 plus 515 times 0.3, and the calculation result is 518.5 meters. Through this method, the credibility and consistency of the fused parameters are effectively improved.

[0030] In summary, in step S11, the application realizes accurate acquisition of terrain elevation, terrain slope and environmental parameters through high-integration multi-source sensing network, rasterized data structure and multi-dimensional feature fusion algorithm.

[0031] In step S12, multi-source data fusion and feature extraction are performed according to the terrain elevation, the terrain slope and the environmental parameters to obtain a preliminary feature set, including: According to the terrain elevation, the terrain slope and the environmental parameters, rasterized mapping and resampling processing are performed to obtain a spatially consistent multi-source surveying and mapping feature matrix; Based on the random forest algorithm, the spatial consistent multi-source surveying and mapping feature matrix is subjected to feature importance evaluation and terrain type classification, to obtain a fusion feature set of different regional terrain forms. Based on the support vector machine algorithm, the fusion feature set of different regional terrain forms is subjected to integrity judgment, if the judgment result is that the feature dimension is not complete, the regional feature is eliminated, if the judgment result is complete, the regional feature is subjected to weighted fusion with the environmental factor, to obtain a preliminary feature set.

[0032] It is worth noting that after obtaining the terrain elevation, terrain slope and environmental parameters in the survey area, in order to improve the structural consistency, classification accuracy and spatial adaptability of the feature data, a multi-source data fusion and feature extraction process is further performed to generate a preliminary feature set as a key input basis for subsequent dynamic correction, regional division and data optimization. The processing procedure mainly includes three stages: one is to construct a spatial consistent surveying and mapping feature matrix, two is to complete feature importance evaluation and terrain type division based on a classification model, three is to judge the structural integrity of the fusion feature set and fuse the environmental factors, and finally form a preliminary feature set with discrimination ability and data integrity.

[0033] In the first stage, the present application adopts a raster mapping and resampling strategy to unify the data of different sources and different resolutions into a spatial feature matrix with consistent structure. The terrain elevation data is generated by laser radar system or remote sensing image inversion, and is represented in the form of discrete point cloud; the terrain slope is obtained by calculating the change rate of the elevation data in the horizontal direction. In the present application, the elevation difference of each grid cell and its adjacent grid points is selected, combined with the actual horizontal distance, and the central difference method is used to calculate the slope size, which is output in units of angle or percentage, to express the terrain fluctuation degree and slope change trend. It has spatial derivative property; the environmental parameters include vegetation coverage, soil moisture, wind speed, precipitation intensity, solar radiation and other multi-dimensional variables, some of which are obtained from remote sensing images, and some are obtained from meteorological stations or ground sensing nodes. Due to the different timestamps, spatial resolutions and reference systems of various data, unified processing is required before fusion. First, according to the set reference coordinate system of the survey area, all data are projected to a unified spatial reference, then the survey area is divided into fixed grids according to the set grid cell size (such as 10 meters by 10 meters), and all numerical features in each grid cell are statistically aggregated. For numerical data, mean or median is used, for missing areas, spatial interpolation or weighted completion of surrounding grids is used, so that all feature values can be expressed in each grid cell, forming a spatial consistent multi-source surveying and mapping feature matrix. Taking a typical hilly area as an example, the integrated region generates a grid of 1200 rows by 1500 columns, each grid cell contains five basic parameters of elevation, slope, NDVI value, land surface temperature and sunshine duration, and has structural integrity.

[0034] Based on the random forest algorithm, the feature importance of the constructed spatial consistent multi-source mapping feature matrix is evaluated and the terrain type is classified, and finally the fusion feature set of different regional terrain forms is obtained, and a stable and reliable input is provided for the subsequent integrity judgment process. In order to ensure the implementability and data support of the process, the invention uses manually annotated sample dataset as the source of training data, and uses remote sensing images and geological survey maps of typical survey areas as auxiliary basis, and combines with the results of field investigation to label representative terrain type areas, such as plains, hills, terraces, valleys and artificial platforms. More than ten thousand grid samples are selected in each area, and each sample grid is assigned a corresponding terrain type label to form a training set and a validation set.

[0035] The random forest algorithm samples the training set multiple times during runtime, and trains a decision tree for each feature combination in each round. Taking a certain grid in a survey area as an example, its feature vector is: elevation 312 meters, slope 18 degrees, NDVI 0.67, soil moisture 71, and sunshine hours 6.3 hours. In a decision tree, the first node may use "whether the slope is greater than 15 degrees" as the basis for division, and the grid meets the condition and enters the right subtree; the next node is set to "whether the elevation is greater than 300 meters", which also meets the condition and continues to classify downward, and finally the tree outputs "hilly area". The entire forest contains fifty trees, and the final classification result is "hilly area" by majority voting of the outputs of all trees.

[0036] At the same time, the system automatically calculates the feature importance during training. The calculation method is: count the frequency of a certain feature as a division basis in all decision trees, and multiply it by the average information gain of the node. For example, if the slope feature appears twenty times in all fifty trees, and each time brings an average information gain of 0.13, then its importance score is 20 times 0.13, which is 2.6. The system finally generates a complete feature importance ranking table, and takes the top five features as the fusion basis.

[0037] After classification, the system performs fusion based on the dominant features of each terrain type. Taking the hilly area as an example, the slope, elevation and NDVI have high weights, so the system extracts the three feature values of all grids in this area, and normalizes them to standardize the value range to between zero and one, and then performs weighted fusion with the feature importance score as the weight. If the normalized slope of a certain grid is 0.9, the elevation is 0.7, and the NDVI is 0.6, the corresponding weights are 0.45, 0.35 and 0.2 respectively, then the fusion feature value is calculated as: 0.9 times 0.45 plus 0.7 times 0.35 plus 0.6 times 0.2, which is equal to 0.405 plus 0.245 plus 0.12, and the final fusion feature is 0.77. This value together with the grid space position and classification label forms the fusion feature set.

[0038] In the third stage, in order to guarantee the structural rationality and dimensional integrity of the fusion features, the application introduces a support vector machine (SVM) algorithm to perform integrity determination and screening on the fusion feature set. The core basis of the integrity determination is the effectiveness of the feature dimension, that is, whether each grid cell contains all the high-weight discriminant features and participates in the contribution in the model calculation. In actual application, due to remote sensing image occlusion, sensor data missing or abnormal elimination processing, some grids may lack key environmental factors or terrain information. The SVM model identifies abnormal samples falling outside the discriminant boundary by boundary fitting on the existing fusion features in the integrity determination, and marks them as "incomplete feature dimension". The application eliminates the corresponding regions of such grids from the current processing batch to avoid their interference with the subsequent modeling process. For the grid feature units determined to be complete, weighted fusion processing with environmental factors is further performed. The weighting mechanism is dynamically adjusted according to the feature importance scores output in the classification stage to ensure that the optimal feature combination is used to express the grid ontology properties under different terrain types. For example, in the plateau block, the importance of slope and sunshine duration is higher than that of NDVI and humidity factor, and the system automatically increases the fusion weight of the former two and weakens the dependence on the latter two. The weighted feature vector constitutes the final preliminary feature set, which has the characteristics of clear spatial distribution, consistent parameter dimension and complete data quality.

[0039] Taking a survey area in the Qinling Mountains as an example, after the fusion processing of the region, a total of about 1.6 million effective grids are retained, of which 1.49 million grids are determined to be complete features, and about 110,000 grids are eliminated due to missing or abnormal regions. In the classification results, 800,000 grids belong to the forest valley region, about 420,000 belong to the ruin tableland, and the remaining 270,000 belong to the slope construction land. Each grid in the preliminary feature set contains 9 fusion feature dimensions, and is output grouped by type label, providing a unified data structure basis for the subsequent feature correction, regional partition and parallel integration process.

[0040] In summary, this step constructs a spatially consistent feature matrix through the fusion processing of terrain data and environmental factors, and outputs a high-quality preliminary feature set based on classification and integrity determination, which not only guarantees the stability of the data structure in the subsequent process, but also significantly improves the expression ability of the terrain complexity and environmental heterogeneity.

[0041] In step S13, according to the preliminary feature set, dynamic correction is performed to obtain a corrected feature set, including: obtaining a terrain slope correction coefficient; extracting a regional slope value according to the preliminary feature set to obtain a regional slope value feature set; According to the region slope value feature set and the terrain slope correction coefficient, correction is performed to obtain a corrected feature set.

[0042] It is worth mentioning that, after the construction of the preliminary feature set is completed, the region slope feature is further extracted and dynamically corrected, aiming to enhance the terrain expression accuracy and spatial variation consistency of the feature set. The main content of this step includes two aspects: one is to calculate the slope feature value of each spatial unit based on the existing elevation information in the preliminary feature set; the other is to introduce a slope correction coefficient related to the complexity of the region terrain to correct the calculated slope value in a directional manner, so as to obtain a corrected feature set with reasonable structure and reliable numerical value, which is used as the input data source for subsequent residual analysis and optimization judgment.

[0043] In the slope feature extraction link, the elevation data is used for spatial gradient analysis to extract the slope value of each grid cell. In the specific implementation, the central difference method is used to calculate the slope, that is, the elevation difference of the current grid and its adjacent east, west, north or south grid is selected, and the actual horizontal distance is combined to calculate the elevation change rate of the point in the X and Y axis directions, and then the combined slope angle is solved. The central difference method is a classical numerical differentiation method, and its calculation logic is to approximate the derivative value of the current point with the difference value on both sides, so as to improve the calculation accuracy and reduce the boundary error influence. For example, if the center point of a 10-meter grid cell is 210 meters high, the east point is 215 meters, and the west point is 208 meters, the elevation difference in the east-west direction is 7 meters, and the corresponding horizontal distance is 20 meters, the slope component is 0.35, and finally the square sum of the X and Y direction components is calculated, and the square root is taken to calculate the combined slope angle, and then converted to degree or percentage form and stored in the feature set.

[0044] Since slope calculation results are directly affected by elevation sampling quality, spatial resolution, and regional complexity, numerical deviations or distortions may occur under different landform structures. Therefore, this invention introduces a terrain slope correction coefficient mechanism to specifically correct the original slope values ​​for different types of terrain regions. This correction coefficient is not statically set but dynamically generated based on terrain complexity indicators, mainly referring to three aspects: regional elevation standard deviation, local elevation fluctuation frequency, and terrain type label. First, the elevation standard deviation of the area within each grid cell is calculated as a basic indicator for judging the degree of terrain undulation. If the standard deviation is high, it indicates that the terrain in the area changes drastically, and the initially calculated slope value is easily affected by high-frequency fluctuations, requiring amplification processing; conversely, if the standard deviation is low, the area is relatively gentle, and the original slope value tends to be overestimated, requiring appropriate compression. Second, the system counts the slope change frequency of each grid cell within the surrounding nine grid cells to determine whether it is a continuous steep slope, a gentle slope, or an abrupt boundary. Third, by referring to the terrain classification results output by the random forest algorithm in the previous step, a basic correction factor value is assigned to each type of terrain. For example, the upper limit of the correction factor for gully-dense terrain is set at 1.25, while that for terrace or plateau types is set between 0.85 and 0.95.

[0045] In the specific implementation process, the terrain slope correction coefficient is generated by weighted fusion of three indicators. For a certain area, if its elevation standard deviation is 15 meters, the slope variation frequency is moderate, and the terrain type is determined to be "ruins and gullies," then the initial slope correction coefficient generated by the system for this area is 1.18. This coefficient is then multiplied by the original slope value to obtain the corrected slope value, which is used to replace the slope field in the preliminary feature set to form a new corrected feature set. Taking a real-world example, in a survey area located in a typical loess hilly region in northern Shaanxi, the initially extracted slope value is 22.4 degrees, and the regional elevation fluctuation standard deviation is 19 meters, belonging to a medium-density gully landform. The system generates a slope correction coefficient of 1.15 according to the set strategy, and the corrected slope value is 25.8 degrees, significantly enhancing the ability to express the actual undulation characteristics of the area.

[0046] To avoid error accumulation introduced by the correction process, the application sets a maximum adjustment ratio threshold before slope correction. In actual setting, the maximum adjustment ratio threshold is in the range of ±20%, which is determined based on the experience of multiple typical terrain area experiments. The selection of this value mainly considers the following three points: first, according to the analysis of regional terrain change rate and actual measurement error amplitude, the fluctuation range of slope data under the disturbance of elevation data is not more than 20%; second, in the national topographic mapping standard, the upper and lower limits of the error of slope data under the condition of 1:2000 scale are generally controlled at about ±15%, considering the influence of environmental factors, the application is expanded to 20% to accommodate the differences brought by multi-source information; third, too low correction limit will lead to ineffective expression of high relief area, while too high correction amplitude is easy to cause slope surge distortion. Therefore, ±20% is taken as the reasonable and controllable correction limit, which not only guarantees the accuracy of terrain expression, but also controls the stability of the overall slope model.

[0047] In the specific implementation process, in each grid cell to be corrected, the system first calculates the preliminary extracted slope value P1, and combines the terrain complexity, spatial texture index and terrain category label to calculate the correction coefficient K. Then the corrected slope value P2 = P1 × K is generated by multiplication. On this basis, the system judges the relative change rate of P2 value and P1: if |P2 - P1| / P1 ≤ 0.2, it means that the correction meets the change amplitude requirement, and can be directly written into the correction result; if it exceeds the proportion limit, the P2 value is limited to P1 ± (0.2 × P1), that is, forced to compress within 20% to prevent abnormal amplitude. For example, if a slope value is 25 degrees, the calculated correction coefficient is 1.3, then P2 is 32.5 degrees, the change rate is 30%, which exceeds the range. At this time, the system automatically adjusts P2 to 25 + (0.2 × 25) = 30 degrees, and the compressed value is written into the correction feature set.

[0048] At the same time, in order to avoid the mutation of the boundary area in the calculation due to the lack of surrounding support, the application introduces a weighted window mechanism in the slope correction and spatial projection link, which guides the numerical smoothing of the boundary points to ensure continuous expression. This mechanism includes two parts: first, in the slope calculation process, for the cells located at the boundary of the grid, the spatial relationship with the adjacent grid is incomplete, so a weight coefficient is introduced when calculating the gradient to alleviate the deviation problem caused by the lack of point set; second, in the correction strategy, for the boundary points, the system weights the data of the center kernel points in the proportion of 7:3, which improves the numerical stability of the boundary area and avoids the appearance of artificial boundary or jump phenomenon.

[0049] For example, in a certain test area boundary section, the central point slope is 28.5 degrees, and the boundary point extraction value is 32.4 degrees. The system identifies that the difference with the adjacent point is too large. After performing weighted correction, assuming that the average of the adjacent points is 29.0 degrees, the final boundary point correction value is: 32.4x0.7 + 29.0x0.3 = 31.38 degrees. This value is closer to the overall trend in the continuity trend, avoiding the occurrence of isolated values.

[0050] In summary, by setting the maximum adjustment ratio threshold and supplementing the boundary weighted processing, the numerical rationality and spatial continuity of the terrain slope correction process are guaranteed. Not only does it prevent the numerical amplification problem caused by local complex terrain, but it also improves the expression stability of the overall surveying and mapping data at the grid level and the regional level, and constructs a more realistic and robust slope feature set, providing high-quality input basis for subsequent residual analysis and boundary optimization processing.

[0051] In step S14, according to the correction feature set, terrain residual analysis is performed to obtain an optimized feature set, including: According to the correction feature set, residual distribution extraction is performed, and spatial analysis is performed to obtain a residual spatial feature value. If the residual spatial feature value is greater than or equal to a preset dynamic threshold, the correction feature set is optimized based on a random forest algorithm to obtain an optimized feature set.

[0052] It is worth noting that in step S14, in this step, to realize the judgment of the expression accuracy of the current correction feature set in space, the present application proposes a method based on residual distribution extraction and spatial analysis to obtain a residual spatial feature value with spatial continuity description ability. This step first constructs a basic reference model, selects a high-precision elevation product in a public geographic database consistent with the survey area, such as the digital elevation model data provided by the National Basic Geographic Information Center. The system locates the spatial position of each grid cell in the current correction feature set and finds its corresponding position in the reference model to form a one-to-one spatial mapping relationship.

[0053] The process of calculating residual values is as follows: for each grid cell, the system extracts its actual elevation value and slope value in the correction feature set, denoted as H_actual and S_actual, and extracts the standard elevation value and slope value of the location from the reference model, denoted as H_ref and S_ref respectively. The system performs a difference operation on each pair of values to calculate the residual of the grid in elevation as ΔH = H_actual - H_ref, and the residual in slope as ΔS = S_actual - S_ref. The difference value can not only reflect the numerical difference between the current expression and the ideal terrain at a geographic location, but also retain specific physical meaning. For example, a grid cell in a mountainous area has an actual elevation of 528 meters and a standard value of 524 meters, so the residual is 4 meters and the slope residual is 3 degrees.

[0054] To further analyze the aggregation and distribution trend of residuals in space, the present application divides the analysis area in a sliding window manner, and sets each window to be 25 meters by 25 meters. In each window, the system performs statistical analysis on the residual values of all grids contained, calculates the mean, standard deviation and maximum value, and thereby constructs a two-dimensional residual space graph. The graph can not only be used to identify local outliers, but also provide input for subsequent dynamic threshold setting and optimization decisions. Taking a typical valley area as an example, the residual window mean is 3.2 meters and the standard deviation is 0.9 meters, indicating that there is a systematic deviation phenomenon in the area, which needs to be processed in the subsequent residual judgment link.

[0055] To further determine whether the residual space eigenvalue belongs to the acceptable range, the application introduces a dynamic threshold mechanism. The threshold is not a fixed constant, but is generated adaptively according to the feature complexity, terrain type and spatial distribution density in the analysis area. Specifically, in the preprocessing stage, the system will calculate the terrain complexity coefficient according to the slope variation degree of the region, and set the basic residual tolerance rate after normalizing the region elevation standard deviation and maximum slope difference. For example, in the medium complex terrain, the tolerance error is set to ±1.2 meters, in the high complexity area, it is adjusted to ±1.8 meters, and in the low complexity area, it can be tightened to ±0.8 meters. In this way, adaptive error adjustment is realized for different landforms. Taking a typical mountain slope area as an example, after the grid-level residual calculation is completed, a two-dimensional residual space graph is formed, and the system calculates the average elevation residual in the sliding window as 1.7 meters, and the standard deviation as 0.6 meters. After terrain complexity analysis, the dynamic threshold of this area is automatically set to 1.5 meters. Since the average residual is higher than the current threshold range, the system determines that there is a systematic deviation in the feature expression, and needs to enter the subsequent optimization processing flow. On the contrary, in another flat hilly area, the average elevation residual in the sliding window is 0.6 meters, and the standard deviation is 0.3 meters. This area is classified as a low complexity terrain, and the corresponding dynamic threshold is 0.8 meters. Since it is lower than the tolerance error range, the system does not trigger further optimization processing.

[0056] When it is detected that a grid or space piece area exceeds the dynamic threshold, the application uses a random forest algorithm to optimize and adjust the feature set of the area. The optimization process first constructs a training sample set by inputting the spatial position of the current correction feature point, the initial feature value, the residual value and the adjacent grid features. In the training process of the random forest algorithm, the system introduces the residual value as a label variable, the training goal is to predict the residual deviation trend of the point, and accordingly the optimal feature adjustment amplitude is back calculated. For example, the model identifies that some slope residuals are concentrated in the position of relative elevation mutation, and then deduces that the reason may be that the original elevation interpolation is insufficient, and suggests adjusting the slope expression downward.

[0057] The optimization strategy adopts a weighted correction mechanism, that is, according to the predicted residual distribution graph output by the model, a limited amplitude adjustment operation is performed on the target area feature value, and the maximum adjustment ratio is limited to within 15% to avoid introducing new round of error diffusion by optimization operation. At the same time, a sliding average strategy is also introduced for the boundary area to prevent boundary correction discontinuity. Finally, a new optimized feature set is formed, and the residual graph before and after optimization is compared to realize error adaptive adjustment closed loop.

[0058] In summary, this step improves the system's ability to identify abnormal areas and the robustness of feature expression by establishing a residual space analysis system and introducing a dynamic threshold decision mechanism. Especially in areas with sudden terrain changes, slope boundaries, or local reconstruction areas, it can effectively identify local error concentration zones and perform targeted optimization processing accordingly.

[0059] In step S15, according to the optimized feature set, high-frequency feature cache partitioning is performed to obtain a partition feature set, including: The optimized feature set is divided into terrain grids according to a preset spatial scale to obtain initial partition feature data. According to the initial partition feature data, high-frequency area identification and cache optimization are performed to obtain updated partition feature structure. Based on the XGBoost algorithm, the updated partition feature structure is subjected to optimal configuration reasoning and partition reconstruction to obtain a partition feature set.

[0060] It is worth noting that in step S15, according to the optimized feature set, high-frequency feature cache partitioning is performed to obtain the final partition feature set. The processing process mainly focuses on the division strategy of terrain grids, the identification method of high-frequency areas, the configuration mechanism of cache structure, and the application of partition reconstruction algorithm. In the specific implementation process, this step not only serves as a bridge connecting feature data analysis and subsequent parallel processing optimization, but also is a key link to improve the overall mapping data scheduling efficiency and response speed.

[0061] Firstly, the optimized feature set is taken as input data, and the target area is divided into terrain grids according to the preset spatial scale parameter. The principle of division is to ensure that each grid unit has a uniform feature scale in space and has the ability to be independently scheduled and processed in parallel. The spatial scale parameter used for division is a set of adjustable regional side length setting values, with units of meters. For example, in the field mesoscale topographic mapping task, a square grid of 100 meters by 100 meters is selected as the basic unit. After division, the initial partition feature data is formed, and each grid unit stores the corresponding terrain elevation, slope feature, soil moisture, and other optimized mapping features.

[0062] Next, the application performs high-frequency region identification and cache optimization processing steps according to the initial partition feature data. Specifically, the system performs data access frequency statistics operations for each partition grid, and the basis for statistics is the historical access records in the previous multiple processing tasks and the task scheduling information in the current integration process. For example, in the surveying and mapping data processing of a mountainous area, it is found that the characteristic points in the valley low-lying area are accessed many times in the previous residual optimization and spatial correction steps due to the steep change of slope and the influence of hydrological convergence. At this time, the system records the access count of the region grid as 82 times, while the surrounding stable plateau region has only 12 times. On this basis, the system sets a high-frequency identification threshold, which is set to 30 times, indicating that if the access frequency of a region exceeds 30 times, it is considered a high-frequency access region. The identified high-frequency region will be prioritized for optimization configuration of the cache structure, that is, it will be prioritized to allocate more storage space and higher access priority in the cache node.

[0063] After completing the high-frequency region identification, the application further adjusts and reconfigures the high-frequency partition feature structure, with the goal of ensuring that such regions have more stable access efficiency in subsequent task execution. To this end, the system uses an optimal configuration reasoning method based on the XGBoost algorithm to learn and train the updated partition feature structure and infer the optimal solution. The XGBoost algorithm is a boosting decision tree algorithm that performs well in handling high-dimensional sparse data and is suitable for modeling interactions between multiple variables. In the present application, the input of the XGBoost algorithm is multiple dimensional parameters such as feature density, access frequency, and historical task time consumption of each partition, and the output is cache allocation priority, task scheduling granularity, and partition structure reconstruction suggestions.

[0064] For example, a mountain slope region with partition number G12 has an elevation change range of 50 to 260 meters, an average slope of 28 degrees, and an access frequency of 66 times. After evaluation by the XGBoost algorithm, it is recommended to be subdivided into two sub-grids, each focusing on slope jump and stable slope segment, and configured with independent cache strategies. At the same time, the model also recommends configuring a dual-channel asynchronous cache interface to support multi-threaded task concurrent access. Through this combination of structure optimization and model reasoning, the system effectively avoids cache access conflicts and data delay problems, ensuring stable response in subsequent multi-threaded integration and boundary continuity verification processes.

[0065] Finally, after all the optimization configuration and partition reconstruction operations are completed, the system outputs a new partition feature set that not only has a clear structure, efficient access, but also has cache flexibility for subsequent large-scale surveying and mapping data scheduling. It is worth mentioning that before the partition feature set is output, the system will conduct a comprehensive check on the overall spatial range to ensure the continuity of feature expression at the boundary splicing. If it is found that there are some subgrids with feature mutations at the boundary, such as a jump of more than 30 meters in elevation value, the system will back to the previous round of XGBoost algorithm reasoning stage to regenerate the reconstruction recommendations for the region.

[0066] In summary, in step S15, by means of spatial division and high-frequency recognition mechanism of the optimized feature set, combined with the intelligent reasoning ability of XGBoost algorithm, the efficient transition from the original optimized feature to the structured partition feature is realized.

[0067] In step S16, according to the partition feature set, a consistency feature set is obtained by synchronicity verification, including: According to the partition feature set, a structured task unit set is obtained by partition task structured configuration; According to the structured task unit set, a partition boundary area that needs to be optimized is obtained by boundary continuity synchronization verification; Based on the CatBoost algorithm, a consistency feature set is obtained by consistency scoring and feature screening for the partition boundary area that needs to be optimized.

[0068] It is worth mentioning that in step S16, according to the partition feature set, a consistency feature set is obtained by synchronicity verification. This step mainly aims at the partition feature set obtained in the previous stage, and by establishing a structured task unit set, performing boundary continuity verification and feature consistency screening, the final feature set is ensured to have good continuity and logical consistency in spatial expression, providing complete structure and natural transition regional data support for subsequent feature integration and modeling.

[0069] Firstly, this step takes the partition feature set as the initial input data and performs task unit structured configuration. The core goal of this process is to combine partition data and computing resource allocation strategy to generate a set of schedulable and verifiable task units. In specific implementation, the system analyzes each feature partition, identifies its boundary grid, center feature position and spatial coverage range, and combines the grid structure information generated in the previous step to determine the boundary range and main processing area of each task unit. In order to facilitate subsequent multi-thread processing and cross-node synchronization, each task unit not only contains the core feature data of its own region, but also needs to mark the boundary point coordinates, boundary attribute values and historical identification labels between the region and its adjacent regions.

[0070] After the structured configuration is completed, the system performs a boundary continuity synchronization verification on the above-mentioned task unit set. The key of this verification process is to identify the feature value jump at the junction of each task unit, so as to judge whether there is a continuity fault in the spatial data. Based on the bidirectional comparison mechanism, the system extracts the shared boundary of the adjacent two task units as a feature vector set, and respectively calculates the local difference of the boundary points on the elevation, slope, texture and other dimensions.

[0071] In the initial stage of task unit division, each sub-region has an elevation raster matrix constructed at a fixed spatial resolution (such as one meter or five meters). For the corresponding grid points at the shared boundary, the elevation values in task unit A and task unit B are extracted, and the difference is calculated in grid points. For example, if the elevation of the nth boundary point in unit A is 126.8 meters and in unit B is 129.1 meters, then the local elevation difference is 2.3 meters. After calculating the difference values of all boundary points, the mean square deviation index is used for statistical induction to judge the continuity level of the overall boundary segment in the elevation dimension.

[0072] Secondly, the calculation of slope difference needs to rely on the spatial gradient of elevation data. The system uses the central difference method, that is, taking each grid point as the center, combining the elevation values of the adjacent grids in the east-west direction and the north-south direction, respectively calculating the change rate in the horizontal and vertical directions. Further, the direction derivatives are combined in the Euclidean norm to obtain the slope angle value. For example, if the slope of boundary point P in unit A is 15.6 degrees and in unit B is 19.3 degrees, then the slope difference of this point is 3.7 degrees. The statistical analysis method of the slope difference of all boundary points is the same as that of the elevation, and the average difference and the range are calculated to identify whether there is a slope mutation.

[0073] Finally, the calculation of texture difference is based on topographic image data, and the local image texture features are extracted using the gray level co-occurrence matrix. The specific processing procedure is as follows: extract the image gray block in the fixed window range at each boundary point, construct the corresponding gray level co-occurrence matrix, and calculate the texture statistical indicators such as contrast, entropy and energy. Taking the "contrast" index as an example, if the texture contrast of the image window corresponding to boundary point P in unit A is 1.45 and in unit B is 2.02, then the texture difference is 0.57.

[0074] If the difference of the boundary point in each dimension exceeds the system set continuity threshold. The elevation change exceeds 3 meters, the slope change exceeds 10 degrees, and the texture contrast exceeds 0.8. Then determine that the boundary is an area that needs to be optimized, and mark it as "potential consistency loss area". Such areas will be processed in the subsequent steps to improve the overall consistency level.

[0075] After obtaining the partition boundary areas to be optimized, the system calls the CatBoost algorithm to perform consistency evaluation and feature screening on these boundary areas. As an algorithm based on gradient boosting decision trees, CatBoost can effectively handle data structures with both categorical and continuous features, and is particularly suitable for boundary feature classification, anomaly identification, and precision reconstruction. In practical applications, the system inputs the original feature vector of the boundary area into the CatBoost model, and combines its adjacent cell features, spatial position relationships, and historical matching scores to perform multiple rounds of training and prediction. The algorithm outputs a consistency score value for each boundary area and a feature retention suggestion. The score value is used to evaluate whether the boundary meets the preset continuity requirements (e.g., the consistency score needs to be higher than 0.85), and the retention suggestion is used to determine whether the boundary area needs to be reconstructed or merged.

[0076] For example, in a typical application scenario, two partitions located in the southeast region have a slope jump phenomenon at the boundary. The system detects that the average slope values on both sides are 18 degrees and 32 degrees, with a difference far exceeding the continuity threshold, so it is identified as an area to be optimized. The CatBoost model further evaluates its consistency score as 0.67, which is lower than the standard value of 0.85, and gives a feature screening suggestion of "merge and adjust the boundary grid density." Subsequently, the system performs feature fusion on the boundary area according to the model suggestion, including resampling the boundary grid, smoothing the elevation curve, and adjusting the boundary width of the sub-area, thereby eliminating the original jump and achieving smooth transition of the boundary. The final processed score is improved to 0.91, confirming that it meets the consistency standard and is included in the consistency feature set.

[0077] Through the above steps, the system can effectively improve the boundary consistency level between regions while ensuring the completeness of spatial data coverage, avoiding model fitting bias caused by feature jumps or grid faults. It is worth noting that to ensure the effectiveness of the CatBoost model, the system needs to complete the model training process in advance, the training data set needs to cover different terrain types and various partition feature combinations, and the model parameters such as learning rate, depth, and iteration number need to be optimized to improve the accuracy and stability of the prediction.

[0078] In summary, step S16 improves the expression quality of the feature set in terms of spatial continuity and semantic consistency through a series of fine operations such as structured task configuration, boundary continuity comparison, and CatBoost algorithm screening, and lays a solid foundation for subsequent global feature integration and high-precision mapping modeling.

[0079] In step S17, according to the consistency feature set, multi-thread feature integration processing is performed to obtain a global integrated feature set, including: According to the consistency feature set, a multi-thread concurrent integration processing is performed to obtain a preliminary integrated feature set; According to the preliminary integrated feature set, a boundary continuity optimization processing is performed to obtain a smooth optimized integrated feature set; According to the smooth optimized integrated feature set, a load balancing and global integration output is performed to obtain a global integrated feature set.

[0080] It is worth noting that first, after the consistency feature set is obtained, the system performs a preliminary partitioning to meet the prerequisite for parallel computing. The partitioning principle is based on the dual standards of feature value interval division and spatial coordinate division. Specifically, the system takes the terrain slope as the basis for partitioning and divides it into four slope intervals: 0-10 degrees, 10-20 degrees, 20-30 degrees, and more than 30 degrees. Each interval constitutes an independent sub-feature region, and combined with its corresponding spatial coordinate range, it further confirms the physical location range that needs to be processed by the thread. This ensures the functional independence of tasks between threads and avoids repeated processing of spatially adjacent regions.

[0081] Next, the system assigns independent threads to each sub-feature region for processing. The thread performs feature data merging operations during execution, mainly including grid boundary coordinate uniform processing, adjacent feature point integration, and local spatial structure repair. In specific execution, the thread first reconstructs the spatial index of each grid cell in the responsible region, unifies its coordinate expression and adjacent relationship, and calculates the spatial similarity index. As a typical example, if thread 1 is responsible for processing the 0-10 degree slope interval, containing 10,000 grid records, it converts the coordinate of the center point of each grid during processing, and constructs an eight-neighborhood spatial connection graph to ensure clear and conflict-free connection relationships between grids.

[0082] After the initial integration is completed, the system enters the boundary continuity optimization phase. This phase focuses on solving the possible jump problem at the boundary between the results of multi-thread processing. The specific method is to compare the numerical difference of the boundary grid of the adjacent two thread responsible areas, if the slope difference of the adjacent grids at both ends exceeds the set threshold of three degrees, it is determined that there is a boundary jump. The system introduces linear interpolation method or sliding window smoothing method to generate intermediate transition data. The system uses a five-point sliding window to smooth the calculation of the boundary jump section, and actually uses a weighted average method. Specifically, the weighted average refers to setting different weight coefficients for the current processing point and its adjacent multiple data points, then multiplying and summing the values and weights of each point, and finally obtaining the weighted result value with the total weight as the denominator. In this invention, it is assumed that there is a boundary grid with slope values of 9 degrees, 10 degrees, 13 degrees, 14 degrees, and 15 degrees, and the system sets the weight of the intermediate point, i.e. the current jump point, to 0.4, and the other four adjacent points to 0.15. The calculation process is to multiply 9 by 0.15, add 10 by 0.15, add 13 by 0.4, add 14 by 0.15, and add 15 by 0.15, and finally divide the result by the total weight 1. The calculation output value is about 12.3 degrees. This value is significantly lower than the original jump value of 15 degrees, and tends to be close to the average level of the areas before and after the boundary, thereby effectively alleviating the gradient mutation problem caused by multi-thread processing of the boundary, making the integrated results more natural, smooth and continuous in spatial expression.

[0083] During the optimization process, in order to avoid the reduction of processing efficiency caused by unbalanced thread load, the system monitors the progress of each thread task and resource utilization in real time. If it is found that the running time of a certain thread exceeds the average value or the amount of data processed is greater than the average task upper limit, the system will dynamically adjust the task allocation. For example, if thread 2 processes 15 to 30 degree data totaling 16,000, the processing time is three times longer than other threads, and the system will divide 6,000 data in the middle of the task to thread 3 for execution, relieving the load pressure of thread 2.

[0084] After forming the optimized integration feature set, the system also performs local feature resmoothing processing. The purpose of this link is to further improve the spatial trend consistency of regional expression on the basis of ensuring overall boundary continuity. Taking actual data as an example, in a mountainous area, there are 5,000 records with a slope greater than 30 degrees in a steep slope area, but the surrounding area is generally a 15-degree gentle slope. The system smooths the target area data through weighted average calculation. The specific method is to assign a weight of 0.6 to the current grid value, and the remaining adjacent grids are assigned a weight of 0.1. The final output value is distributed between 25 and 29 degrees, and a transition smoothing mechanism is used, i.e. a five-point weighted window containing the current grid and the adjacent grids above, below, left and right is used for mean reconstruction again, further compressing the mutation gradient.

[0085] Finally, the system integrates data according to the optimization of the execution of each thread, and outputs the global integrated feature set. Before output, the system performs difference analysis according to the boundary area of the output results of all threads to ensure seamless connection of data of each thread. If there are still extreme values or jump trends, interpolation reconciliation is performed. Taking a certain survey area as an example, after final reconciliation, the difference in slope between any two adjacent grids in the integrated result is not more than 2 degrees, and the overall spatial trend is continuous and smooth.

[0086] In summary, step S17 realizes the overall integration of spatial feature information through the processes of consistent feature set division, multi-thread concurrent integration, boundary continuity analysis and optimization, local feature re-smoothing, and final output structure unification. It also balances efficiency and quality, ensuring the integrity of the basic data structure required for high-precision data modeling in subsequent steps.

[0087] In step S18, according to the global integrated feature set, abnormal area feature reconstruction and high-precision optimization are performed to obtain a high-precision feature set, including: According to the global integrated feature set, abnormal area detection and extraction are performed to obtain a preliminary high-precision feature set. According to the preliminary high-precision feature set, consistency verification and feature reconstruction processing are performed to obtain an optimized high-precision feature set. According to the optimized high-precision feature set, boundary redrawing and precision verification optimization are performed to obtain a high-precision feature set.

[0088] It is worth noting that first, the global integrated feature set is taken as the main processing object, focusing on spatial continuity abnormal areas such as elevation mutation and slope steepness. The formation process of the global integrated feature set has ensured the coordination of the overall features in terms of multi-thread integration, boundary optimization, and load balancing. However, due to complex topography, discontinuous original data, or sparse sampling in local areas, there may still be areas with high error points or feature fragmentation. Therefore, in this step, the system needs to identify these problem areas through an abnormal area detection mechanism and implement targeted feature reconstruction and precision optimization operations to ensure that the final output survey features have high consistency and high precision.

[0089] In the actual processing process, the system first performs abnormal area detection and extraction operation. Specifically, the system identifies the local area with a large elevation difference by analyzing the rate of change of elevation in the global integrated feature set. For example, if the elevation difference between consecutive grid cells exceeds 30 meters, the system marks this area as an abnormal elevation area and defines a buffer boundary centered on it, for example, all feature points within a radius of 200 meters are listed as areas to be optimized. In one practical scenario, the system identifies a region with a continuous elevation difference of 48 meters in a mountain slope area, and further defines a processing range including three local sub-regions, each containing about 5000 to 8000 feature points.

[0090] After extracting the preliminary abnormal area, the system then performs feature reconstruction processing. This process takes the preliminary high-precision feature set as input and determines whether the extraction result of each sub-region deviates significantly from the original data through a consistency checking mechanism. The checking process uses a similarity calculation method, and sets the matching degree preset threshold to 90%. If the matching degree of the features in a certain region with the original elevation grid data is lower than the threshold, the system automatically marks it as inconsistent data and needs to enter the feature reconstruction stage. For example, the average elevation extraction value of a sub-region is 165 meters, while the average value of the original data is 150 meters, with an error of more than 10%, the system determines that feature reconstruction is needed. In the reconstruction process, the system uses a sliding window method to smooth the abnormal value points in the region, and specifically calculates the weighted average value of the mutation point and the average value of the adjacent five cells to make the reconstructed data consistent with the neighborhood trend. After processing, the average value of the region is adjusted to 152 meters, and the matching degree is improved to 92%, meeting the system's quality requirements.

[0091] After feature reconstruction, the system continues to perform high-precision optimization processing, focusing on the rationality of boundary division and the uniformity of feature point distribution. To avoid precision loss caused by ambiguous boundaries, the system uses a sub-region subdivision strategy to further refine the reconstructed region and re-performs boundary division and partition configuration. For example, in the aforementioned abnormal area, the original data structure divided into three sub-regions is optimized to five regions, two new independent sub-regions are added for the transition zone, and each is configured with about 2000 feature data. This optimization method ensures the structural continuity of the boundary region and improves the spatial details of feature expression.

[0092] It is worth noting that in the optimization after the reconstruction is completed, the system is not simply using static judgment, but again performing integrity verification and checking process, ensuring that the final feature set and the original terrain data meet the set accuracy requirements at global and local levels. Specifically, the system performs comparative analysis region by region, checks its spatial trend, consistency score and boundary accuracy. In one example, the five sub-regions after reconstruction and refinement are subjected to secondary matching degree test, all results exceed 93%, among which the boundary region with original matching degree of only 85% is improved to 94%, and finally a high-precision feature set with unified structure is formed.

[0093] In summary, in step S18, the application systematically completes the structure correction of the high-error area through abnormal area detection, feature reconstruction and boundary optimization.

[0094] In step S19, according to the high-precision feature set, parallel task decomposition and accelerated feature integration are performed to obtain high-precision surveying and mapping data, including: According to the high-precision feature set, task decomposition and parallel integration processing are performed to obtain preliminary surveying and mapping data; According to the preliminary surveying and mapping data, consistency checking is performed to obtain optimized feature integration results; According to the optimized feature integration results, task re-decomposition and boundary reintegration operations are performed to obtain final high-precision surveying and mapping data.

[0095] It is worth noting that in step S19, the application is based on the high-precision feature set formed in the preceding steps, and further performs task decomposition and accelerated integration operations, aiming to solve the scheduling bottleneck and limited processing accuracy problems existing in the final output stage of multi-source complex surveying and mapping data. This step adopts parallel computing and dynamic adjustment mechanism to ensure that the system still has fast and accurate data integration capability under high load scenarios, and outputs the final surveying and mapping results that meet the requirements of spatial continuity and boundary accuracy.

[0096] Firstly, the system pre-processes the spatial distribution and feature type of the high-precision feature set, evaluates the distribution balance and feature fluctuation degree of the current data to be processed among various spatial blocks. In order to facilitate processing, the system divides the high-precision feature set into multiple calculation units, each unit containing data regions with similar geomorphic structure or similar elevation gradient. The division strategy is based on grid index, slope distribution and spatial density index, for example, in a mountainous data set containing about one hundred thousand feature points, the system may divide the data into ten units, each unit containing ten thousand data records, and ensuring that the data fluctuation degree within each unit is within the set range.

[0097] After the initial division, the system enters the parallel task scheduling phase. The scheduling mechanism takes into account the processing capacity of each computing node, the current load state, and the differences in the data volume of each task, and performs dynamic thread allocation for each unit. In one implementation, node A has a processing capacity of 500 records per second, and node B has a processing capacity of 800 records per second. The system allocates the units with larger data volumes to node B to achieve optimal overall computing time. Each parallel task calculates the average slope, elevation variance, and spatial gradient within the region during execution, and performs preliminary integration. This process is called parallel integration processing, and the output is the preliminary mapping data.

[0098] After the preliminary integration results are generated, the system immediately performs consistency verification on the data. The verification process focuses on the boundary continuity between adjacent units in space and the deviation control of the overall data from the original mapping data. For example, if the elevation difference between the boundaries of two adjacent units exceeds the system's set threshold of 3 meters, or the data fitting error exceeds the set upper limit of 3%, the data unit will be marked as a deviation anomaly area and subsequent optimization operations will be performed. At this time, the system will automatically call up the preliminary data of the area and rebuild the feature interpolation model to perform boundary fitting or local smoothing on the elevation jump area.

[0099] In one specific example, the elevation change in a boundary area of the preliminary mapping data jumps from 120 meters to 145 meters, and the spatial gradient analysis finds that the change rate is much higher than the average value of the surrounding area. The system determines that this jump is not reasonable in terms of landform, and uses the moving average method to reconstruct the data in the boundary area. After processing, the elevation change in this area is adjusted to 120 meters to 130 meters, and the fitting error is reduced from 6.2% to 2.8%, meeting the requirements of the final output.

[0100] After all the abnormal areas are optimized, the system performs spatial consistency analysis again on the optimized feature integration results and builds a new task decomposition strategy. This strategy reassigns task weights and thread structures based on the time distribution of the previous parallel processing and the frequency of abnormal areas to further reduce the overall integration time. For example, if three of the ten tasks in the preliminary processing phase, tasks 3, 6, and 9, take significantly longer, the system will increase the number of parallel threads for these three units in the next round of scheduling, or split them into smaller sub-tasks and execute them in parallel on multiple nodes, achieving both load balancing and efficiency improvement.

[0101] The final system forms an adjusted high-precision feature set and enters an output integration stage. The stage takes boundary feature stability as the basis for judgment, and further evaluates the variation degree of boundary points in the elevation, slope, texture and other dimensions through the random forest algorithm to determine the output boundary range. In a typical example, when the system evaluates the surveying and mapping data of a certain area, it determines the area with an elevation variation of less than 2 meters and a slope fluctuation of less than 4 degrees as the final output boundary through boundary feature variation threshold analysis, eliminates 34 jump feature points, and corrects 250 boundary data. This method ensures that the output surveying and mapping results have high boundary continuity, high spatial integrity and high-precision feature expression capability.

[0102] The final generated high-precision feature set not only contains stable feature parameters such as elevation, slope, texture and residual error of each region, but also carries the time consumption, complexity label and continuity score of each sub-region in the integration process. The system identifies which regions are computing hotspots and which regions have the risk of boundary discontinuity or residual accumulation through these labels. According to the identification result, the system can perform fine task decomposition, independently split high-complexity regions into smaller task units and assign them to different threads for execution, thereby avoiding the situation where a thread is occupied for too much time by a complex region, slowing down the overall integration speed.

[0103] In summary, the cloud computing-based surveying and mapping data processing method provided by the present application constructs a full-link optimization process from data acquisition to high-precision surveying and mapping data output when facing complex mountainous terrain, multi-source heterogeneous data input and dynamic scheduling requirements. Through the linkage implementation of step-by-step feature extraction, dynamic correction, regional reconstruction, parallel integration and task rescheduling mechanism, the system effectively solves the problems of feature discontinuity, inaccurate scheduling and low processing efficiency in the prior art. Especially in high-frequency feature change and complex boundary regions, the present application realizes fine-grained identification and rapid response to terrain changes through multi-algorithm cooperation and module-level regulation, thereby solving the problem of limited scheduling accuracy under multi-source complex data.

[0104] Reference Figure 2 The second embodiment of the present application provides a cloud computing-based surveying and mapping data processing system, comprising: A data acquisition module for acquiring terrain elevation, terrain slope and environmental parameters; A preliminary feature set module for performing multi-source data fusion and feature extraction based on the terrain elevation, terrain slope and environmental parameters to obtain a preliminary feature set; A corrected feature set module for performing dynamic correction and integrity verification based on the preliminary feature set to obtain a corrected feature set; An optimized feature set module for performing terrain residual error analysis based on the corrected feature set to obtain an optimized feature set; The partition feature set module is configured to perform high-frequency feature cache partitioning according to the optimization feature set, and obtain a partition feature set. The consistency feature set module is configured to perform synchronization verification according to the partition feature set, and obtain a consistency feature set. The global integration feature set module is configured to perform multi-thread feature integration processing according to the consistency feature set, and obtain a global integration feature set. The high-precision feature set module is configured to perform abnormal area feature reconstruction and high-precision optimization according to the global integration feature set, and obtain a high-precision feature set. The high-precision surveying and mapping data module is configured to perform parallel task decomposition and accelerated feature integration according to the high-precision feature set, and obtain high-precision surveying and mapping data.

[0105] It should be noted that the cloud computing-based surveying and mapping data processing system provided by the embodiments of the present application is used to perform all process steps of the cloud computing-based surveying and mapping data processing method of the above-mentioned embodiments, and the working principles and beneficial effects of the two are one-to-one correspondence, so they will not be repeated here.

[0106] The embodiments of the present application also provide an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a cloud computing-based surveying and mapping data processing program. The processor implements the steps in each of the above cloud computing-based surveying and mapping data processing method embodiments when executing the computer program, such as Figure 1 The steps S11 shown. Alternatively, the processor implements the functions of each module / unit in each of the above device embodiments when executing the computer program, such as the high-precision surveying and mapping data module.

[0107] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units can be a series of computer program instruction segments that can complete a specific function, which are used to describe the execution process of the computer program in the electronic device.

[0108] The electronic device can be a desktop computer, a notebook computer, a palm computer, and a smart tablet computer, etc. The electronic device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device, and can include more or fewer components than the above, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus, etc.

[0109] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is a control center of the electronic device, and connects various parts of the electronic device through various interfaces and lines.

[0110] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the electronic device by running or executing the computer program and / or modules stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, an image playing function, etc.), etc.; and the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory device.

[0111] The modules / units integrated in the electronic device, if realized in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer readable medium can include any entity or device, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. that can carry the computer program code. It should be noted that the contents included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0112] It should be noted that the above-described device embodiments are only schematic, and the units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment. In addition, the connection relationship between the modules in the device embodiment provided by the present application indicates that there is a communication connection between them, which can be realized as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0113] The above-described specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above-described specific embodiments are only for the specific embodiments of the present application and do not limit the protection scope of the present application. It is particularly pointed out that any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A cloud computing-based surveying data processing method, characterized in that, include: Obtain terrain elevation, terrain slope, and environmental parameters; Based on the terrain elevation, terrain slope and environmental parameters, multi-source data fusion and feature extraction are performed to obtain a preliminary feature set; Based on the preliminary feature set, dynamic correction and integrity verification are performed to obtain the correction feature set; Based on the corrected feature set, terrain residual analysis is performed to obtain the optimized feature set; Based on the optimized feature set, high-frequency feature cache partitioning is performed to obtain partitioned feature sets; Based on the partition feature set, a synchronization verification is performed to obtain a consistency feature set; Based on the consistent feature set, multi-threaded feature integration processing is performed to obtain a global integrated feature set; Based on the global integrated feature set, abnormal region features are reconstructed and optimized with high precision to obtain a high-precision feature set. Based on the high-precision feature set, parallel task decomposition and accelerated feature integration are performed to obtain high-precision mapping data.

2. The cloud computing-based mapping data processing method according to claim 1, characterized in that, The process involves multi-source data fusion and feature extraction based on the terrain elevation, terrain slope, and environmental parameters to obtain a preliminary feature set, including: Based on the terrain elevation, terrain slope, and environmental parameters, rasterization mapping and resampling are performed to obtain a spatially consistent multi-source mapping feature matrix; Based on the random forest algorithm, feature importance assessment and terrain type classification are performed on the spatially consistent multi-source mapping feature matrix to obtain a fusion feature set of terrain morphology in different regions. Based on the support vector machine algorithm, the completeness of the fused feature set of the terrain morphology of different regions is judged. If the judgment result is that the feature dimension is incomplete, the feature of that region is removed; if the judgment result is complete, the feature of that region is weighted and fused with the environmental factors to obtain a preliminary feature set.

3. The cloud computing-based mapping data processing method according to claim 1, characterized in that, The step of performing dynamic correction based on the preliminary feature set to obtain a corrected feature set includes: Obtain the terrain slope correction coefficient; Based on the preliminary feature set, regional slope values ​​are extracted to obtain a regional slope value feature set; Based on the regional slope value feature set and the terrain slope correction coefficient, correction is performed to obtain the correction feature set.

4. The cloud computing-based mapping data processing method according to claim 1, characterized in that, The step of performing terrain residual analysis based on the corrected feature set to obtain an optimized feature set includes: Based on the correction feature set, residual distribution is extracted and spatial analysis is performed to obtain residual spatial feature values; If the residual space feature value is greater than or equal to a preset dynamic threshold, the correction feature set is optimized based on the random forest algorithm to obtain an optimized feature set.

5. The cloud computing-based mapping data processing method according to claim 1, characterized in that, The step of dividing the high-frequency feature cache according to the optimized feature set to obtain a partitioned feature set includes: The optimized feature set is divided into terrain grids according to a preset spatial scale to obtain initial partition feature data; Based on the initial partition feature data, high-frequency region identification and cache optimization are performed to obtain the updated partition feature structure; Based on the XGBoost algorithm, optimal configuration reasoning and partition reconstruction are performed on the updated partition feature structure to obtain the partition feature set.

6. The cloud computing-based mapping data processing method according to claim 1, characterized in that, The step of performing synchronization verification based on the partition feature set to obtain a consistency feature set includes: Based on the partition feature set, partition tasks are configured in a structured manner to obtain a set of structured task units; Based on the structured set of task units, perform boundary continuity synchronous verification to obtain the partition boundary region that needs to be optimized; Based on the CatBoost algorithm, consistency scoring and feature selection are performed on the partition boundary regions that need to be optimized to obtain a consistency feature set.

7. The cloud computing-based mapping data processing method according to claim 1, characterized in that, The step of performing multi-threaded feature integration processing based on the consistent feature set to obtain a globally integrated feature set includes: Based on the consistency feature set, multi-threaded concurrent integration processing is performed to obtain a preliminary integration feature set; Based on the preliminary integrated feature set, boundary continuity optimization processing is performed to obtain a smooth optimized integrated feature set; Based on the smooth optimized integration feature set, load balancing and global integration output are performed to obtain the global integration feature set.

8. The cloud computing-based mapping data processing method according to claim 1, characterized in that, The step of reconstructing and optimizing abnormal region features based on the globally integrated feature set to obtain a high-precision feature set includes: Based on the global integrated feature set, abnormal regions are detected and extracted to obtain a preliminary high-precision feature set. Based on the preliminary high-precision feature set, consistency verification and feature reconstruction are performed to obtain an optimized high-precision feature set. Based on the optimized high-precision feature set, boundary redrawing and precision verification optimization are performed to obtain the high-precision feature set.

9. The cloud computing-based mapping data processing method according to claim 1, characterized in that, The process of performing parallel task decomposition and accelerated feature integration based on the high-precision feature set to obtain high-precision mapping data includes: Based on the high-precision feature set, task decomposition and parallel integration processing are performed to obtain preliminary mapping data; Based on the preliminary mapping data, a consistency check is performed to obtain the optimized feature integration result; Based on the optimized feature integration results, task re-decomposition and boundary remodeling operations are performed to obtain the final high-precision mapping data.

10. A cloud computing-based surveying data processing system, characterized in that, include: The data acquisition module is used to acquire terrain elevation, terrain slope, and environmental parameters; The preliminary feature set module is used to perform multi-source data fusion and feature extraction based on the terrain elevation, terrain slope and environmental parameters to obtain a preliminary feature set; The correction feature set module is used to perform dynamic correction and integrity verification based on the preliminary feature set to obtain the correction feature set; The optimized feature set module is used to perform terrain residual analysis based on the corrected feature set to obtain the optimized feature set; The partition feature set module is used to perform high-frequency feature cache partitioning based on the optimized feature set to obtain the partition feature set; The consistency feature set module is used to perform synchronization verification based on the partition feature set to obtain a consistency feature set; The global integrated feature set module is used to perform multi-threaded feature integration processing based on the consistent feature set to obtain the global integrated feature set; The high-precision feature set module is used to reconstruct and optimize the abnormal region features based on the globally integrated feature set to obtain a high-precision feature set. The high-precision mapping data module is used to perform parallel task decomposition and accelerated feature integration based on the high-precision feature set to obtain high-precision mapping data.

Images