Land utilization planning investigation project surveying and mapping data acquisition and analysis system

Abstract

The invention discloses a surveying and mapping data acquisition and analysis system for a land utilization planning investigation project. The surveying and mapping data acquisition and analysis system comprises a multi-source data acquisition module, a data preprocessing module, a data fusion module, an intelligent analysis module and a result output module. According to the invention, a macroscopic spectral image is obtained through a hyperspectral satellite, local details are focused through multi-spectral photography of the unmanned aerial vehicle, soil data are monitored in real time by a ground sensor, and multi-source data complement each other, so that the comprehensiveness and accuracy of the data are greatly improved; in coastal zone monitoring, satellite remote sensing, shipborne sonar and a shore-based sensor are integrated, air-sea-shore three-dimensional monitoring is achieved, information covering the terrain, the ecological environment and the like is obtained, and compared with single coastal image collection of a common method, dynamic changes of the coastal zone can be reflected more comprehensively.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of land surveying and mapping analysis, and particularly relates to a land use planning survey project mapping data collection and analysis system. BACKGROUND

[0002] The accuracy, perfection and overallness of land planning cannot be achieved without land surveying and mapping technology. The land surveying and mapping technology can provide reliable information for land planning, thereby greatly improving the work efficiency of land planning.

[0003] For the current data collection system, the traditional data collection method mainly relies on manual measurement and single sensor equipment, and has problems of low efficiency, large error, single data dimension and the like. With the development of geographic information systems (GIS), global positioning systems (GPS) and remote sensing technology, although the efficiency and accuracy of data collection are improved, there are still deficiencies in multi-source data fusion, real-time processing and intelligent analysis. The formats of different types of data are not unified, data redundancy leads to slow processing speed, and there is a lack of data analysis algorithm for specific needs of land use planning, which is difficult to meet the complex and changeable land use planning survey requirements. Accordingly, the application provides a land use planning survey project mapping data collection and analysis system. SUMMARY

[0004] The application aims to solve the problems in the prior art and provides a land use planning survey project mapping data collection and analysis system.

[0005] To achieve the above-mentioned purpose, the application adopts the following technical solutions:

[0006] A land use planning survey project mapping data collection and analysis system comprises:

[0007] A multi-source data collection module is used for integrating satellite remote sensing equipment, unmanned aerial laser radar, ground mobile measurement system and Internet of Things sensors, respectively acquiring large-range land image data, high-precision three-dimensional point cloud data, road and surrounding terrain and object data and environment data, and transmitting the data through a wireless communication module;

[0008] A data preprocessing module is used for performing format conversion, noise removal and data normalization processing on original data;

[0009] A data fusion module is used for uniformly converting different types of data into the same coordinate system by using an algorithm based on feature matching;

[0010] An intelligent analysis module is used for analyzing the fused data by using a land use type classification algorithm and a land use change detection algorithm;

[0011] Result output module: for outputting analysis results in the form of maps, reports and three-dimensional models, and supporting users to view and download through the Web or mobile terminal

[0012] Preferably, in the high-precision farmland planning scene, the multi-source data acquisition module uses a hyperspectral satellite remote sensing device to obtain spectral images, uses a multi-spectral camera-equipped unmanned aerial vehicle to perform low-altitude photography, and simultaneously arranges soil humidity, nitrogen, phosphorus and potassium content sensors in the farmland to collect soil data.

[0013] Preferably, the data preprocessing module uses the FLAASH algorithm for atmospheric correction for hyperspectral images, voxel filtering and downsampling for laser radar point clouds, and geometric correction and orthographic correction for satellite images.

[0014] Preferably, in the urban old city reconstruction scene, the data fusion module uses an octree-based method to fuse laser radar point clouds and oblique photography images, to realize texture mapping of the point clouds.

[0015] Preferably, in the mountain resource survey scene, the intelligent analysis module uses a terrain factor analysis algorithm to calculate terrain factors such as slope, slope direction and curvature, and uses a mobile curved surface fitting algorithm to perform terrain filtering on the laser radar point clouds.

[0016] Preferably, in the coastal zone monitoring scene, the intelligent analysis module uses a U-Net network model to extract the coastline, and uses a regression analysis model combined with time series analysis to predict the trend of ecological environment changes.

[0017] Preferably, the data fusion module uses a spatio-temporal data fusion algorithm to calculate weights based on the spatial distance and collection time of data points, to realize fusion of data of different time and spatial resolutions.

[0018] Preferably, the satellite remote sensing device in the multi-source data acquisition module has the ability to periodically obtain multi-spectral images of different resolutions, and the image coverage ranges from visible light to near-infrared band, for macro monitoring and analysis of land use conditions.

[0019] The present application has the following beneficial effects:

[0020] 1. In high-precision farmland planning, a "sky-ground" integrated acquisition mode is adopted, a hyperspectral satellite obtains macro spectral images, a multi-spectral camera-equipped unmanned aerial vehicle focuses on local details, and ground sensors monitor soil data in real time, so that multi-source data complement each other, greatly improving the comprehensiveness and accuracy of data; in coastal zone monitoring, satellite remote sensing, shipborne sonar and shore-based sensors are integrated to realize air-sea-land stereoscopic monitoring, and obtain information covering terrain, ecological environment and other aspects, which can more comprehensively reflect the dynamic changes of the coastal zone compared with the single coastal image acquisition of ordinary methods.

[0021] 2. Voxel filtering and downsampling of vehicle-mounted laser radar point cloud and octree-based texture mapping fusion, distortion correction of oblique photography image, to ensure data quality; in mountain resource survey, mobile curved surface fitting algorithm is used to separate laser radar point cloud ground points and non-ground points, satellite image is orthorectified combined with digital elevation model, to effectively improve data availability, and ordinary method is difficult to realize such accurate data processing.

[0022] 3. In high-precision farmland planning, support vector machine is combined with spectrum and soil data to realize accurate crop type classification; in urban old city reconstruction, three-dimensional point cloud change detection algorithm is used to quantify the difference before and after reconstruction; in mountain resource survey, terrain factor analysis algorithm is used to deeply mine terrain characteristics; in coastal monitoring, deep learning and regression analysis are used to realize coastline extraction and ecological environment trend prediction, compared with ordinary method, it can provide more in-depth and scientific decision basis for land use planning. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 A module structure schematic diagram of a land use planning survey project surveying and mapping data collection and analysis system is provided for the present application;

[0024] Figure 2 A part code display diagram of atmospheric correction in the land use planning survey project surveying and mapping data collection and analysis system is provided for the present application;

[0025] Figure 3 A part code display diagram of point cloud voxel filtering in the land use planning survey project surveying and mapping data collection and analysis system is provided for the present application;

[0026] Figure 4 A part code display diagram of laser radar point cloud terrain filtering in the land use planning survey project surveying and mapping data collection and analysis system is provided for the present application;

[0027] Figure 5 A part code display diagram of coastline extraction based on U-Net in the land use planning survey project surveying and mapping data collection and analysis system is provided for the present application. DETAILED DESCRIPTION

[0028] 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, not all the embodiments.

[0029] Embodiment one: high-precision farmland planning data collection and analysis based on hyperspectral remote sensing

[0030] In the high-precision farmland planning scenario, the system constructs a "sky-ground" integrated data acquisition system, coupled with intelligent data processing and analysis module. High-spectral satellites as "sky-based" data sources provide macro-scale spectral coverage of farmland; unmanned aerial vehicles equipped with multi-spectral cameras perform "air-based" data acquisition, focusing on local details of farmland; ground soil sensor network realizes "ground-based" real-time monitoring, and the data of the three are finally converged to the central processing platform, after preprocessing, fusion and intelligent analysis, the farmland planning decision information is output.

[0031] Data acquisition: Choose satellites with high-spectral imaging capabilities, such as Gaofen-5 satellites, with spectral resolution of 5nm, covering 400-2500nm visible light to near-infrared band. Through the satellite ground receiving station, the farmland spectral image is obtained with a specific imaging period (such as 3 days once), and the image spatial resolution is 30 meters, which can be used to identify the overall crop distribution and growth of farmland;

[0032] Deploy multi-spectral unmanned aerial vehicles equipped with multi-spectral cameras, fly at a height of 100 meters above the farmland according to the preset route. The camera is configured with red, green, blue, red edge, and near-infrared 5 bands to obtain crop canopy images with a spatial resolution of 0.1 meters, which can capture the subtle differences of individual crops;

[0033] Lay soil sensor nodes in the farmland with a grid spacing of 50m x 50m, each node integrates soil moisture sensors (such as Decagon EC-5) and nitrogen, phosphorus, and potassium content sensors (such as Sentek EnviroPro). The sensor transmits real-time data to the aggregation node at an interval of 15 minutes through LoRa wireless communication technology, and then uploads it to the central processing platform.

[0034] Data preprocessing: High-spectral images are affected by atmospheric molecules and aerosol scattering and absorption, resulting in spectral distortion. The FLAASH algorithm is based on radiation transfer theory and corrects by the following steps:

[0035] Calculate the atmospheric transmittance: According to the solar zenith angle θ s , observation zenith angle, and atmospheric profile data (such as atmospheric water vapor content, ozone content), use the MODTRAN radiation transfer model to calculate the atmospheric transmittance τ λ of each band.

[0036] Calculate the atmospheric path radiance: By solving the radiation transfer equation, the atmospheric path radiance L path,λ is obtained, and the correction formula is Where L TOA is the atmospheric top radiation brightness, and after correction, the value is closer to the true reflection radiation of the ground; L λ is the sensor observation radiation brightness; d is the distance from the earth to the sun, which can be obtained from the astronomical calendar table according to the date; E s,λFor solar irradiance, the calibration parameters of satellite sensors are related.

[0037] Due to the factors such as the flight attitude of the unmanned aerial vehicle and the terrain undulation, the multispectral image has geometric distortion. A polynomial transformation model is adopted to establish the mapping relationship between the original image coordinates (x, y) and the geographic coordinates (X, Y) through ground control points (GCPs):

[0038]

[0039] At least 6 GCPs are selected for uniform distribution, and the polynomial coefficients a ij and b ij are solved by the least square method to realize the geometric precise correction of the image.

[0040] Data fusion: the spectral angle matching (SAM) algorithm is used to fuse the hyperspectral image and the soil sensor data. The spectral data is regarded as a vector, and the similarity is measured by calculating the angle between the vectors: where R 1i and R 2i are the reflectivity of the i-th waveband of the hyperspectral image spectral vector and the soil nutrient spectral feature vector, and θ is the spectral angle. In practical application, an angle threshold (such as 15°) is set, and the region corresponding to the spectral vector less than the threshold is regarded as similar, and the soil sensor data is fused into the corresponding position of the hyperspectral image to generate a fused data cube.

[0041] Intelligent analysis: the support vector machine (SVM) algorithm is used for crop type classification. The spectral features (such as vegetation index NDVI, EVI) and soil nutrient data (nitrogen, phosphorus, and potassium content) of the fused data are used as the input feature vector x, and the data is mapped to a high-dimensional space through the kernel function K(x i , x) (such as the radial basis function K(x i , x) = exp (-γ||x i -x|| 2 ), and γ is the kernel parameter) to construct an optimal classification hyperplane:

[0042] where α i is the Lagrange multiplier, which is obtained by solving a quadratic programming problem; y i is the sample class label (such as wheat, corn, soybean, etc.); and b is the bias term. During the training process, the cross-validation method is used to optimize the kernel parameter γ and the penalty parameter C to improve the classification accuracy.

[0043] Embodiment two: three-dimensional modeling and change detection of urban old city reconstruction

[0044] Focusing on the needs of urban renewal, a data acquisition module was built using vehicle-mounted LiDAR and drone-mounted oblique photography, combined with 3D modeling and change detection algorithms. The vehicle-mounted LiDAR dynamically scans along city streets, acquiring point cloud data of roads and surrounding buildings. Drones capture images of buildings from multiple angles. Both data are pre-processed and integrated into a model on a central processing platform, ultimately enabling the detection and visualization of changes before and after renovation.

[0045] Data Collection: A Velodyne VLP-16 lidar was selected, mounted on the top of the vehicle and driven along city streets at a speed of 20 km / h to collect point cloud data. The radar has a vertical field of view of 30° (-15°-15°) and a horizontal angular resolution of 0.1°. It can generate approximately 300,000 points per second, with a point cloud density of 100 points per square meter, capable of fully recording 3D information such as building facades and road facilities.

[0046] A drone equipped with a five-lens oblique camera (such as the RCD305) was used at an altitude of 120 meters above the building to capture images from five angles: vertical, front, back, left, and right. This captured all-around texture information of the building. The camera had a resolution of 20 megapixels and a ground sampling distance (GSD) of approximately 2 centimeters, ensuring clear texture details.

[0047] Data preprocessing: The amount of point cloud data collected by the vehicle-mounted laser radar is huge and needs to be downsampled. The voxel filtering algorithm is used to divide the three-dimensional space into For a voxel grid (e.g. 0.1m×0.1m×0.1m), the centroid of each voxel is calculated as the representative point. The formula is: Where V is the voxel coordinate, is the point cloud coordinate, This method can reduce the amount of point cloud data by 80%-90% while retaining the main geometric features.

[0048] Oblique photography cameras are subject to lens distortion, so the Zhang calibration method is used to calibrate the camera. By capturing an image of a checkerboard calibration plate, corner coordinates are extracted, and the camera's intrinsic parameter matrix (focal length, principal point coordinates) and distortion parameters (radial distortion, tangential distortion) are calculated. This distortion correction is then applied to the image, improving subsequent modeling accuracy.

[0049] Data fusion: Data fusion

[0050] An octree-based point cloud and image fusion method is used. The octree structure recursively divides the three-dimensional space into eight equal subspaces, until the number of points or the size of the subspace reaches the set threshold. In the fusion process, the following steps are taken to achieve texture mapping; a feature-based registration method is used to extract the PFH (point feature histogram) features of the point cloud and the SIFT features of the image, and the initial transformation parameters are calculated by the nearest neighbor matching algorithm. Then, the iterative closest point (ICP) algorithm is used for fine registration to establish the spatial correspondence between the point cloud and the image. For the point cloud in each octree node, the corresponding pixel is found on the image according to its spatial coordinates, and the pixel color value is assigned to the point cloud to achieve texture mapping and generate a three-dimensional point cloud model with real texture.

[0051] Intelligent analysis: a change detection algorithm based on three-dimensional point cloud is used to calculate the mean absolute error (MAE) of the point cloud before and after the transformation: Where p 1i and p 2i are the coordinates of the corresponding points before and after the transformation, and n is the number of corresponding points. In specific implementation, the following steps are taken for change detection: use the iterative closest point (ICP) algorithm to register the point clouds before and after the transformation to the same coordinate system; use the KD-Tree data structure to quickly find the corresponding points in the two point clouds; set the MAE threshold (such as 0.5 meters), and when the MAE of the corresponding points exceeds the threshold, determine that the area has changed. The change area is visualized and displayed on the three-dimensional model through color coding.

[0052] Implementation mode three: multi-scale terrain analysis for mountain resource survey

[0053] For mountain resource survey, the system constructs a "macro-meso-micro" multi-scale data acquisition system. Satellite remote sensing provides a macro view of the terrain and geomorphology; unmanned aerial vehicle laser radar realizes mesoscale high-precision terrain modeling; ground artificial measurement obtains key node micro data. After preprocessing and fusion, the data are analyzed by terrain factor analysis algorithm to provide scientific basis for mountain resource assessment.

[0054] Data acquisition: Choose high two satellite, get spatial resolution of 1 meter panchromatic image and 4 meters multispectral image, cover the whole survey area, for identifying mountain terrain, vegetation cover and other macro features, using RieglVUX-1UAV laser radar, in the complex mountainous area with 150 meters flight height, laser scanning frequency up to 240 kHz, point cloud density up to 50 points per square meter, get high precision terrain point cloud data, especially suitable for steep slopes, valleys and other artificial terrain measurement in areas difficult to reach, using Trimble S9 total station in key terrain nodes (such as mountain top, valley, ridge line) measurement, measurement accuracy up to ± 1 mm + 1 ppm, get control point coordinates, for unmanned aerial vehicle point cloud data geographic coordinate conversion and accuracy verification.

[0055] Data preprocessing: assuming the ground is a local smooth surface, using moving surface fitting algorithm to separate ground points and non-ground points, selecting a neighborhood point with a radius of r (such as 2 meters) as the center, using least squares method to fit quadratic surface:

[0056]

[0057] Wherein, z is the elevation value.a ij The fitting coefficient.

[0058] Calculate the vertical distance d of the point to the fitted surface, set the distance threshold (such as 0.5 meters), the points less than the threshold are determined as ground points, otherwise as non-ground points, based on digital elevation model (DEM), using digital differential correction technology to orthorectify satellite image, eliminate the image deformation caused by terrain undulation, generate orthophoto map, convenient for other data superposition analysis.

[0059] Data fusion: use pyramid model to fuse data of different scales. The high-resolution unmanned aerial vehicle point cloud data and low-resolution satellite image data are used to construct a multi-resolution pyramid. The unmanned aerial vehicle point cloud data and satellite image are down-sampled respectively to generate data of different resolution levels. Starting from high-resolution data, the low-resolution data is generated by up-sampling layer by layer to form a pyramid structure. Each layer of data maintains data continuity through interpolation algorithm (such as bilinear interpolation). In each layer of the pyramid, the elevation information of the point cloud data and the texture information of the image data are fused according to the spatial position relationship of the data, and finally the multi-scale fusion data is obtained.

[0060] Intelligent analysis: using terrain factor analysis algorithm, calculate slope, slope direction, curvature and other terrain factors, slope represents the inclination of the ground unit, the calculation formula is: Through differential calculation of digital elevation model (DEM), get And (derivative of elevation in x and y direction), and then the slope is calculated, the result is in degree.

[0061] Aspect is the direction of the normal of the tangent plane at a point on the surface, projected onto the horizontal plane (0° for the positive north direction), the formula is:

[0062] Where arctan 2 is the four quadrant arctangent function, the result is normalized to the range of 0-360°.

[0063] Curvature reflects the bending degree of the terrain surface, which is calculated by the second derivative of elevation, such as profile curvature, the formula is:

[0064]

[0065] Where z xx ,z xy ,z yy are the second derivative of elevation z with respect to x, the mixed second derivative of x and y, and the second derivative of y, respectively. By analyzing these terrain factors, combined with geological, hydrological and other data, the potential of mountain resources development is evaluated.

[0066] Embodiment four: dynamic data acquisition and analysis of coastal monitoring

[0067] Data preprocessing: due to satellite orbit deviation, earth curvature and other factors, the image has geometric distortion. A polynomial transformation model is used to establish the mapping relationship between the original image coordinates (x, y) and geographic coordinates (X, Y) through at least 6 ground control points (such as obvious ground points on the coastline, lighthouse, breakwater endpoint, etc.):

[0068] The polynomial coefficients a ij and b ij are solved by least squares method, generally n=2 can meet the geometric correction accuracy requirement in most scenarios. The corrected image can be accurately fitted into the geographic coordinate system, which is convenient for subsequent analysis.

[0069] The original water depth data collected by the shipborne sonar contains noise and outliers. First, median filtering is used to remove salt and pepper noise. For each water depth measurement point, the median of the water depth values in its neighborhood (such as a 3*3 grid centered on the point) is taken as the filtered value, the formula is:

[0070] f(x,y) = median{g(s,t), (s,t) ∈ N xy} where f(x,y) is the water depth value of the filtered point (x,y), g(s,t) is the neighborhood N xyThe water depth value of the inner point (s, t). Then, by statistical analysis, identify and remove outliers, calculate the mean μ and standard deviation σ of all water depth values, and remove the water depth values deviating from the mean by more than 3σ.

[0071] The data collected by the shore-based sensor may have problems such as data missing and jumping. For missing values, linear interpolation method is used to fill in. If the data at time t is missing, and the data at time t1 and t2 before and after it are known as y1 and y2 respectively, then the interpolation y at time t is:

[0072] For jumping data, set a reasonable threshold range (such as water temperature change amplitude not more than 5 degrees Celsius per hour), and determine the data outside the threshold range as abnormal and correct or remove it.

[0073] Data fusion: Use spatio-temporal data fusion algorithm to fuse satellite images, sonar data and sensor data with different time and spatial resolution. Based on spatio-temporal weight distribution, the formula is: Where D f is the fused data, D i is the original data, w i is the spatio-temporal weight coefficient, and d ij is the Euclidean distance between each data point and the target fusion point. The closer the distance, the higher the weight. The formula for calculating the spatial weight w sij is:

[0074] Determine the time weight w tij according to the time difference Δt i between data collection time and target fusion time. The smaller the time difference, the higher the weight. The formula is:

[0075] The final spatio-temporal weight coefficient w i is the weighted average of spatial weight and time weight, i.e. w i = αw sij + (1-α)w tij , where α is an adjustment parameter, which is determined according to the actual situation (generally between 0.4 and 0.6).

[0076] Intelligent analysis: Use U-Net network model to extract coastline. U-Net network adopts encoder-decoder structure, the encoder part is composed of multiple convolution layers and pooling layers, which is used to extract the features of the image; the decoder part restores the feature map to the original image size through upsampling and convolution operation, and outputs the probability map of the coastline position.

[0077] The network input is the pre-processed and fused multi-spectral image data. In the training process, the manually labeled coastline data is used as the label, and the cross-entropy loss function L is used for optimization, and the formula is:

[0078]

[0079] Where x, y are image pixel coordinates, y xy is the true label (coastline pixels are 1, and non-coastline pixels are 0), p xy is the probability that the model predicts that the pixel belongs to the coastline. Through the back propagation algorithm, the network parameters are updated to minimize the loss function, and the accuracy of coastline extraction is improved.

[0080] Through the analysis of marine environmental data, the regression analysis model is used to predict the trend of ecological environment change, such as the chlorophyll concentration prediction formula: Where C is the chlorophyll concentration, X i is the influencing factor (such as temperature, salinity, dissolved oxygen, light intensity, etc.), β i is the regression coefficient, and ò is the error term. The multiple linear regression method is used to solve the regression coefficient β i , so that the mean square error of the predicted value and the actual value is minimized. In practical application, time series analysis can also be combined to consider the time correlation of data, and ARIMA model can be used to dynamically predict ecological environment parameters, providing decision support for coastal zone ecological protection and resource management.

[0081] Table 1: Comparison of specific detection report data of each embodiment

[0082]

[0083] From the above table, it can be seen that the four embodiments break through the traditional single data collection mode and build a multi-source collaborative collection system. In high-precision farmland planning, the "sky-ground" integrated collection method combines hyperspectral satellites, unmanned aerial multi-spectral photography and ground soil sensors to realize comprehensive data acquisition from macro to micro and from spectral image to soil parameters; in urban old city reconstruction, vehicle-mounted laser radar and unmanned aerial oblique photography are used to quickly collect building three-dimensional point cloud and texture information; in mountain resource exploration, satellite remote sensing, unmanned aerial laser radar and ground manual measurement are used to cover large-scale terrain and key node data; in coastal monitoring, satellite, shipborne sonar and shore-based sensors are integrated to complete air-sea-land three-dimensional monitoring. This multi-source collection mode makes the data collection efficiency improve by 5-10 times, the collection dimension increase by 2-3 times, and the data coverage range expand by several to several dozen times, providing rich and comprehensive data basis for subsequent analysis.

[0084] Further, the processing scheme is customized according to data characteristics. For example, atmospheric correction is performed on hyperspectral images, voxel filtering is performed on laser radar point clouds, and geometric correction is performed on satellite images, effectively removing noise and correcting distortion, and significantly improving data quality. In terms of data fusion, high-precision farmland planning uses spectral angle matching algorithm to fuse multi-source data; urban old city reconstruction uses octree-based method to realize point cloud and image fusion; coastal zone monitoring integrates data of different temporal and spatial resolutions through spatio-temporal data fusion algorithm. These innovative fusion algorithms reduce data registration error by 80%-90%, greatly improve data integrity and consistency, and lay a solid foundation for intelligent analysis.

[0085] Further, advanced algorithms are used in a targeted manner. High-precision farmland planning uses support vector machines to achieve high-precision classification of crop types, with an accuracy improvement of 20%-30%; urban old city reconstruction uses three-dimensional point cloud change detection algorithm to accurately quantify the difference before and after reconstruction, with an accuracy of 90%-96%; mountain resource survey uses terrain factor analysis algorithm to accurately extract features such as slope and aspect, with an error reduction of 70%-80%; coastal zone monitoring uses deep learning and regression analysis to achieve accurate extraction of coastline and prediction of ecological environment trends, with an accuracy improvement of 18%-24% and 25%-32%, respectively. These algorithms deeply mine data value and provide scientific and accurate decision-making basis for land use planning.

[0086] Further, the system shows wide applicability and great benefits in different scenarios. In terms of efficiency, full-process automation reduces project cycle by 60%-80%; in terms of cost, by reducing manual input and data reuse, project cost is reduced by 40%-60%. In addition, the system also has good scalability, modular architecture can quickly adapt to new scenarios, and transfer learning function reduces algorithm development cost. In the field of ecological protection and sustainable development, the system helps to reduce pollution through precise fertilization of farmland, ecological restoration of coastal zone, and rational development of mountain resources, which is of great significance to sustainable utilization of land resources and ecological environment protection.

[0087] In summary, the data collection link breaks the traditional single mode, and in a multi-source collaborative way, according to the needs of different scenarios, it combines satellite remote sensing, unmanned aerial vehicles, ground sensors and other equipment, so that the collection efficiency is improved by 5-10 times, and the data dimension is increased by 2-3 times; in data processing, the customized preprocessing scheme combined with innovative fusion algorithm reduces the data registration error by 80%-90%; data analysis with the help of advanced algorithms such as support vector machine and deep learning can greatly improve the accuracy of classification, detection and prediction in various scenarios, and provide accurate data support for land use planning survey. From the application value, the system shows significant advantages. Its full-process automation reduces the project cycle by 60%-80%, reduces the project cost by 40%-60% through reducing manual input and data reuse, and at the same time, the modular architecture and transfer learning function give it strong scalability. In terms of ecological protection and sustainable development, the system helps precision fertilization in farmland, ecological restoration in coastal zone, and rational development of mountain resources, and plays an important role in promoting sustainable use of land resources and ecological environment protection, and has broad application prospects and far-reaching social and economic benefits.

[0088] Specifically, in various embodiments, key code examples written in Python language and corresponding analysis are taken.

[0089] Further, as shown in Figure 2-Figure 3 , the code realizes effective preprocessing and feature extraction according to different data source characteristics. For example, in high-precision farmland planning, the atmospheric correction code simulates the calculation of atmospheric top radiation brightness, providing a basis for eliminating atmospheric interference of hyperspectral images; the point cloud voxel filtering code in urban old city reconstruction reduces the amount of point cloud data while retaining geometric features; the laser radar point cloud terrain filtering code in mountain resource survey separates ground points through moving surface fitting algorithm, which improves the quality of original data and lays a solid foundation for subsequent analysis.

[0090] Further, as shown in Figure 3-Figure 4 , in terms of data fusion and registration, the codes of various embodiments each show their abilities. The spectral angle matching algorithm for high-precision farmland planning realizes the similarity measurement and fusion of multi-source data characteristics; the iterative closest point algorithm for urban old city reconstruction completes accurate registration of point clouds; the spatio-temporal data fusion code for coastal zone monitoring considers spatial distance and collection time to calculate weights, realizes fusion of multi-source heterogeneous data, and ensures the consistency and availability of data.

[0091] Further, as shown in Figure 2-Figure 5As shown, the intelligent analysis link adopts classical machine learning and deep learning algorithm to construct models. High-precision farmland planning uses support vector machine for crop type classification; coastal monitoring uses U-Net to construct a deep learning model to extract the coastline. Through learning and training of the data, these models mine the potential value of the data to provide scientific and accurate analysis results and decision basis for land use planning survey.

[0092] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can make equivalent replacements or changes within the technical range disclosed by the present application according to the technical solution and inventive concept of the present application, which should be covered within the protection scope of the present application.

Claims

1. A land use planning and surveying project surveying and mapping data acquisition and analysis system, characterized in that: include: Multi-source data acquisition module: used to integrate satellite remote sensing equipment, drone-mounted lidar, ground mobile measurement systems, and IoT sensors to obtain large-scale land image data, high-precision 3D point cloud data, road and surrounding terrain data, and environmental data, and transmit the data through wireless communication modules; Data preprocessing module: used to convert the format of raw data, remove noise and perform data normalization; Data fusion module: used to unify different types of data into the same coordinate system using a feature matching-based algorithm; Intelligent analysis module: used to analyze the fused data through land use type classification algorithm and land use change detection algorithm; Result output module: used to output analysis results in the form of maps, reports, and 3D models, and supports users to view and download them through the web or mobile terminals.

2. A land use planning and investigation project surveying and mapping data acquisition and analysis system according to claim 1, characterized in that: In the high-precision farmland planning scenario, the multi-source data acquisition module uses hyperspectral satellite remote sensing equipment to obtain spectral images, uses drones equipped with multispectral cameras for low-altitude photography, and simultaneously deploys soil moisture and nitrogen, phosphorus, and potassium content sensors in the farmland to collect soil data.

3. The land use planning and investigation project surveying and mapping data acquisition and analysis system according to claim 1, characterized in that: The data preprocessing module uses the FLAASH algorithm to perform atmospheric correction on hyperspectral images, performs voxel filtering and downsampling on lidar point clouds, and performs geometric correction and orthorectification on satellite images.

4. A land use planning and investigation project surveying and mapping data acquisition and analysis system according to claim 1, characterized in that: In the urban old town renovation scenario, the data fusion module uses an octree-based method to fuse the lidar point cloud with the oblique photography image to achieve texture mapping of the point cloud.

5. The land use planning and investigation project surveying and mapping data acquisition and analysis system according to claim 1, characterized in that: In the mountain resource survey scenario, the intelligent analysis module uses a terrain factor analysis algorithm to calculate terrain factors such as slope, aspect, and curvature, and performs terrain filtering on the lidar point cloud through a moving surface fitting algorithm.

6. The land use planning and investigation project surveying and mapping data acquisition and analysis system according to claim 1, characterized in that: In the coastal monitoring scenario, the intelligent analysis module uses the U-Net network model to extract the coastline and uses the regression analysis model combined with time series analysis to predict the trend of ecological environment changes.

7. The land use planning and investigation project surveying and mapping data acquisition and analysis system according to claim 1 is characterized in that: The data fusion module adopts a spatiotemporal data fusion algorithm to calculate weights based on the spatial distance and acquisition time of data points, thereby achieving the fusion of data with different temporal and spatial resolutions.

8. The land use planning and investigation project surveying and mapping data acquisition and analysis system according to claim 1, characterized in that: The satellite remote sensing equipment in the multi-source data acquisition module has the ability to regularly acquire multispectral images of different resolutions, and the images cover the visible light to near-infrared bands, which are used for macro-monitoring and analysis of land use conditions.

Images