Land reclamation project progress monitoring system based on time sequence image analysis

Abstract

The invention relates to the field of land reclamation engineering, in particular to a land reclamation engineering progress monitoring system based on time sequence image analysis, which comprises a data storage module, an engineering progress image recognition engine, an engineering quality detection module, a meteorological condition compensation module and an early warning module. A Riemannian geometric model of a terrain curved surface is constructed, a project progress is analyzed and quantified through a geodesic network, a project progress image recognition engine comprises an image registration module, a Riemannian geometric terrain modeling module and a geodesic analysis module, and the Riemannian geometric terrain modeling module constructs a terrain surface model through manifold parameterization and measurement tensor calculation. The geodesic analysis module quantifies engineering progress indexes such as the earthwork completion rate, the road hardening rate and the greening coverage rate based on geodesic network changes, the system further integrates quality evaluation functions such as slope stability detection and drainage system integrity detection, and the reliability of a monitoring result is improved through meteorological condition compensation.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of land consolidation engineering, in particular to a land consolidation engineering progress monitoring system based on time series image analysis, and more particularly to using differential geometry theory to analyze multi-temporal images collected by unmanned aerial vehicles to realize automatic monitoring of land consolidation engineering progress. BACKGROUND

[0002] Land consolidation engineering is an important measure for land space development and utilization and ecological protection. It usually has large scale, long cycle and wide distribution. Traditional manual field inspection methods have low efficiency, strong subjectivity and inaccurate data. In recent years, with the development of unmanned aerial vehicle technology and remote sensing technology, using unmanned aerial vehicles to collect land consolidation engineering image data for monitoring has become a trend.

[0003] Currently, there are two main ways to monitor land consolidation engineering progress: one is regular manual field measurement, and the other is manual interpretation after obtaining images by unmanned aerial vehicles. The former is time-consuming and labor-intensive, and the accuracy is limited. Although the latter improves the data acquisition efficiency, it still mainly relies on manual interpretation in the image analysis process, which is difficult to meet the accurate monitoring needs of large-scale engineering.

[0004] Some existing automatic monitoring methods are mainly based on simple pixel comparison or region segmentation technology. These methods are difficult to provide stable and reliable monitoring results when facing complex terrain, variable lighting conditions and various interference factors. In addition, existing methods often lack specific consideration of land consolidation engineering characteristics, making it difficult to extract and quantify key engineering indicators, resulting in a large deviation between monitoring results and actual engineering progress.

[0005] Therefore, there is an urgent need for a land consolidation engineering progress monitoring system that can overcome the above shortcomings, based on advanced mathematical theory, to achieve high-precision, stable and reliable land consolidation engineering progress monitoring. SUMMARY

[0006] The purpose of the present application is to provide a land consolidation engineering progress monitoring system based on time series image analysis, which introduces differential geometry theory to analyze and process multi-temporal images collected by unmanned aerial vehicles, to realize accurate monitoring and quality detection of land consolidation engineering progress.

[0007] The present application provides a land consolidation engineering progress monitoring system based on time series image analysis, which includes:

[0008] A data storage module for storing historical image data of land consolidation engineering projects collected by unmanned aerial vehicles and corresponding geographic information;

[0009] An engineering progress image recognition engine, in communication connection with the data storage module, is configured to receive the image data provided by the data storage module, construct a terrain surface model based on differential geometry theory, and extract engineering key technology parameters from the terrain surface model; the engineering progress image recognition engine comprises an image registration module, a Riemannian geometry terrain modeling module, and a geodesic line analysis module, wherein the image registration module is configured to register images at different time points to a unified coordinate system, the Riemannian geometry terrain modeling module is configured to construct a parameterized representation of a manifold representing a terrain surface and calculate differential invariants, and the geodesic line analysis module is configured to construct a geodesic line network and quantify engineering progress based on changes in the geodesic line network.

[0010] An engineering quality detection module, in communication connection with the engineering progress image recognition engine, is configured to receive the terrain surface model and differential geometry features provided by the engineering progress image recognition engine and evaluate engineering construction quality.

[0011] A meteorological condition compensation module, in communication connection with the engineering progress image recognition engine and the engineering quality detection module, is configured to compensate monitoring results according to meteorological conditions such as light and precipitation at the time of collection.

[0012] An early warning module, in communication connection with the engineering progress image recognition engine and the engineering quality detection module, is configured to generate early warning information based on engineering progress and engineering quality detection results.

[0013] Preferably, the historical image data of land consolidation engineering projects collected by the unmanned aerial vehicle and stored by the data storage module comprises a flight map and a panoramic map; the geographic information comprises geographic feature coordinates, elevations, and photographing times, and is configured to provide meteorological compensation parameters for the meteorological condition compensation module.

[0014] Preferably, the image registration module comprises:

[0015] A key point extraction unit configured to select points with invariant features from images as key points;

[0016] A feature description unit, in communication connection with the key point extraction unit, is configured to describe feature information of the key points using feature vectors;

[0017] A key point matching unit, in communication connection with the feature description unit, is configured to establish a corresponding relationship between homonymous points in images at different time points;

[0018] An image conversion unit, in communication connection with the key point matching unit, is configured to convert two-dimensional plane coordinates of different images to a unified plane coordinate system.

[0019] Preferably, the Riemannian geometry terrain modeling module comprises:

[0020] a manifold parameterization unit configured to map the terrain point cloud data to a parameter space and construct a two-dimensional manifold parameterization representation;

[0021] a Riemannian metric tensor computation unit communicatively connected to the manifold parameterization unit and configured to compute a metric tensor of the terrain surface to characterize local geometric properties of the surface;

[0022] a curvature tensor analysis unit communicatively connected to the Riemannian metric tensor computation unit and configured to compute Gaussian and mean curvature distributions of the terrain surface to identify key topographic features;

[0023] a differential invariant extraction unit communicatively connected to the curvature tensor analysis unit and configured to extract terrain geometric features that are robust to viewing angle, lighting, and scanning conditions.

[0024] Preferably, the geodesic analysis module comprises:

[0025] a key point selection unit configured to select geodesic network nodes based on curvature and terrain features;

[0026] a geodesic computation unit communicatively connected to the key point selection unit and configured to compute geodesics connecting the nodes on the terrain surface;

[0027] a geodesic flux analysis unit communicatively connected to the geodesic computation unit and configured to compare topological and geometric changes of the geodesic network at different time points to quantify terrain changes;

[0028] an engineering progress quantification unit communicatively connected to the geodesic flux analysis unit and configured to convert the terrain changes into engineering progress indicators such as earthwork completion rate, road hardening rate, and green coverage rate.

[0029] Preferably, the engineering quality detection module comprises:

[0030] an image segmentation unit configured to subdivide the remote sensing image into multiple sub-regions;

[0031] a slope stability detection unit communicatively connected to the image segmentation unit and configured to obtain slope grade, height, and length of the slope and compute a slope stability coefficient;

[0032] a drainage system detection unit communicatively connected to the image segmentation unit and configured to detect integrity and functionality of the drainage system;

[0033] a flatness detection unit communicatively connected to the image segmentation unit and configured to evaluate ground surface flatness based on differential geometric features.

[0034] Preferably, the meteorological condition compensation module comprises:

[0035] The illumination compensation unit is used to compensate for the impact of illumination on the monitoring results during image capture;

[0036] The precipitation compensation unit is used to compensate for the impact of precipitation on the monitoring results during image capture.

[0037] Ground change compensation unit is used to compensate for the impact of building construction on monitoring results within the monitoring range;

[0038] The illumination compensation unit, precipitation compensation unit, and ground change compensation unit acquire regional meteorological data based on the longitude, latitude, and time extracted from the geographic information, and perform targeted compensation processing based on the project progress and quality inspection results.

[0039] Preferably, the early warning module includes:

[0040] The parameter warning unit is used to compare the engineering progress indicators generated by the engineering progress image recognition engine with a preset threshold.

[0041] A quality early warning unit is used to compare the quality assessment results generated by the engineering quality inspection module with the quality standards.

[0042] The integrated early warning unit is communicatively connected to the parameter early warning unit and the quality early warning unit, and is used to generate multi-level early warning information based on the overall project progress and quality status.

[0043] The early warning push unit is communicatively connected to the integrated early warning unit and is used to send early warning information to users via SMS, email, or application push.

[0044] Preferably, the differential invariant extraction unit is specifically used for:

[0045] Calculate the family of differential invariants of topographic surfaces at multiple scales;

[0046] Constructing a multi-scale representation based on the Laplace-Beltrami operator;

[0047] Generate differential geometric fingerprints of feature surfaces to achieve robust comparisons of terrain at different time points;

[0048] A spatiotemporal covariance comparative analysis method was used to process the changes in perspective and illumination between images acquired at different times.

[0049] As a preferred embodiment, the project progress quantification unit is specifically used for:

[0050] Establish a mapping relationship between engineering design objectives and terrain change indicators;

[0051] The project completion evaluation function is defined based on geodetic distance.

[0052] Based on multi-temporal observation results, the differential geometric flow method is used to smooth the progress changes in the time dimension;

[0053] Build a progress prediction model to predict future progress based on historical trends.

[0054] This invention treats the land reclamation project site as a Riemannian manifold and uses tensor field theory and geodesic theory from differential geometry to achieve accurate modeling and quantification of changes in the terrain surface. This overcomes the problem that traditional pixel comparison methods are greatly affected by environmental conditions, and significantly improves monitoring accuracy and stability.

[0055] The present invention has the following beneficial effects:

[0056] 1. Improved monitoring accuracy: Through terrain surface modeling technology based on differential geometry, this invention can accurately describe terrain features, and the accuracy of engineering progress measurement can reach within ±5%, which is far superior to traditional methods.

[0057] 2. Enhanced environmental adaptability: The time-series change detection mechanism based on differential invariants makes the system more robust to environmental factors such as changes in illumination and differences in viewing angle, making it suitable for monitoring tasks under various complex weather conditions.

[0058] 3. Automatic calculation of multiple indicators: Through geodesic network analysis, the system can automatically calculate key indicators such as earthwork completion rate, road hardening rate and green coverage rate, reducing manual intervention and improving monitoring efficiency.

[0059] 4. Provides quality early warning function: The system can automatically detect quality problems such as slope stability and drainage system integrity, and provide timely warnings to support decision-making for project management.

[0060] 5. Reduce project risks: By detecting quality problems early, the system can issue warnings 7 to 10 days in advance, effectively reducing project risks and improving overall project quality. Attached Figure Description

[0061] Figure 1 This is a system architecture diagram of one embodiment of the present invention;

[0062] Figure 2 This is a schematic diagram of the engineering progress image recognition engine of the present invention;

[0063] Figure 3 This is a flowchart illustrating the workflow of the Riemannian geometric terrain modeling module of the present invention.

[0064] Figure 4 This is a flowchart illustrating the workflow of the geodesic analysis module of the present invention.

[0065] Figure 5 This is a schematic diagram of the engineering quality inspection module of the present invention;

[0066] Figure 6 This is a flowchart of the early warning module of the present invention. Detailed Implementation

[0067] Please refer to the attached document. Figures 1-6 The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0068] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0069] Reference Figure 1 The land consolidation project progress monitoring system based on time-series image analysis provided by the present invention includes a data storage module 1, a project progress image recognition engine 2, a project quality detection module 3, a meteorological condition compensation module 4, and an early warning module 5.

[0070] Data storage module 1 stores historical image data and corresponding geographic information of land consolidation projects collected by UAVs. Project progress image recognition engine 2 communicates with data storage module 1, receives image data from data storage module 1, constructs a terrain surface model based on differential geometry theory, and extracts key technical parameters of the project from the terrain surface model. Project quality inspection module 3 communicates with project progress image recognition engine 2, receives the terrain surface model and differential geometric features provided by project progress image recognition engine 2, and evaluates the construction quality of the project. Meteorological condition compensation module 4 communicates with project progress image recognition engine 2 and project quality inspection module 3, and compensates the monitoring results based on meteorological conditions such as illumination and precipitation at the time of data collection. Early warning module 5 communicates with project progress image recognition engine 2 and project quality inspection module 3, and generates early warning information based on project progress and project quality inspection results.

[0071] In a preferred embodiment of the present invention, the historical image data of land consolidation projects collected by the UAV and stored in the data storage module 1 include flight charts and panoramic images. Flight charts are orthophotos collected by the UAV following a preset flight path, while panoramic images are 360-degree panoramic images taken by the UAV hovering at a specific location. These two types of images complement each other and can comprehensively reflect the terrain features of the project area. Furthermore, the geographic information includes feature coordinates, elevation, and photography time, used to provide meteorological compensation parameters for the meteorological condition compensation module 4. Feature coordinates and elevation information are typically provided by the GPS / RTK positioning system and airborne radar on the UAV, with accuracy down to the centimeter level. The photography time records the specific time of image acquisition, providing a basis for subsequent time-series analysis and meteorological condition compensation.

[0072] The project progress image recognition engine 2 is the core component of this system, including the image registration module 21, the Riemannian geometric terrain modeling module 22, and the geodesic analysis module 23. The following will combine... Figure 2 This document provides a detailed introduction to the internal structure and working principle of the Engineering Progress Image Recognition Engine 2.

[0073] The image registration module 21 is used to register images from different time points to a unified coordinate system. In one embodiment of the present invention, the image registration module 21 includes a key point extraction unit 211, a feature description unit 212, a key point matching unit 213, and an image conversion unit 214.

[0074] The keypoint extraction unit 211 is used to select points with invariant features from the image as keypoints. Preferably, the keypoint extraction unit 211 uses an improved SIFT (Scale-Invariant Feature Transform) algorithm to extract keypoints. In this embodiment, the density of keypoint extraction is dynamically adjusted according to the terrain complexity. Typically, 3 to 5 keypoints are selected per 100 square meters in flat areas, and 8 to 12 keypoints are selected per 100 square meters in complex terrain areas to ensure sufficient coverage of key feature points.

[0075] The feature description unit 212 is communicatively connected to the key point extraction unit 211 and is used to describe the feature information of key points using feature vectors. In this embodiment, the feature description unit 212 generates a 128-dimensional feature vector for each key point, describing the gradient information of the area surrounding the key point. Furthermore, considering the characteristics of land reclamation projects, the feature description unit 212 also incorporates terrain elevation features to enhance the discriminative power of the feature description.

[0076] The keypoint matching unit 213 is communicatively connected to the feature description unit 212 and is used to establish corresponding relationships between images at different time points. In this embodiment, the keypoint matching unit 213 uses a nearest neighbor matching algorithm combined with the RANSAC (RandomSampleConsensus) algorithm to filter matching points. The matching threshold is set to 0.75 (feature distance ratio), which is the optimal value verified by a large number of experiments, and can retain a sufficient number of matching points while ensuring matching accuracy.

[0077] The image conversion unit 214 is communicatively connected to the keypoint matching unit 213 and is used to convert the two-dimensional planar coordinates of different images to a unified planar coordinate system. In this embodiment, the image conversion unit 214 calculates the affine transformation matrix or homography matrix (depending on the geometric changes of the scene) and uses a bicubic interpolation algorithm to resample the images, ensuring the quality and accuracy of the registered images.

[0078] The Riemannian geometric terrain modeling module 22 is used to construct a manifold parameterized representation of the terrain surface and to compute differential invariants. (See reference...) Figure 3 In one embodiment of the present invention, the Riemannian geometric terrain modeling module 22 includes a manifold parameterization unit 221, a Riemannian metric tensor calculation unit 222, a curvature tensor analysis unit 223, and a differential invariant extraction unit 224.

[0079] The manifold parameterization unit 221 is used to map terrain point cloud data to a parameter space, constructing a two-dimensional manifold parameterized representation. In this embodiment, the manifold parameterization unit 221 first performs density adaptive sampling on the point cloud data to ensure a higher density of key geomorphic feature points. Then, a local tangent plane projection algorithm is used to map the three-dimensional point cloud to the two-dimensional parameter domain. This parameterization method can effectively preserve the topological relationships and geometric features of the terrain. Specifically, for points... Coordinates in the parameter domain The mapping relationship can be expressed as:

[0080] ,

[0081] in: A point in three-dimensional space The coordinates are in the two-dimensional parameter domain. For point The Cartesian coordinate components in three-dimensional space. This mapping establishes a correspondence between the parameter domain and the three-dimensional terrain surface, enabling us to analyze and process the terrain in the parameter domain.

[0082] The Riemannian metric tensor computation unit 222 is communicatively connected to the manifold parameterization unit 221 and is used to compute the metric tensor of the terrain surface, characterizing the local geometric properties of the surface. In this example, the Riemannian metric tensor computation unit 222 constructs a local coordinate system for each point on the surface, computes the equation of the local tangent plane, and derives the metric tensor for each point on the surface. Metric Tensor (The first basic form) can be represented as:

[0083] ,

[0084] in: For the metric tensor (first fundamental form), it is a symmetric positive definite matrix; , indicating parameters The square of the first derivative in the direction; , indicating parameters and The dot product of the first derivatives in the direction; , indicating parameters The square of the first derivative in the direction; and Representing parametric surfaces right and The partial derivative of the tangent vector. Metric tensor. It describes the length change when a small displacement in the parameter domain is mapped to a three-dimensional surface, and is a fundamental tool in Riemannian geometry for characterizing the local properties of surfaces.

[0085] The curvature tensor analysis unit 223 is communicatively connected to the Riemannian metric tensor calculation unit 222, and is used to calculate the Gaussian curvature and mean curvature distribution of the terrain surface to identify key geomorphic features. In this example, the curvature tensor analysis unit 223 introduces a second basic form to calculate the surface curvature distribution:

[0086] ,

[0087] in: As the second fundamental form, it is symmetric matrix; , indicating parameters The dot product of the second derivative of the direction and the normal vector; , indicating parameters and The dot product of the mixed second derivative and the normal vector; , indicating parameters The dot product of the second derivative of the direction and the normal vector; For the surface at point The unit normal vector at that location is calculated using the following formula: ,in Represents the cross product of vectors. This represents the magnitude of the vector. Second fundamental form. It describes the degree of curvature of the surface in the direction of the normal vector.

[0088] Gaussian curvature can be calculated based on the first and second fundamental forms. and mean curvature :

[0089] ,

[0090] ,

[0091] in: Let be the Gaussian curvature, representing the intrinsic curvature of the surface, which is an important invariant in surface geometry; The mean curvature represents the average degree of curvature of the surface in all directions; , , The components of the first basic form; , , These are the components of the second fundamental form. Gaussian curvature and mean curvature are important indicators for describing the geometric properties of surfaces and can be used to identify topographic features such as ridges, valleys, and plains. In land reclamation projects, changes in these curvature values ​​directly reflect the progress of earthwork.

[0092] The differential invariant extraction unit 224 is communicatively connected to the curvature tensor analysis unit 223, and is used to extract terrain geometric features that are robust to viewpoint, illumination, and scanning conditions. In this embodiment, the differential invariant extraction unit 224 calculates a family of differential geometric invariants at multiple scales, including Gaussian curvature, mean curvature, and their derivatives. These invariants exhibit strong stability to changes in image acquisition conditions and can provide a reliable description of terrain features.

[0093] To further enhance the discriminative power of features, the differential invariant extraction unit 224 also constructs a multi-scale representation based on the Laplace-Beltrami operator. The Laplace-Beltrami operator... It is a generalization of the Laplace operator on Riemannian manifolds, defined as:

[0094] ,

[0095] in: It is the Laplace-Beltrami operator, a second-order differential operator on a Riemannian manifold; A function defined on a manifold, usually an elevation function; For measuring tensors The components of the inverse matrix; For measuring tensors The determinant of; and These are local coordinates on the manifold, corresponding to the parameters in this example. Represents coordinates The partial derivative operator; the summation symbol follows the Einstein summation convention, i.e., summing over repeated superscripts and subscripts (from 1 to the coordinate dimension, which is 2 in this case). The Laplace-Beltrami operator is an intrinsic differential operator that is independent of the embedding of the manifold and is therefore invariant to parametric transformations.

[0096] By calculating the Laplace-Beltrami operator responses at different scales, a multi-scale feature description of the terrain can be constructed, providing a foundation for subsequent change detection.

[0097] The geodesic analysis module 23 is used to construct a geodesic network and quantify project progress based on changes in the geodesic network. (Refer to...) Figure 4 In one embodiment of the present invention, the geodesic analysis module 23 includes a key point selection unit 231, a geodesic calculation unit 232, a geodesic flow analysis unit 233, and a project progress quantification unit 234.

[0098] The key point selection unit 231 is used to select geodesic network nodes based on curvature and terrain features. In this embodiment, the key point selection unit 231 identifies points with significant geometric features as network nodes by analyzing the curvature distribution of the terrain. Preferably, the node selection follows the following strategy: densely arranging nodes along feature lines such as ridges and valleys; evenly distributing nodes in flat areas; and increasing node density in key engineering areas (such as roads and slopes). Typically, the node density is set to 3-5 nodes per hectare in flat areas and 8-12 nodes per hectare in feature areas, ensuring that the geodesic network can fully describe the terrain features and engineering conditions.

[0099] The geodesic calculation unit 232 is communicatively connected to the key point selection unit 231 and is used to calculate the geodesics connecting nodes on the terrain surface. In this embodiment, the geodesic calculation unit 232 uses an improved FastMarching algorithm to calculate the geodesics. A geodesic is the shortest path between two points on a Riemannian manifold, defined as a curve that satisfies the following geodesic equation:

[0100] ,

[0101] in: Represent the geodesic parameter equation, The value ranges from 1 to the coordinate dimension (2 in this example), representing the coordinate components of a point on the curve; For curve parameters; Representing coordinates For parameters The second derivative of , i.e., acceleration; and Representing coordinates and For parameters The first derivative of , i.e., velocity; Christoffel symbols describe the local geometric properties of a surface; summation symbols follow Einstein's summation conventions, summing repeated subscripts and superscripts (from 1 to the coordinate dimension). The geodesic equation indicates that the acceleration of the curve is determined solely by the geometric properties of the surface, with no tangential acceleration component.

[0102] Christopher Symbol Defined as:

[0103] ,

[0104] in: The Christopher symbol has three indices and a total of eight components (when the coordinate dimension is 2). For measuring tensors The components of the inverse matrix; , , For measuring tensors The amount; Represents the metric tensor components coordinates The partial derivatives; the summation notation follows Einstein's summation convention, and repeating subscripts and superscripts are not considered. Summation is performed (from 1 to the coordinate dimension). Christofer notation describes the local geometric properties of the surface and is a key parameter for calculating geodesics.

[0105] The geodesic flow analysis unit 233 is communicatively connected to the geodesic calculation unit 232, and is used to compare the topological and geometric changes of the geodesic network at different time points to quantify terrain changes. In this embodiment, the geodesic flow analysis unit 233 analyzes changes in geometric properties such as geodesic line length and curvature, and calculates geodesic flow indicators characterizing terrain changes. Specifically, for connecting nodes... and The change in length of a geodesic can be expressed as:

[0106] ,

[0107] in: This represents the change in the length of the geodesic line; and They represent time respectively and Time connection node and The length of the geodesic line; and These are two distinct time points, typically two adjacent monitoring periods. Changes in the length of the geodesic line directly reflect changes in the terrain.

[0108] Based on the change in geodesic line length, geodesic flow indexes can be defined. :

[0109] ,

[0110] in: It is a geodetic flow index, a comprehensive measure characterizing the topographic changes of the entire region; This represents the set of edges in a geodesic network, i.e., the set of all geodesics. Indicates the connection node and geodesics; This is a weighting coefficient used to adjust the importance of geodesic line changes in different regions, and is usually between 0 and 1. Represents the absolute value of the change in geodesic length; summation symbol. This represents the summation of all geodesics in the network. In land consolidation projects, key areas (such as slopes and roads) are usually assigned higher weights, for example, key areas are weighted at 0.8-1.0, while general areas are weighted at 0.4-0.6.

[0111] The project progress quantification unit 234 is communicatively connected to the geodetic flow analysis unit 233, and is used to convert terrain changes into project progress indicators such as earthwork completion rate, road hardening rate, and green coverage rate. In this embodiment, the project progress quantification unit 234 establishes a mapping relationship between project design objectives and terrain change indicators, and defines a project completion evaluation function based on geodetic distance.

[0112] Regarding the completion rate of earthwork projects The calculation formula is:

[0113] ,

[0114] in: This represents the completion rate of earthwork projects, with a value ranging from 0 to 1, where 1 indicates complete completion. Indicates the first The design earthwork volume for each area is in cubic meters; This represents the completion coefficient of the area, determined by geodetic flow indicators, and its value ranges from 0 to 1. Total number of regions; summation symbol This indicates summation over all regions. Completion coefficient. Usually related to geodetic flow indicators (area It is directly proportional to the geodetic flow rate, which can be determined by the following relationship: ,in To achieve the target geodesic flow value when the state is completed.

[0115] Road hardening rate and green coverage rate The calculations are similar, determined based on geodetic flow indicators for roads and green areas, respectively.

[0116] Furthermore, the project progress quantification unit 234 also constructs a progress prediction model to predict future progress based on historical trends. In this embodiment, a weighted moving average combined with a differential geometric flow method is used to smooth progress changes over time, improving prediction accuracy. The time window of the prediction model is typically set to 3 to 5 monitoring periods, effectively balancing short-term fluctuations and long-term trends.

[0117] Reference Figure 5 The engineering quality inspection module 3 is used to evaluate the construction quality of the project based on the terrain surface model and differential geometric features provided by the engineering progress image recognition engine 2. In one embodiment of the present invention, the engineering quality inspection module 3 includes an image segmentation unit 31, a slope stability detection unit 32, a drainage system detection unit 33, and a flatness detection unit 34.

[0118] Image segmentation unit 31 is used to subdivide remote sensing images into multiple sub-regions. In this embodiment, image segmentation unit 31 employs a segmentation algorithm based on differential geometric features to segment the image into regions with similar geometric characteristics. The segmentation process considers differential geometric features such as Gaussian curvature and mean curvature, which can effectively identify different functional areas such as slopes, roads, and flat land.

[0119] The slope stability detection unit 32 is communicatively connected to the image segmentation unit 31 and is used to acquire the slope gradient, slope height, and slope length of the slope, and to calculate the slope stability coefficient. In this embodiment, the slope stability detection unit 32 extracts slope parameters based on a differential geometric model and uses the Bishop simplified method to calculate the slope stability coefficient. :

[0120] ,

[0121] in: The slope stability coefficient represents the ratio of resistance force to sliding force. The effective cohesion of the soil is expressed in kPa. The effective internal friction angle of the soil is expressed in degrees. For the first The weight of each strip, in kN; The pressure is the pore water pressure acting on the bottom surface of the strip, expressed in kPa. The width of the strip is in meters (m). The angle between the base of the strip and the horizontal plane, in degrees; summation symbol. This indicates summing over all bars. Because... If it appears on both sides of the equation, it usually needs to be solved by an iterative method. The initial value is set to 1.0, and the iteration error threshold is set to 0.001.

[0122] Under normal circumstances, when At that time, the slope was considered stable; when At that time, the slope is in a critical stable state; when At that time, the slope is at risk of instability. By setting different thresholds, the slope stability detection unit 32 can detect potential slope stability problems in a timely manner.

[0123] The drainage system detection unit 33 is communicatively connected to the image segmentation unit 31 and is used to detect the integrity and functionality of the drainage system. In this embodiment, the drainage system detection unit 33 identifies drainage ditches and checks their connectivity by analyzing the differential geometric features of the terrain surface. Specifically, the drainage system detection unit 33 first identifies the ditch area based on curvature features, and then checks the integrity of the drainage system through geodesic connectivity analysis.

[0124] For ditch identification, the characteristic of the mean curvature H is mainly utilized, as ditch areas typically have a large negative mean curvature. The connectivity check of the drainage system is achieved by analyzing the geodesic topology of the ditch network. If a breakpoint or blockage exists, the drainage system detection unit 33 will mark the corresponding location and generate an early warning message.

[0125] The surface smoothness detection unit 34 is communicatively connected to the image segmentation unit 31 and is used to evaluate surface smoothness based on differential geometric features. In this embodiment, the surface smoothness detection unit 34 evaluates smoothness based on local curvature changes of the terrain surface. Smoothness index Defined as:

[0126] ,

[0127] in: It is a flatness index, with units of 1 / m, representing the average absolute value of the average curvature per unit area; The area to be assessed is in square meters; For point The average curvature at that point, expressed in units of 1 / m; Represents the area element in the parameter field; integral Indicates the entire assessment area Integrating over [the given area], in numerical computation, integration is often discretized into summation: ,in This represents the number of sampling points within the region. For the first The parameter coordinates of each sampling point.

[0128] For an ideally flat surface The value is close to 0; The higher the value, the more uneven the surface. In practical applications, a flatness threshold is set according to engineering requirements. For example, for general land leveling projects, the flatness index threshold is set to 0.05 / m; for road foundations, the threshold is set to 0.02 / m; and for building foundations, the threshold is set to 0.01 / m. The flatness detection unit 34 evaluates the surface flatness quality by comparing the actual flatness with the threshold.

[0129] The meteorological condition compensation module 4 is used to compensate the monitoring results based on meteorological conditions such as sunlight and precipitation at the time of data collection. In a preferred embodiment of the present invention, the meteorological condition compensation module 4 includes a sunlight compensation unit, a precipitation compensation unit, and a ground change compensation unit.

[0130] The illumination compensation unit is used to compensate for the impact of lighting conditions during image capture on the monitoring results. In this embodiment, the illumination compensation unit estimates the lighting conditions during image acquisition based on geographical information and the capture time, and applies an illumination compensation model to adjust the differential geometric features. The illumination compensation model considers factors such as solar altitude angle and azimuth angle to reduce the impact of shadows and uneven lighting on feature extraction.

[0131] The precipitation compensation unit is used to compensate for the impact of precipitation on the monitoring results during image capture. In this embodiment, the precipitation compensation unit estimates the impact of precipitation on surface reflectivity and differential geometric characteristics based on meteorological data, and applies a compensation model for correction. The precipitation compensation model sets different compensation parameters for different types of surfaces (such as bare soil, vegetation, hardened pavement, etc.) to improve the compensation effect.

[0132] The ground change compensation unit is used to compensate for the impact of building construction within the monitoring area on the monitoring results. In this embodiment, the ground change compensation unit identifies non-engineering changes (such as temporary facilities, material stacking, etc.) to eliminate their interference with the project progress assessment. Ground change compensation is based on land feature classification and time series analysis, which can effectively distinguish between engineering changes and non-engineering changes.

[0133] In practical applications, the meteorological condition compensation module 4 acquires regional meteorological data based on longitude, latitude, and time extracted from geographic information, and performs targeted compensation processing based on project progress and quality inspection results. Compensation parameters are typically obtained through fitting with a large amount of experimental data. For example, for illumination compensation, the characteristic intensity needs to be adjusted by 5%–8% for every 10-degree change in solar altitude angle; for precipitation compensation, the reflectance change caused by each millimeter of precipitation is approximately 2%–4%, depending on the surface type.

[0134] Reference Figure 6 The early warning module 5 is used to generate early warning information based on the project progress and project quality inspection results. In one embodiment of the present invention, the early warning module 5 includes a parameter early warning unit 51, a quality early warning unit 52, a comprehensive early warning unit 53, and an early warning push unit 54.

[0135] The parameter early warning unit 51 is used to compare the project progress indicators generated by the project progress image recognition engine 2 with preset thresholds. In this embodiment, the parameter early warning unit 51 monitors indicators such as earthwork completion rate, road hardening rate, and green coverage rate. When these indicators deviate from the planned values ​​to a certain extent, it generates early warning information of corresponding levels. For example, when the actual progress of a certain area lags behind the planned progress by 10% to 20%, a general early warning is issued; when it lags behind by 20% to 30%, an important early warning is issued; and when it lags behind by more than 30%, an emergency early warning is issued.

[0136] The quality early warning unit 52 is used to compare the quality assessment results generated by the engineering quality inspection module 3 with the quality standards. In this embodiment, the quality early warning unit 52 monitors quality indicators such as slope stability, drainage system integrity, and surface flatness. When a quality problem is detected, it generates early warning information of different levels according to the severity.

[0137] The integrated early warning unit 53 is communicatively connected to the parameter early warning unit 51 and the quality early warning unit 52, and is used to generate multi-level early warning information based on the overall project progress and quality status. In this embodiment, the integrated early warning unit 53 determines the final early warning level based on the combined impact of progress deviations and quality problems. The early warning levels are divided into three levels: general early warning (blue), important early warning (yellow), and emergency early warning (red).

[0138] The early warning push unit 54 is communicatively connected to the integrated early warning unit 53 and is used to send early warning information to users via SMS, email, or application push. In this embodiment, the early warning push unit 54 selects different notification methods and frequencies according to the early warning level. General early warnings are sent via in-application messages; important early warnings are sent via application push and email; and emergency early warnings are sent via multiple channels including application push, email, and SMS to ensure that relevant personnel are informed of the early warning information in a timely manner.

[0139] In addition, the early warning push unit 54 will customize different early warning content for users with different roles. For example, for on-site construction personnel, it will mainly push the specific location of quality problems and handling suggestions; for project management personnel, it will push the overall progress status and potential risk analysis; and for decision-makers, it will provide a summary report on the overall project status.

[0140] The differential invariant extraction unit 224 is specifically used for: calculating the family of differential invariants of terrain surfaces at multiple scales; constructing a multi-scale representation based on the Laplace-Beltrami operator; generating differential geometric fingerprints of feature surfaces to achieve robust comparison of terrain at different time points; and using a spatiotemporal covariant comparative analysis method to handle changes in perspective and illumination differences between images acquired at different times.

[0141] In a preferred embodiment of the present invention, the differential invariant extraction unit 224 first calculates a family of differential invariants including Gaussian curvature, mean curvature, principal curvature, and their derivatives. These invariants are rotationally, translationally, and scale-invariant in describing terrain features, and are stable and reliable terrain feature descriptors.

[0142] Multi-scale representation is achieved by applying Laplace-Beltrami operators of different scales to the terrain surface. In practical calculations, the discrete Laplace-Beltrami operator approximation is used:

[0143] ,

[0144] in: For the discrete Laplace-Beltrami operator at the vertex Pair function The effect of the action; As vertices The area of ​​the surrounding region is usually the vertex. The area of ​​the mixed Voronoi region; As vertices The set of adjacent vertices, i.e., the set of vertices adjacent to the vertex. Directly connected vertices; For the edge The weight of the vertex represents the weight of the vertex. and The strength of the connection between them; and functions respectively At the apex and The value at the point; the summation symbol Represents a vertex Summing is performed on all adjacent vertices. In practical applications, weights... Typically set as cotangent weight:

[0145] ,

[0146] in: For the edge The weights; and For the edge The angles on both sides, in radians; This represents the cotangent function. This weighting method reflects the geometric properties of the discrete Laplace-Beltrami operator and can better preserve the geometric characteristics of the surface.

[0147] By analyzing terrain functions (e.g., elevation function) applying different scales The Laplace-Beltrami operator can be used to construct multi-scale representations. :

[0148] ,

[0149] in: For function In scale The following is a representation; The thermonuclear operator is an exponential operator, representing the Laplace-Beltrami operator. conduct The thermal diffusion process over time; This is a scale parameter that controls the smoothness, measured in square meters. In practical applications, it is typically selected as... Multiple scales are used to capture terrain features at different scales. In numerical computation, the thermal kernel operator is typically approximated through the following iterations: ,in The time step is usually set to .

[0150] Based on multi-scale differential invariants, the differential invariant extraction unit 224 generates a differential geometric fingerprint of the feature surface. In this embodiment, for each point p on the terrain surface, its differential geometric fingerprint... Defined as the set of differential invariants at different scales:

[0151] ,

[0152] in: For point The differential geometric fingerprint is a multidimensional feature vector; For scale Lower point Gaussian curvature at that point; For scale Lower point The average curvature at that point; and respectively scale Lower point The maximum and minimum principal curvature at the point; For the selected set of scales, for example Differential geometric fingerprints contain curvature information of the terrain at different scales, and can comprehensively describe the geometric characteristics of the terrain.

[0153] To address the variations in viewing angle and illumination between images acquired at different times, the differential invariant extraction unit 224 employs a spatiotemporal covariant comparative analysis method. This method, based on a geodesic feature point matching algorithm and parallel transport theory, transforms local geometric quantities at different time points to the same coordinate system for comparison. Parallel transport, a fundamental concept in Riemannian geometry, is used to move tangent vectors parallel to a curve. In terrain analysis, parallel transport can be used to compare curvature changes at the same location at different time points, eliminating the influence of viewing angle variations.

[0154] Through these techniques, the differential invariant extraction unit 224 can provide a robust description of terrain features in response to changes in environmental conditions, laying the foundation for accurate change detection.

[0155] The project progress quantification unit 234 is specifically used for: establishing the mapping relationship between project design objectives and terrain change indicators; defining a project completion evaluation function based on geodetic distance; smoothing the progress changes in the time dimension using the differential geometric flow method based on multi-temporal observation results; and constructing a progress prediction model to predict future progress based on historical change trends.

[0156] In a preferred embodiment of the present invention, the project progress quantification unit 234 first establishes a mapping relationship between project design objectives and terrain change indicators. For land consolidation projects, the focus is mainly on three aspects: earthwork, road hardening, and greening. Project design objectives typically include target elevation models, road layouts, and greening plans, which need to be mapped to change indicators based on geodesic networks.

[0157] Specifically, for earthwork projects, completion assessment is based on the geodetic distance between the actual terrain and the target elevation model. Geodetic distance Defined as two surfaces and Minimum distance between:

[0158] ,

[0159] in: curved surface and The geodetic distance between them, in meters; and It consists of two curved surfaces; curved surface The point on; curved surface The point on; Point and Geodetic distance between them; This represents the infimum (minimum) operator, which is the minimum distance among all possible point pairs. In practical calculations, the geodesic distance between two surfaces is usually approximated by sampling and discretization methods.

[0160] Project completion evaluation function based on geodetic distance Defined as:

[0161] ,

[0162] in: This is a function for evaluating the completion of the project, with a value range of [0,1]. The larger the value, the higher the completion rate. Indicates the current terrain surface; Represents the target terrain surface; Represents the initial terrain surface; This represents the geodetic distance between the current terrain and the target terrain. This represents the geodesic distance between the initial terrain and the target terrain. This function measures how close the current terrain is to the target state relative to the initial state; a completion rate of 0 indicates no progress, and a completion rate of 1 indicates that the target state has been fully achieved.

[0163] To smooth out changes in progress over time, the project progress quantification unit 234 employs a differential geometric flow method. In this embodiment, MeanCurvatureFlow is used to smooth the time series:

[0164] ,

[0165] in: Representing a curved surface Over time The rate of change; The mean curvature is expressed in units of 1 / m. The unit normal vector is a three-dimensional unit vector perpendicular to the surface; the negative sign indicates that the flow direction is opposite to the curvature direction, meaning the flow moves towards the direction of lower average curvature. Average curvature flow can effectively remove noise while preserving key features, making it suitable for smoothing changes in flow rate.

[0166] In practical applications, the choice of smoothing parameters is crucial. Typically, the smoothing time step is set to 0.05-0.1, and the number of iterations is 5-10. These parameters can remove short-term fluctuations while preserving the true trend of progress changes.

[0167] Finally, the project progress quantification unit 234 constructs a progress prediction model based on historical trends. In this embodiment, a weighted moving average model combined with trend analysis is used for prediction:

[0168] ,

[0169] in: The degree of completion of the prediction, i.e., the time point. The degree of prediction completion; Historical completion, i.e., point in time. The actual completion rate; These are weighting coefficients, representing the importance of data at different time points; The number of time points used for historical data is typically set to 3-5; summation symbol. Indicates to A weighted sum of historical time points. Weight coefficients. satisfy This ensures that the scale of the predicted values ​​is consistent with the actual level of completion.

[0170] The weighting coefficients are typically set using an exponential decay method:

[0171] ,

[0172] in: For the first Weighting coefficients for each historical time point; This is a decay factor that controls the rate at which the weights decay. As a normalization factor, it ensures that the sum of all weights is 1; and These are the indices for time points, from 1 to... Larger A value (e.g., 0.9) means that the weight of historical data decays more slowly, which is suitable for stable project schedules; a smaller value... Values ​​(such as 0.7) are suitable for projects with rapidly changing schedules.

[0173] In this way, the project schedule quantification unit 234 can provide accurate project schedule assessment and reliable schedule forecast, providing decision support for project management.

[0174] The embodiments described above are merely illustrative of specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A land consolidation project progress monitoring system based on time-series image analysis, characterized in that, include: The data storage module is used to store historical image data of land consolidation projects collected by drones and the corresponding geographic information; An engineering progress image recognition engine, communicatively connected to the data storage module, is used to receive image data provided by the data storage module, construct a terrain surface model based on differential geometry theory, and extract key engineering technical parameters from the terrain surface model. The engineering progress image recognition engine includes an image registration module, a Riemannian geometric terrain modeling module, and a geodesic analysis module. The image registration module is used to register images at different time points to a unified coordinate system. The Riemannian geometric terrain modeling module is used to construct a manifold parameterized representation of the terrain surface and calculate differential invariants. The geodesic analysis module is used to construct a geodesic network and quantify the engineering progress based on changes in the geodesic network. The engineering quality inspection module is communicatively connected to the engineering progress image recognition engine and is used to receive the terrain surface model and differential geometric features provided by the engineering progress image recognition engine to evaluate the engineering construction quality. The meteorological condition compensation module is communicatively connected to the engineering progress image recognition engine and the engineering quality detection module, and is used to compensate the monitoring results according to meteorological conditions such as light and precipitation at the time of collection. The early warning module is communicatively connected to the project progress image recognition engine and the project quality inspection module, and is used to generate early warning information based on the project progress and project quality inspection results.

2. The land consolidation project progress monitoring system based on time-series image analysis according to claim 1, characterized in that: The data storage module stores historical image data of land consolidation projects collected by drones, including aerial flight maps and panoramic images; the geographic information includes the coordinates of ground features, elevation, and photography time, which is used to provide meteorological compensation parameters for the meteorological condition compensation module.

3. The land consolidation project progress monitoring system based on time-series image analysis according to claim 1, characterized in that: The image registration module includes: The key point extraction unit is used to select points with invariant features from the image as key points; The feature description unit is communicatively connected to the key point extraction unit and is used to describe the feature information of the key points using feature vectors. The key point matching unit is communicatively connected to the feature description unit and is used to establish a correspondence between corresponding points in images at different time points; The image conversion unit is communicatively connected to the key point matching unit and is used to convert the two-dimensional plane coordinates of different images to a unified plane coordinate system.

4. The land consolidation project progress monitoring system based on time-series image analysis according to claim 1, characterized in that: The Riemannian geometric terrain modeling module includes: Manifold parameterization unit is used to map terrain point cloud data to parameter space and construct a two-dimensional manifold parameterization representation; The Riemann metric tensor computation unit is communicatively connected to the manifold parameterization unit and is used to compute the metric tensor of the terrain surface to characterize the local geometric properties of the surface. The curvature tensor analysis unit is communicatively connected to the Riemannian metric tensor calculation unit and is used to calculate the Gaussian curvature and mean curvature distribution of the terrain surface and identify key geomorphic features. The differential invariant extraction unit is communicatively connected to the curvature tensor analysis unit and is used to extract terrain geometric features that are robust to viewpoint, illumination, and scanning conditions.

5. The land consolidation project progress monitoring system based on time-series image analysis according to claim 1, characterized in that: The geodesic analysis module includes: Key point selection unit, used to select geodesic network nodes based on curvature and terrain features; The geodesic calculation unit is communicatively connected to the key point selection unit and is used to calculate the geodesic lines connecting the nodes on the terrain surface. The geodesic flow analysis unit is communicatively connected to the geodesic calculation unit and is used to compare the topological and geometric changes of the geodesic network at different time points to quantify terrain changes. The project progress quantification unit is communicatively connected to the geodetic flow analysis unit and is used to convert topographic changes into project progress indicators such as earthwork completion rate, road hardening rate, and green coverage rate.

6. The land consolidation project progress monitoring system based on time-series image analysis according to claim 1, characterized in that: The engineering quality inspection module includes: Image segmentation unit is used to subdivide remote sensing images into multiple sub-regions; The slope stability detection unit is communicatively connected to the image segmentation unit and is used to obtain the slope, slope height and slope length of the slope, and calculate the slope stability coefficient. A drainage system detection unit is communicatively connected to the image segmentation unit and is used to detect the integrity and functionality of the drainage system. The flatness detection unit is communicatively connected to the image segmentation unit and is used to evaluate the flatness of the ground surface based on differential geometric features.

7. The land consolidation project progress monitoring system based on time-series image analysis according to claim 1, characterized in that: The meteorological condition compensation module includes: The illumination compensation unit is used to compensate for the impact of illumination on the monitoring results during image capture; The precipitation compensation unit is used to compensate for the impact of precipitation on the monitoring results during image capture. Ground change compensation unit is used to compensate for the impact of building construction on monitoring results within the monitoring range; The illumination compensation unit, precipitation compensation unit, and ground change compensation unit acquire regional meteorological data based on the longitude, latitude, and time extracted from the geographic information, and perform targeted compensation processing based on the project progress and quality inspection results.

8. The land consolidation project progress monitoring system based on time-series image analysis according to claim 1, characterized in that: The early warning module includes: The parameter warning unit is used to compare the engineering progress indicators generated by the engineering progress image recognition engine with a preset threshold. A quality early warning unit is used to compare the quality assessment results generated by the engineering quality inspection module with the quality standards. The integrated early warning unit is communicatively connected to the parameter early warning unit and the quality early warning unit, and is used to generate multi-level early warning information based on the overall project progress and quality status. The early warning push unit is communicatively connected to the integrated early warning unit and is used to send early warning information to users via SMS, email, or application push.

9. The land consolidation project progress monitoring system based on time-series image analysis according to claim 4, characterized in that: The differential invariant extraction unit is specifically used for: Calculate the family of differential invariants of topographic surfaces at multiple scales; Constructing a multi-scale representation based on the Laplace-Beltrami operator; Generate differential geometric fingerprints of feature surfaces to achieve robust comparisons of terrain at different time points; A spatiotemporal covariance comparative analysis method was used to process the changes in perspective and illumination between images acquired at different times.

10. The land consolidation project progress monitoring system based on time-series image analysis according to claim 5, characterized in that: The project progress quantification unit is specifically used for: Establish a mapping relationship between engineering design objectives and terrain change indicators; The project completion evaluation function is defined based on geodetic distance. Based on multi-temporal observation results, the differential geometric flow method is used to smooth the progress changes in the time dimension; Build a progress prediction model to predict future progress based on historical trends.

Images