A method and device for long-term comparison and analysis based on landscape three-dimensional scenes

Abstract

The present invention discloses a long time series comparative analysis method and device based on landscape three-dimensional scene, which belongs to the field of landscape data analysis, wherein the method includes: reading multi-source landscape data and performing pre-processing and fitting with homologous data, performing timestamp identification, and determining effective scene data; performing big data mining and building a definition state library; combining the definition state library to establish a long time series comparison model; performing low-feature three-dimensional simulation based on effective scene data, building a landscape three-dimensional scene, performing qualitative and quantitative spatial-temporal evolution comparative analysis and mutual verification, determining long time series evolution trends, and adding long time series comparative analysis results. This application solves the technical problem that the prior art lacks effective multi-source data integration and long time series comparative analysis means, and it is difficult to comprehensively and accurately depict the temporal and spatial evolution laws of landscape patterns, and achieves the technical effect of comprehensively and accurately depicting the temporal and spatial evolution laws of landscape patterns through data fusion and qualitative and quantitative spatial-temporal evolution analysis.

Description

Technical Field

[0001] The present invention relates to the field of landscape data analysis, and in particular to a long-term time series comparative analysis method and device based on a three-dimensional landscape scene. Background Art

[0002] With the increasing impact of environmental changes and human activities, the landscape pattern is undergoing rapid changes. Accurately grasping the dynamic changes of landscape in time and space and predicting future landscape evolution trends are of great significance for regional sustainable development. Traditional landscape pattern analysis methods mainly rely on two-dimensional plane data, such as satellite images, land use maps, etc., which are difficult to fully reflect the three-dimensional structural characteristics of the landscape. At the same time, due to the lack of large-scale and long-term quantitative analysis methods, the current understanding of the dynamic change process of the landscape is still relatively limited. Although there are methods that try to use three-dimensional scenes and time series data for landscape analysis, there are still deficiencies in the effective integration of multi-source heterogeneous data and long-term quantitative comparative analysis, which makes it difficult to fully and accurately describe the spatiotemporal evolution of the landscape pattern. Summary of the invention

[0003] This application provides a long-term comparative analysis method and device based on a three-dimensional landscape scene, aiming to solve the technical problem that the prior art lacks effective multi-source data integration and long-term comparative analysis methods, making it difficult to comprehensively and accurately depict the temporal and spatial evolution laws of landscape patterns.

[0004] In view of the above problems, the present application provides a long-term time series comparative analysis method and device based on landscape three-dimensional scenes.

[0005] The first aspect disclosed in the present application provides a long-term time series comparative analysis method based on a three-dimensional landscape scene, the method comprising: reading multi-source landscape data and performing preprocessing and fitting with homologous data, performing timestamp identification, and determining effective scene data, wherein the effective scene data is the spatiotemporal distribution data of the monitoring area; performing big data mining, building a defined state library, wherein the defined state library is determined based on the landscape pattern type-pattern characteristics-driving factors-factor driving relationship; combining with the defined state library, establishing a long-term time series comparison model, wherein the long-term time series comparison model includes a qualitative comparison branch and a quantitative comparison branch in parallel, and a lateral interaction branch is established; connecting to a visualization simulation platform, based on the effective scene data, Low-feature three-dimensional simulation is performed to build a three-dimensional landscape scene, wherein the three-dimensional simulation includes physical simulation and virtual simulation, and wherein the three-dimensional landscape scene has seasonal and temporal characteristics; based on the long-term comparison model and taking the standardized landscape as the baseline, the three-dimensional landscape scene is subjected to qualitative and quantitative spatial-temporal evolution comparative analysis and mutual verification to determine the long-term evolution trend, wherein the long-term evolution trend includes real-time deviation trend and predicted deviation trend, and the long-term evolution trend is synchronously updated with the passage of time; the long-term evolution trend is integrated and added into the long-term comparison analysis results, and the deviation exceeding limit positioning and landscape maintenance management are performed based on the real-time deviation trend and the predicted deviation trend.

[0006] Another aspect disclosed in the present application provides a long-term time series comparison and analysis device based on a three-dimensional landscape scene, the device comprising: a scene data determination unit, used to read multi-source landscape data and perform preprocessing and fitting with homologous data, perform timestamp identification, and determine effective scene data, wherein the effective scene data is the spatiotemporal distribution data of the monitoring area; a state library construction unit, used to perform big data mining, build a defined state library, and the defined state library is determined based on the landscape pattern type-pattern characteristics-driving factors-factor driving relationship; a model building unit, used to combine the defined state library to establish a long-term time series comparison model, the long-term time series comparison model includes a qualitative comparison branch and a quantitative comparison branch in parallel, and a lateral interaction branch is established; a three-dimensional scene construction unit, used to connect the visualization simulation The platform performs low-feature three-dimensional simulation based on effective scene data and builds a three-dimensional landscape scene, wherein the three-dimensional simulation includes entity simulation and virtual simulation, and wherein the three-dimensional landscape scene has seasonal and temporal characteristics; an evolution trend determination unit is used to perform qualitative and quantitative spatial-temporal evolution comparison analysis and mutual verification of the three-dimensional landscape scene based on a long-term comparison model and a standardized landscape as a baseline to determine the long-term evolution trend, wherein the long-term evolution trend includes a real-time deviation trend and a predicted deviation trend, and the long-term evolution trend is synchronously updated with the passage of time; an analysis result acquisition unit is used to integrate the long-term evolution trend, add the long-term comparison analysis result, and perform deviation limit positioning and landscape maintenance management based on the real-time deviation trend and the predicted deviation trend.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The method adopts the method of reading multi-source landscape data and performing preprocessing and fitting with homologous data, performing timestamp identification, and determining effective scene data. The effective scene data is the spatiotemporal distribution data of the monitoring area. By establishing a unified spatiotemporal benchmark, it lays the foundation for subsequent data analysis; big data mining is carried out, and a definition state library is built. The definition state library is determined based on the landscape pattern type-pattern characteristics-driving factors-factor driving relationship, so as to realize the semantic expression and structured organization of landscape elements and form an analysis-oriented definition state library; combined with the definition state library, a long time series comparison model is established. The long time series comparison model includes qualitative comparison branches and quantitative comparison branches in parallel, and lateral interactive branches are established to realize the comparison analysis paradigm combining qualitative and quantitative, which can not only intuitively present the evolution process of the landscape pattern, but also accurately depict its dynamic change law; connected to the visualization simulation platform, low-feature three-dimensional simulation is carried out based on the effective scene data, and a three-dimensional landscape scene is built. Among them, the three-dimensional landscape scene has seasonal and temporal characteristics, which is for intuitive display and interaction. The system provides a platform for operation; based on the long-term comparison model and taking the standardized landscape as the baseline, the three-dimensional landscape scene is subjected to qualitative and quantitative temporal and spatial evolution comparative analysis and mutual verification to determine the long-term evolution trend. The long-term evolution trend includes the real-time deviation trend and the predicted deviation trend. The long-term evolution trend is updated synchronously with the passage of time. By identifying the historical evolution trend and the predicted evolution trend, the pattern evolution and the process mechanism are closely combined to realize the dynamic monitoring and simulation of the landscape across time phases, and reveal the temporal and spatial coupling mechanism of the landscape pattern and the ecological process; the long-term evolution trend is integrated, and the long-term comparative analysis results are added. The technical solution of deviation over-limit positioning and landscape maintenance management based on the real-time deviation trend and the predicted deviation trend is solved, which solves the technical problem that the existing technology lacks effective multi-source data integration and long-term comparative analysis methods, and it is difficult to comprehensively and accurately describe the temporal and spatial evolution laws of the landscape pattern, and achieves the technical effect of comprehensively and accurately describing the temporal and spatial evolution laws of the landscape pattern through data fusion and qualitative and quantitative temporal and spatial evolution analysis.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flowchart of a long-term comparison and analysis method based on a landscape three-dimensional scene is provided for an embodiment of the present application;

[0011] Figure 2A structural schematic diagram of a long-term time series comparison and analysis device based on a landscape three-dimensional scene is provided for an embodiment of the present application.

[0012] Explanation of the reference numerals: scene data determination unit 11 , state library construction unit 12 , model establishment unit 13 , three-dimensional scene construction unit 14 , evolution trend determination unit 15 , analysis result acquisition unit 16 . DETAILED DESCRIPTION

[0013] The overall idea of ​​the technical solution provided by this application is as follows:

[0014] The embodiment of the present application provides a method and device for long-term comparative analysis based on a three-dimensional landscape scene. First, through timestamp identification, multi-source heterogeneous landscape data are integrated into a unified spatiotemporal benchmark, and a structured definition state library is constructed to realize the semantic expression and reasoning analysis of landscape information. Secondly, under a normalized spatiotemporal framework, a long-term comparison model containing two branches of qualitative comparison and quantitative comparison is designed to establish the intrinsic connection between landscape pattern and ecological process, and to reveal the dynamic evolution mechanism of landscape from multiple angles and levels. Thirdly, three-dimensional visualization technology is introduced to convert abstract data into realistic landscape scenes, build a three-dimensional landscape scene, and make the analysis results more intuitive and easy to understand. Afterwards, based on the long-term comparison model, the three-dimensional landscape scene is subjected to qualitative and quantitative spatiotemporal evolution comparative analysis and mutual verification, and the future development trend of the landscape is intelligently predicted to obtain comprehensive and accurate long-term comparative analysis results.

[0015] In summary, this application overcomes the limitations of previous landscape spatiotemporal analysis through a variety of technical means such as data-driven, qualitative and quantitative analysis, realizes cross-time, cross-regional, and cross-scale pattern evolution analysis, and improves the accuracy and comprehensiveness of the analysis.

[0016] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification. Example

[0017] like Figure 1 As shown, the embodiment of the present application provides a long time series comparison and analysis method based on a landscape three-dimensional scene, the method comprising:

[0018] S1: Read multi-source landscape data and perform preprocessing and fitting with homologous data, perform timestamp identification, and determine effective scene data, which is the spatiotemporal distribution data of the monitoring area.

[0019] Furthermore, valid scene data is determined, including:

[0020] Guided by the data preprocessing requirements, supervised training of the preprocessing model is performed, wherein the preprocessing model includes a source processing block and a multi-source fusion block; based on the source processing block, redundant processing and registration enhancement processing are performed on the multi-source landscape data to determine the effective multi-source data, wherein the redundant processing includes redundant screening and redundant dimensionality reduction; the effective multi-source data is traversed, data coaxial transformation and homologous data fusion are performed based on a unified coordinate system, and the effective scene data is integrated and determined.

[0021] In the embodiment of the present application, first, multi-channel multi-source landscape data is used, for example, the landscape information covering the whole area and multiple phases is obtained by remote sensing acquisition of the ground landscape by sensors carried on satellites; the detailed information of the local area is acquired with high resolution by the sensor equipment carried by low-altitude flying vehicles; the landscape video picture data collected by the fixed or mobile video surveillance equipment deployed in the area is used to obtain the near-ground sensing data, reflecting the dynamic change process of the landscape. Then, according to the data preprocessing requirements, that is, the requirements of landscape three-dimensional scene construction and long-term comparative analysis, the goals and standards of data preprocessing are determined, such as the spatiotemporal resolution, coverage, feature type, etc. of the data. Based on these requirements, the data preprocessing model is trained by supervised learning. The model consists of two parts: a source processing block and a multi-source fusion block. The source processing block is responsible for processing landscape data from a single source, such as noise removal, feature enhancement, etc. The multi-source fusion block is responsible for aligning and fusing data from different sources to form a consistent data expression.

[0022] Subsequently, the source processing block of the preprocessing model is used to perform redundancy processing and registration enhancement processing on the landscape data from each source. Among them, redundancy processing includes two sub-steps: redundancy screening and redundancy dimensionality reduction. Redundancy screening refers to removing duplicate information in the data, such as repeated collection of the same area by different sensors at the same time. Redundancy dimensionality reduction refers to reducing the dimension of the data while maintaining the essential characteristics of the data, such as band selection for hyperspectral data. In addition to redundancy processing, registration enhancement processing is also required, that is, preliminary spatiotemporal registration is performed at the raw data level to prepare for subsequent fine registration. After being processed by the source processing block, the original multi-source landscape data is simplified into effective multi-source data with smaller data volume and higher quality. Then, the effective multi-source data obtained by traversing is deeply processed using the multi-source fusion block of the preprocessing model. All data are converted to a unified spatiotemporal benchmark, that is, data coaxial conversion is performed. This process needs to consider the projection mode, coordinate system and timestamp of different data, and perform necessary conversion and interpolation. On this basis, multi-source fusion algorithms, such as weighted average and principal component analysis, are used to fuse landscape representations from different sources to form more comprehensive and accurate homologous data. Afterwards, the homologous data are integrated, and unqualified data are eliminated to form the final effective scene data, which is the spatiotemporal distribution data of the monitoring area, recording the distribution status of landscape elements at specific time points and specific geographical locations, and reflecting the dynamic change characteristics of landscape patterns in time and space.

[0023] S2: Conduct big data mining and build a definition state library, where the definition state library is determined based on landscape pattern type-pattern characteristics-driving factors-factor driving relationships.

[0024] Furthermore, S2 includes:

[0025] Identify the mapping relationship between the pattern characteristics and driving factors, and determine multiple mapping groups; traverse the multiple mapping groups, measure the evolution of the pattern characteristics under the unit driving step based on single-factor driving and multi-factor joint driving, and determine multiple groups of factor driving relationships, wherein the multiple mapping groups correspond one-to-one to the multiple groups of factor driving relationships; integrate the multiple groups of factor driving relationships, and perform occasional marking of the pattern characteristics-driving factors.

[0026] In the embodiments of the present application, big data mining technology is used to discover hidden patterns and laws from massive landscape spatiotemporal data, and to construct a definition state library that can reflect the mechanism of landscape evolution. The state library is built with landscape pattern type-pattern characteristics-driving factors-factor-driven relationship as the basic framework. Among them, landscape pattern type refers to the components of the landscape and its spatial configuration, such as forest, farmland, city and other landscape types; pattern characteristics refer to the spatial geometric characteristics and topological relationships of the landscape pattern, such as the number of patches, area, shape, connectivity, etc.; driving factors refer to the internal reasons that cause changes in the landscape pattern, including both natural factors such as topography, soil, climate, etc., and human factors such as land use methods; factor-driven relationship describes the interaction mechanism between driving factors and landscape pattern, that is, how the landscape pattern evolves under the action of specific driving factors.

[0027] First, starting from the basic framework of defining the state library, the attribute set of pattern characteristics and driving factors is extracted. Then, the association rule algorithm in data mining, such as Apriori, FP-growth, etc., is used to analyze the association strength and support between pattern characteristics and driving factors, and identify significant mapping relationships. For example, it can be expressed as "IF driving factor A THEN pattern feature B", which means that under the action of driving factor A, the landscape pattern shows feature B. By clustering and merging association rules, multiple mapping groups are obtained, each of which consists of multiple related mapping relationships, reflecting the many-to-many mapping relationship between driving factors and pattern characteristics. Then, multiple mapping groups are traversed, and one mapping group is obtained each time. Taking the mapping group as a unit, the effects of single-factor driving and multi-factor joint driving on the evolution of landscape pattern characteristics are examined respectively. For each mapping group, it is first assumed that only one driving factor changes and the other factors remain unchanged. The response of pattern characteristics under the unit driving step (such as every 1°C increase in temperature) is measured to obtain the feature evolution sequence under single-factor driving; then, considering the situation where multiple driving factors change at the same time, the interaction and cumulative effect between them are analyzed, and the evolution of pattern characteristics under joint driving is measured. By comparing the differences in the effects of single-factor driving and multi-factor driving, we can identify driving factors and their primary and secondary relationships, and form multiple factor-driven relationship groups. Each mapping group corresponds to a set of factor-driven relationships, revealing the causal logic and mechanism of action between the driving factors and the evolution of pattern characteristics within the mapping group. Subsequently, the similarities and differences in the factor-driven relationships between different mapping groups are compared to identify common driving patterns, such as the driving effects of factors such as economic development level and traffic conditions in the process of urban expansion in different regions. Then, logical reasoning and causal modeling are used to mark the pattern characteristics and driving factors, that is, to indicate the "IF-THEN" logical relationship and causal transmission path between them, to form a driving rule library suitable for the corresponding landscape pattern type and scale, and to build a defined state library.

[0028] S3: In combination with the defined state library, a long time series comparison model is established, wherein the long time series comparison model includes a qualitative comparison branch and a quantitative comparison branch in parallel, and a lateral interaction branch is established.

[0029] In the embodiments of the present application, a long time series comparison model is established by taking the constructed definition state library as the knowledge basis and adopting a framework combining qualitative analysis with quantitative analysis and coordinating time scale with spatial scale, so as to realize multi-dimensional and multi-granular analysis of the landscape pattern evolution process.

[0030] The long-term comparison model consists of two parallel branches, namely the qualitative comparison branch and the quantitative comparison branch, and also has a lateral interaction branch. The qualitative comparison branch and the quantitative comparison branch characterize and compare the spatiotemporal evolution characteristics of the landscape pattern from different perspectives, respectively, and realize information fusion and joint analysis through the lateral interaction branch. Among them, the qualitative comparison branch focuses on describing and comparing the process and stage characteristics of the landscape pattern evolution. Based on the landscape pattern type and factor-driven relationship in the defined state library, this branch uses time series matching, pattern recognition and other methods to identify the key nodes and turning points of landscape evolution, divide the stages and trends of evolution, and qualitatively compare the landscape patterns of different periods, revealing the directionality and mutation of evolution, and obtaining an overall understanding and stage diagnosis of the long-term evolution of the landscape pattern. The quantitative comparison branch focuses on measuring and comparing the quantitative characteristics and rate characteristics of landscape pattern evolution. Based on the pattern characteristics and driving factors in the defined state library, this branch uses time series analysis, spatial statistics and other methods to quantitatively evaluate the changes in composition, structure and function of landscape patterns at different temporal and spatial scales, characterize the quantitative characteristics such as the amplitude, speed and acceleration of evolution, and compare the evolution rates in different regions and periods to reveal the regional differences and stages of evolution, and obtain a refined characterization of the long-term evolution of landscape patterns and the spatiotemporal differentiation of change rates.

[0031] In addition to the two parallel branches of qualitative comparison branch and quantitative comparison branch, the long-term comparison model also sets up a lateral interaction branch to coordinate and integrate the analysis results of the qualitative comparison branch and the quantitative comparison branch. On the one hand, the results of qualitative comparison provide the definition of time segments and semantic description of evolutionary stages for quantitative comparison, helping to explain the causes and mechanisms of changes in quantitative indicators; on the other hand, the results of quantitative comparison provide quantitative basis and statistical verification for qualitative comparison, improving the reliability of evolutionary trend judgment. Through lateral interaction, qualitative branches and quantitative branches can verify and complement each other, forming a three-dimensional and panoramic depiction of the long-term evolution law of landscape pattern.

[0032] S4: Connecting to a visualization simulation platform, performing low-feature three-dimensional simulation based on the effective scene data, and building a three-dimensional landscape scene, wherein the three-dimensional simulation includes entity simulation and virtual simulation, and wherein the three-dimensional landscape scene has seasonal and temporal characteristics.

[0033] Furthermore, low-feature three-dimensional simulation is performed based on the effective scene data to build a landscape three-dimensional scene, including:

[0034] Based on the evolution requirement, the scene feature refinement is determined, and the evolution requirement is the low feature processing standard; based on the scene feature refinement, the effective scene data is subjected to scene feature simplification and feature conversion processing to determine the scene low feature distribution; based on the scene low feature distribution, the landscape is restored in three-dimensional simulation to determine the landscape three-dimensional scene.

[0035] Furthermore, based on the low-feature distribution of the scene, a three-dimensional realistic restoration of the landscape is performed, including:

[0036] The low-feature distribution of the scene is traversed to determine the physical feature distribution and the virtual feature distribution; based on the physical feature distribution, landscape entity modeling is performed to determine the physical three-dimensional model; based on the light change as a reference, virtual periodic evolution modeling is performed in combination with the virtual feature distribution to determine the virtual three-dimensional model, wherein the light change period is the virtual modeling standard; the physical three-dimensional model and the virtual three-dimensional model are coordinated to generate the landscape three-dimensional scene.

[0037] In a preferred embodiment, in order to intuitively display the spatiotemporal evolution of the landscape pattern, an external visualization simulation platform needs to be connected. The visualization simulation platform is a software system that integrates a variety of three-dimensional visualization and simulation analysis functions, which can model, render and interact with complex geographic spatial data and processes, and provide a good analysis environment. By connecting to the visualization simulation platform, the effective scene data is converted into a vivid and realistic three-dimensional visualization scene, and a variety of perspectives and operation methods are provided for interactive exploration and analysis. For example, using the virtual earth platform Google Earth, remote sensing images, land use maps, etc. of different periods are superimposed and displayed, and the evolution of landscape types and patterns is dynamically observed through the timeline control; using the three-dimensional GIS platform ArcGIS Pro, the terrain undulations, surface cover, etc. of the landscape pattern are three-dimensionally modeled and rendered, and a physical model of the landscape is generated, and the realism of the landscape is improved by adding environmental elements such as lighting and materials.

[0038] By connecting to an external visualization simulation platform, using the acquired effective scene data, and adopting low-feature 3D modeling technology, a digital twin that is highly similar to the real landscape is constructed, namely, a 3D landscape scene. The construction process of a 3D landscape scene includes two types of 3D simulation: physical simulation and virtual simulation. Physical simulation emphasizes the realistic reproduction of the physical space form of the landscape, while virtual simulation emphasizes the dynamic simulation of the landscape environment atmosphere. Physical simulation refers to the 3D reconstruction of the physical elements of the landscape, such as terrain, landforms, and buildings. The geometric information and texture information of the landscape entity elements are extracted using high-resolution remote sensing images, lidar point clouds and other data. Then, 3D modeling software such as 3dsMax and SketchUp are used to perform low-polygon modeling of the landscape entity. Under the premise of ensuring the overall effect of the landscape, the number of faces and texture size of the model are minimized, thereby improving the rendering efficiency and interactive performance of the scene. Virtual reality is to simulate the non-physical elements of the landscape, such as lighting, clouds and other environmental effects. According to the geographical location and seasonal characteristics of the landscape, the azimuth, altitude and light intensity of sunlight at different times are simulated, and the light and dark changes on the surface of the landscape entity are calculated to generate lighting effects that conform to natural laws; the evolution process of atmospheric phenomena such as clouds and fog over the landscape is simulated to improve the realism of the landscape. At the same time, the three-dimensional landscape scene has seasonal and temporal characteristics, that is, when constructing the three-dimensional landscape scene, the dynamic changes of landscape elements with seasons and time are considered and truly reproduced. Among them, seasonality refers to the periodic change law of landscape pattern with the change of seasons. In different seasons, the state of landscape elements such as vegetation and water bodies will change significantly, such as vegetation growing leaves and dense leaves in spring and summer, withering and falling leaves in autumn and winter; water bodies freeze in winter and flow freely in summer. Temporal characteristics refer to the long-term and continuous change trend of landscape pattern over time, which reflects the evolution process of landscape, such as urban expansion leading to a decrease in vegetation area year by year, and water conservancy projects changing the morphology of rivers and lakes. In order to truly reproduce the seasonal and temporal changes of the landscape, when constructing the three-dimensional scene, the parameters of the landscape element model, such as material, color, volume and other attribute parameters, are dynamically adjusted according to the seasonal and year information described by the monitoring data, so that it can realistically show the changes of the four seasons and annual changes. When dealing with seasonal changes, the periodic phenological process of the landscape is simulated, such as simulating the budding, flowering, and leaf-falling cycles of vegetation; when dealing with temporal changes, the continuous year-to-year changes of the landscape are simulated to ensure that the landscape patterns of different years can be reasonably connected and smoothly transitioned. By making the three-dimensional landscape scene have seasonal and temporal characteristics, it can truly reflect the dynamic changes of the landscape pattern with seasons and time, so that it has good spatiotemporal continuity and authenticity, and provides a virtual simulation basis for subsequent long-term comparative analysis.

[0039] When low-feature three-dimensional simulation is performed based on effective scene data and a three-dimensional landscape scene is constructed, the level of detail and performance accuracy of the simulation should be set according to the evolution requirements, and the scene feature precision should be determined. Since this application focuses on the overall evolutionary characteristics of the landscape pattern, the evolution requirements are set to low-feature processing standards during modeling, that is, the geometric accuracy and texture resolution of individual landscape elements are reduced, and the macroscopic morphology and distribution patterns of the landscape pattern are highlighted to balance realism and computational efficiency. For example, for the interannual changes in forest coverage across the entire region, a rough scene feature expression is used to highlight the geometric outline and topological relationship of the landscape, and to downplay the texture details of local features, which can both reduce data redundancy and highlight the macroscopic pattern of global evolution. Then, according to the determined scene feature precision, the acquired effective landscape data is subjected to low-feature processing, key features that have an important impact on evolutionary analysis are extracted, and converted into data formats and organizational forms suitable for three-dimensional modeling, a low-feature expression of the landscape three-dimensional scene is generated, and the scene low-feature distribution is determined.

[0040] Then, based on the low-feature distribution of the scene, the physical features and virtual features of the landscape elements are further distinguished. The so-called physical features refer to the physical entity parts of the landscape elements, such as terrain, objects, buildings, etc. that can be directly observed and measured; virtual features refer to non-physical effects attached to physical objects, such as lighting shadows, reflection and refraction, atmospheric smog, etc. Traverse the low-feature distribution of the scene, and judge whether it is physical or virtual according to the spatial dimensions, temporal characteristics, physical properties, etc. of the low-feature of the scene. For example, elements with clear spatial coordinates and relatively stable forms, such as DEM elevation points and vegetation patches, are physical features; while elements that are closely related to lighting and atmospheric conditions and show dynamic changes, such as shadows and halos, can be classified as virtual features. By traversing the low-feature distribution of the scene, the physical feature distribution and the virtual feature distribution are determined. Subsequently, the corresponding physical 3D model is generated using 3D reconstruction technology for the physical feature distribution of the landscape. First, the terrain is reconstructed. For example, using terrain data such as elevation points and contour lines, continuous terrain surfaces are generated through methods such as TIN and GRID. The terrain surfaces are smoothed and texture mapped to simulate real terrain undulations and landform features. Hydrological analysis tools are used to generate water elements such as river networks and lakes on the terrain surface. Then, the objects are constructed. According to the classification information in the entity features, such as vegetation type and building type, the corresponding object models are generated on the terrain surface using the three-dimensional symbol library and parametric modeling technology. For irregular natural elements such as vegetation, the random graphics method is used to simulate their spatial distribution and morphological characteristics; for artificial elements such as buildings, the voxel method and CSG are used to depict their geometric structure and topological relationship. The next step is landscape combination. On the basis of generating terrain and object models, the arrangement and combination of various landscape elements are reasonably organized according to the spatial adjacency relationship in the distribution of entity features, and the elevation matching and collision detection between terrain and objects are considered to make the landscape combination conform to the real spatial pattern and logical relationship, and obtain a solid three-dimensional model.

[0041] At the same time, it focuses on the three-dimensional dynamic simulation of the virtual feature distribution of the landscape, emphasizes the lighting conditions as the dominant factor, depicts the virtual effect of the landscape under the change of light and shadow, and obtains the virtual three-dimensional model. The core of virtual modeling is to grasp the law of light change, and simulate the virtual presentation of landscape elements under different lighting conditions. Specifically, first obtain the geographical location, time range and other information of the landscape location, and refer to the astronomical calendar to calculate the solar altitude angle and azimuth at different times, and then combine the terrain and landform information in the landscape entity model, use ray tracing, light mapping and other technologies to simulate the reflection, refraction, scattering and other processes of sunlight on the surface of the landscape entity, and generate the landscape light and shadow effects that change with time, realize the virtual periodic evolution modeling, and obtain the virtual three-dimensional model. After that, check the consistency of the physical three-dimensional model and the virtual three-dimensional model in the spatial references such as coordinate system and scale to ensure that the two types of models can be accurately superimposed, and then comprehensively analyze the parameters such as the surface material of the entity and the orientation of the virtual light source to coordinate the light and dark transition of the entity lighting and the virtual light and shadow. Through the full coordination of physical 3D models and virtual 3D models, a high-fidelity landscape 3D scene with spatial accuracy, light and shadow texture, and dynamic artistic conception is formed.

[0042] S5: Based on the long-term time series comparison model and taking the standardized landscape as the baseline, the three-dimensional landscape scene is subjected to qualitative and quantitative spatiotemporal evolution comparative analysis and mutual verification to determine the long-term time series evolution trend. The long-term time series evolution trend includes a real-time deviation trend and a predicted deviation trend. The long-term time series evolution trend is synchronously updated with the passage of time.

[0043] Furthermore, the embodiment of the present application also includes:

[0044] Among them, the qualitative comparison branch includes a twin network structure with the quantitative comparison branch, which respectively includes a time span block and a space span block. The time span blocks between the branches establish a first lateral interaction channel, and the space span blocks between the branches establish a second lateral interaction channel. Based on the qualitative comparison branch and the quantitative comparison branch, parallel evolution analysis and lateral interaction analysis are performed on the three-dimensional landscape scene, and time-space fitting is performed to determine the long-term evolution trend.

[0045] Further, performing spatiotemporal fitting to determine the long-term time series evolution trend includes:

[0046] Based on the first evolution trend of the time span, multiple mutation evolution nodes are determined; the multiple mutation evolution nodes are traversed to determine the node evolution trend of the spatial span as the second evolution trend; the first evolution trend is integrated with the second evolution trend to determine the historical evolution trend; based on the historical evolution trend and the historical prediction bias as a constraint, landscape evolution prediction is performed in a preset time zone to determine the prediction deviation trend, and the prediction deviation trend is determined based on the standardized bias of the prediction evolution trend.

[0047] In the embodiment of the present application, the long-term comparison model established before is used to analyze and compare the spatial-temporal evolution of the landscape three-dimensional scene at both qualitative and quantitative levels, and the two analysis results are mutually verified and integrated to determine the long-term evolution trend of the landscape pattern, including the real-time deviation trend and the predicted deviation trend. At the same time, a dynamic update mechanism for the long-term evolution trend is designed to ensure that the analysis results are synchronized with the actual changes in the landscape pattern. Specifically, with a certain standardized landscape as a reference, the three-dimensional landscape scene to be analyzed is input into the long-term comparison model, the spatial-temporal evolution of the landscape pattern is qualitatively described and the process is analyzed, and the changes in the quantitative indicators of the pattern are measured and evaluated. Through the mechanism of mutual interaction, the quantitative and qualitative analysis results are verified and integrated to form a comprehensive understanding of the long-term evolution law of the landscape pattern. On this basis, the long-term evolution trend of the landscape pattern is output, including the real-time deviation trend and the predicted deviation trend. Among them, the real-time deviation trend reflects the deviation of the current landscape pattern from the long-term trend; and the predicted deviation trend predicts the abnormal changes that may occur in the landscape in the future period. At the same time, as new monitoring data is continuously acquired, the long-term comparison model automatically reruns, updates the evolution trajectory of the historical stage, recalculates the real-time deviation trend, and re-simulates the predicted deviation trend, thereby obtaining a long-term evolution trend description synchronized with the latest data. In a preferred embodiment, after acquiring the long-term comparison model and the three-dimensional landscape scene, the constructed long-term comparison model is used to perform multi-dimensional and multi-scale spatiotemporal comparison analysis on the generated three-dimensional landscape scene through the qualitative comparison branch, quantitative comparison branch and lateral interaction branch in the long-term comparison model, so as to comprehensively characterize the evolution law and development trend of the landscape pattern. Among them, the qualitative comparison branch includes the quantitative comparison branch, which is composed of a time span block and a space span block; the time span block is responsible for processing the feature extraction and change characterization of landscape data in different periods; the space span block is responsible for processing the pattern recognition and association analysis of landscape data in different regions. Between the branches, the time span block realizes information exchange and coordination through the first lateral interaction channel. For example, the typical landscape images of a certain historical period in the qualitative comparison branch can be used as reference samples to input into the corresponding period of the quantitative comparison branch to guide the latter to optimize the extraction method of temporal features; conversely, the landscape temporal index curve of the quantitative comparison branch can also be fed back to the qualitative comparison branch as background knowledge to help calibrate the process division of landscape evolution. Similarly, spatial span blocks are matched and integrated through the second lateral interaction channel. For example, the heterogeneous partitioning of the regional landscape pattern identified by the qualitative comparison branch can be used as prior knowledge to guide the quantitative comparison branch to carry out spatial clustering and differentiation law mining; conversely, the spatial correlation intensity of landscape pattern elements revealed by the quantitative comparison branch also supplements the qualitative comparison branch with a new perspective to explain the similarity of regional landscapes.Through the design of twin network topology and interactive channels, the long-term comparison model can achieve functional complementarity and advantage fusion of qualitative comparison branches and quantitative comparison branches in the two key dimensions of time and space, thereby forming a more comprehensive and in-depth characterization of the long-term evolution of landscape patterns.

[0048] Then, the three-dimensional landscape scene is input into the long-term comparison model. The long-term comparison model inputs the three-dimensional landscape scene into the qualitative comparison branch and the quantitative comparison branch respectively. The parallel evolution analysis is carried out by using the respective time span blocks and space span blocks. The typical characteristics of the landscape pattern in different periods and regions are extracted and compared simultaneously and independently to reveal the evolutionary laws from their respective perspectives. On the basis of parallel analysis, the qualitative comparison branch and the quantitative comparison branch carry out lateral interaction analysis with the help of the first lateral interaction channel and the second lateral interaction channel, breaking the independent parallel state of the two branches. Within the two span blocks of time and space, the analysis results of each other are targeted to form an integrated evolutionary cognition, thereby determining the long-term evolutionary trend. Specifically, first, the time span of the three-dimensional landscape scene is determined, and the evolutionary trends of the two branches in the time span block are integrated as the first evolutionary trend. At the same time, the time span is analyzed for the rate of change and divided into stages to clarify the mutation characteristics of the landscape pattern within the time span, that is, the time point where the rate of change is much higher than the normal value, and it is marked as a mutation evolution node, and multiple mutation evolution nodes are determined. Among them, the formation of mutation evolution nodes may be caused by abnormal fluctuations in natural driving forces, such as the outbreak of extreme climate events; or it may be caused by the concentrated release of human driving forces, such as the implementation of major engineering projects. Then, traverse multiple mutation evolution nodes, obtain a mutation evolution node each time, and for each mutation evolution node, extract the landscape pattern characteristics of the corresponding spatial range from the spatial span blocks of the qualitative comparison branch and the quantitative comparison branch, respectively, to form a multi-regional and multi-indicator node evolution trend in the spatial span of the node, thereby determining the second evolution trend.

[0049] Afterwards, the first evolutionary trend and the second evolutionary trend are nested and interactively iterated step by step by cross-validation. The spatial laws identified in the second evolutionary trend are fed back to the macro process of the first evolutionary trend for verification, and the mutation evolution nodes in the first evolutionary trend are fed back to the space of the second evolutionary trend to examine their discrimination. After multiple rounds of cross-validation and optimization adjustment, a historical evolutionary trend is formed in the time dimension and space dimension. Then, the historical evolutionary trend is used as a benchmark, and the bias of landscape evolution in the historical stage is analyzed, that is, whether there is a trend of continuous deviation in a certain direction compared with the normal development track, and this bias is used as a constraint. Next, a variety of scenarios are set and simulated for landscape evolution in the future time interval, including various internal and external factors and their combinations that may cause changes in landscape pattern, such as climate change, land use policies, and changes in economic levels. For each scenario, the development trajectory of the landscape pattern under the scenario conditions is simulated. Afterwards, the landscape evolution trajectory obtained under each scenario is compared with the normal evolution trajectory deduced based on the historical evolution trend, the deviation between the two is calculated, and the deviation is standardized to determine the predicted deviation trend.

[0050] S6: Integrate the long-term evolution trend, add the long-term comparative analysis results, and perform deviation exceeding limit positioning and landscape maintenance management based on the real-time deviation trend and the predicted deviation trend.

[0051] In the embodiment of the present application, for the acquired long-term evolution trend, a unified description framework is used to systematically integrate various types of long-term evolution trends, forming an overall understanding of the long-term evolution of the landscape pattern as the result of the long-term comparative analysis. Then, from the long-term comparative analysis results, two subsets of real-time deviation trend and predicted deviation trend are extracted. Among them, the real-time deviation trend reflects the degree of deviation of the actual state of the current landscape pattern from the historical normal evolution trajectory; the predicted deviation trend predicts the situation that the landscape pattern may change abnormally in the future under the existing conditions. According to the real-time deviation trend, it is judged whether the current landscape state has exceeded the acceptable normal change range. For those abnormal states that exceed the threshold, the deviation over-limit positioning work is carried out, including determining the specific spatial location, time node, deviation magnitude and other detailed information of the abnormal state. On this basis, combined with the analysis and prediction of the predicted deviation trend on the future development direction, the landscape maintenance management plan and implementation measures are formulated in a targeted manner to curb the further deviation of the landscape pattern and promote its recovery to the normal track as soon as possible. After the implementation of landscape maintenance and management measures, continue to track and monitor landscape dynamics, evaluate maintenance effects, promptly discover and solve new problems, adjust and optimize maintenance strategies when necessary, form a closed-loop control, and ensure the long-term healthy and orderly development of the landscape pattern.

[0052] In summary, the long-term time series comparative analysis method based on landscape three-dimensional scenes provided in the embodiments of the present application has the following technical effects:

[0053] Read multi-source landscape data and perform preprocessing and homologous data fitting, perform timestamp identification, determine effective scene data, realize landscape data assimilation, and establish effective scene data with consistent time, unified coordinates, and standardized semantics. The effective scene data is the spatiotemporal distribution data of the monitoring area, reflecting the dynamic change characteristics of the landscape pattern in time and space, and laying a data foundation for subsequent analysis. Carry out big data mining and build a defined state library. The defined state library is determined based on the relationship between landscape pattern type-pattern characteristics-driving factors-factor driving, revealing the semantic connection and causal law between various landscape elements, and providing knowledge guidance for qualitative and quantitative analysis. Combined with the defined state library, a long-term comparison model is established. The long-term comparison model includes qualitative comparison branches and quantitative comparison branches in parallel, and a lateral interactive branch is established to achieve a comprehensive and accurate analysis of landscape data. Connect to the visualization simulation platform, perform low-feature three-dimensional simulation based on effective scene data, and build a three-dimensional landscape scene. Among them, three-dimensional simulation includes entity simulation and virtual simulation. The three-dimensional landscape scene has seasonal and temporal characteristics. Through the integrated modeling and rendering of landscape entity objects and related information, and combined with the dynamic changes of landscape patterns with seasons and time, an accurate analysis environment is generated. Based on the long-term comparison model, with the standardized landscape as the baseline, the three-dimensional landscape scene is subjected to qualitative and quantitative temporal and spatial evolution comparison analysis and mutual verification to determine the long-term evolution trend. The long-term evolution trend includes real-time deviation trend and predicted deviation trend. The long-term evolution trend is updated synchronously with the passage of time, and is mutually verified through qualitative and quantitative analysis. The temporal and spatial evolution characteristics of the landscape are objectively and comprehensively portrayed, and the real-time update of landscape data is realized. Integrate the long-term evolution trend, add the long-term comparison analysis results, perform deviation over-limit positioning and landscape maintenance management for the real-time deviation trend and predicted deviation trend, and realize real-time maintenance management of landscape data. Example

[0054] Based on the same inventive concept as the long time series comparative analysis method based on landscape three-dimensional scene in the aforementioned embodiment, Figure 2 As shown, the embodiment of the present application provides a long time series comparison and analysis device based on a landscape three-dimensional scene, the device comprising:

[0055] A scene data determination unit 11 is used to read multi-source landscape data, perform pre-processing and homologous data fitting, perform timestamp identification, and determine effective scene data, wherein the effective scene data is the spatiotemporal distribution data of the monitoring area;

[0056] A state library building unit 12 is used to perform big data mining and build a defined state library, wherein the defined state library is determined based on landscape pattern type-pattern characteristics-driving factors-factor driving relationships;

[0057] A model building unit 13, used to build a long time series comparison model in combination with the defined state library, wherein the long time series comparison model includes a qualitative comparison branch and a quantitative comparison branch in parallel, and a lateral interaction branch is built;

[0058] The three-dimensional scene building unit 14 is used to connect to the visualization simulation platform, perform low-feature three-dimensional simulation based on the effective scene data, and build a landscape three-dimensional scene, wherein the three-dimensional simulation includes entity simulation and virtual simulation, wherein the landscape three-dimensional scene has seasonal and temporal characteristics;

[0059] The evolution trend determination unit 15 is used to perform a qualitative and quantitative spatiotemporal evolution comparison analysis and mutual verification on the three-dimensional landscape scene based on the long-term comparison model and the standardized landscape as a baseline to determine the long-term evolution trend, wherein the long-term evolution trend includes a real-time deviation trend and a predicted deviation trend, and the long-term evolution trend is synchronously updated with the passage of time;

[0060] The analysis result acquisition unit 16 is used to integrate the long-term evolution trend, add the long-term comparative analysis result, and perform deviation exceeding limit positioning and landscape maintenance management based on the real-time deviation trend and the predicted deviation trend.

[0061] Furthermore, the scene data determination unit 11 includes the following execution steps:

[0062] Guided by data preprocessing requirements, supervise the training of a preprocessing model, wherein the preprocessing model includes a source processing block and a multi-source fusion block;

[0063] Based on the source processing block, redundancy processing and registration enhancement processing are performed on the multi-source landscape data to determine effective multi-source data, wherein the redundancy processing includes redundancy screening and redundancy dimensionality reduction;

[0064] The valid multi-source data are traversed, data coaxial transformation and homologous data fusion are performed based on a unified coordinate system, and the valid scene data are integrated and determined.

[0065] Furthermore, the state library building unit 12 includes the following execution steps:

[0066] Identifying the mapping relationship between the pattern characteristics and driving factors, and determining a plurality of mapping groups;

[0067] Traversing the plurality of mapping groups, taking single factor driving and multi-factor joint driving as the benchmark, measuring the pattern feature evolution under the unit driving step, and determining the plurality of groups of factor driving relationships, wherein the plurality of mapping groups correspond one to one to the plurality of groups of factor driving relationships;

[0068] Integrate the multiple sets of factor-driven relationships and perform occasional marking of the pattern characteristics-driving factors.

[0069] Furthermore, the three-dimensional scene building unit 14 includes the following execution steps:

[0070] Based on the evolution requirement, the scene feature refinement is determined, and the evolution requirement is the low feature processing standard;

[0071] Based on the scene feature fineness, performing scene feature simplification and feature conversion processing on the effective scene data to determine the scene low feature distribution;

[0072] Based on the low-feature distribution of the scene, a three-dimensional simulated restoration of the landscape is performed to determine the three-dimensional landscape scene.

[0073] Furthermore, the three-dimensional scene building unit 14 also includes the following execution steps:

[0074] Traversing the low-feature distribution of the scene to determine the entity feature distribution and the virtual feature distribution;

[0075] Based on the entity feature distribution, landscape entity modeling is performed to determine the entity three-dimensional model;

[0076] Taking the light change as a benchmark and combining the virtual body feature distribution, a virtual body periodic evolution modeling is performed to determine a virtual body three-dimensional model, wherein the light change period is a virtual body modeling standard;

[0077] The three-dimensional landscape scene is generated by coordinating the entity three-dimensional model and the virtual three-dimensional model.

[0078] Further, the evolution trend determination unit 15 includes the following execution steps:

[0079] The qualitative comparison branch includes a twin network structure with the quantitative comparison branch, which respectively includes a time span block and a space span block, the time span blocks between the branches establish a first lateral interaction channel, and the space span blocks between the branches establish a second lateral interaction channel;

[0080] Based on the qualitative comparison branch and the quantitative comparison branch, parallel evolution analysis and lateral interaction analysis are performed on the three-dimensional landscape scene, and spatiotemporal fitting is performed to determine the long-term evolution trend.

[0081] Furthermore, the evolution trend determination unit 15 further includes the following execution steps:

[0082] Based on the first evolution trend of the time span, a plurality of mutation evolution nodes are determined;

[0083] Traversing the plurality of mutation evolution nodes, and determining a node evolution trend of a spatial span as a second evolution trend;

[0084] fusing the first evolution trend and the second evolution trend to determine a historical evolution trend;

[0085] Taking the historical evolution trend as a benchmark and the historical prediction bias as a constraint, the landscape evolution prediction of the preset time zone is carried out to determine the prediction deviation trend, and the prediction deviation trend is determined based on the standardized bias of the prediction evolution trend.

[0086] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0087] Furthermore, the first or second mentioned above may not only represent an order relationship, but may also represent a specific concept, and / or refer to multiple elements that can be selected individually or in full. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A long-term comparative analysis method based on landscape three-dimensional scenes, characterized in that: The method comprises: Read multi-source landscape data and perform preprocessing and fitting with homologous data, perform timestamp identification, and determine effective scene data, which is the spatiotemporal distribution data of the monitoring area; Conduct big data mining and build a definition state library, which is determined based on landscape pattern type-pattern characteristics-driving factors-factor-driving relationship, where the landscape pattern type refers to the components of the landscape and their spatial configuration; the pattern characteristics refer to the spatial geometric characteristics and topological relationships of the landscape pattern; the driving factors refer to the internal reasons that lead to changes in the landscape pattern; and the factor-driving relationship describes the interaction mechanism between the driving factors and the landscape pattern; In combination with the defined state library, a long time series comparison model is established, wherein the long time series comparison model includes a qualitative comparison branch and a quantitative comparison branch in parallel, and a lateral interaction branch is established; Connecting to a visualization simulation platform, performing low-feature three-dimensional simulation based on the effective scene data, and building a three-dimensional landscape scene, wherein the three-dimensional simulation includes entity simulation and virtual simulation, and wherein the three-dimensional landscape scene has seasonal and temporal characteristics; Based on the long-term comparison model, taking the standardized landscape as the baseline, the three-dimensional landscape scene is subjected to qualitative and quantitative temporal and spatial evolution comparative analysis and mutual verification to determine the long-term evolution trend, wherein the long-term evolution trend includes a real-time deviation trend and a predicted deviation trend, and the long-term evolution trend is synchronously updated with the passage of time; Integrate the long-term evolution trend, add the long-term comparative analysis results, and perform deviation exceeding limit positioning and landscape maintenance management on the real-time deviation trend and the predicted deviation trend; The defined state library is determined based on landscape pattern type-pattern characteristics-driving factors-factor driving relationships, including: Identifying the mapping relationship between the pattern characteristics and driving factors, and determining a plurality of mapping groups; Traversing the plurality of mapping groups, taking single factor driving and multi-factor joint driving as the benchmark, measuring the pattern feature evolution under the unit driving step, and determining the plurality of groups of factor driving relationships, wherein the plurality of mapping groups correspond one to one to the plurality of groups of factor driving relationships; Integrate the multiple groups of factor-driven relationships and perform occasional marking of the pattern characteristics-driving factors; The determining of the long-term time series evolution trend comprises: The qualitative comparison branch and the quantitative comparison branch are twin network structures, respectively including a time span block and a space span block, the time span blocks between the branches establish a first lateral interaction channel, and the space span blocks between the branches establish a second lateral interaction channel; Based on the qualitative comparison branch and the quantitative comparison branch, parallel evolution analysis and lateral interaction analysis are performed on the three-dimensional landscape scene, and spatiotemporal fitting is performed to determine the long-term evolution trend.

2. The method according to claim 1, characterized in that The determining of the effective scene data comprises: Guided by data preprocessing requirements, supervise the training of a preprocessing model, wherein the preprocessing model includes a source processing block and a multi-source fusion block; Based on the source processing block, redundancy processing and registration enhancement processing are performed on the multi-source landscape data to determine effective multi-source data, wherein the redundancy processing includes redundancy screening and redundancy dimensionality reduction; The valid multi-source data are traversed, data coaxial transformation and homologous data fusion are performed based on a unified coordinate system, and the valid scene data are integrated and determined.

3. The method according to claim 1, characterized in that Based on the effective scene data, low-feature three-dimensional simulation is performed to build a landscape three-dimensional scene, including: Based on the evolution requirement, the scene feature refinement is determined, and the evolution requirement is the low feature processing standard; Based on the scene feature fineness, performing scene feature simplification and feature conversion processing on the effective scene data to determine the scene low feature distribution; Based on the low-feature distribution of the scene, a three-dimensional simulated restoration of the landscape is performed to determine the three-dimensional landscape scene.

4. The method according to claim 3, characterized in that Based on the low-feature distribution of the scene, a three-dimensional realistic restoration of the landscape is performed, including: Traversing the low-feature distribution of the scene to determine the entity feature distribution and the virtual feature distribution; Based on the entity feature distribution, landscape entity modeling is performed to determine the entity three-dimensional model; Taking the light change as a benchmark and combining the virtual body feature distribution, a virtual body periodic evolution modeling is performed to determine a virtual body three-dimensional model, wherein the light change period is a virtual body modeling standard; The three-dimensional landscape scene is generated by coordinating the entity three-dimensional model and the virtual three-dimensional model.

5. The method according to claim 1, characterized in that The performing spatiotemporal fitting to determine the long-term time series evolution trend includes: Based on the first evolution trend of the time span, a plurality of mutation evolution nodes are determined; Traversing the plurality of mutation evolution nodes, and determining a node evolution trend of a spatial span as a second evolution trend; fusing the first evolution trend and the second evolution trend to determine a historical evolution trend; Taking the historical evolution trend as a benchmark and the historical prediction bias as a constraint, the landscape evolution prediction of the preset time zone is carried out to determine the prediction deviation trend, and the prediction deviation trend is determined based on the standardized bias of the prediction evolution trend.

6. A long-term time series comparison and analysis device based on a three-dimensional landscape scene, characterized in that: For implementing the long-term time series comparative analysis method based on a landscape three-dimensional scene according to any one of claims 1 to 5, the device comprises: A scene data determination unit, the scene data determination unit is used to read multi-source landscape data and perform pre-processing and fitting with homologous data, perform timestamp identification, and determine effective scene data, the effective scene data being the spatiotemporal distribution data of the monitoring area; A state library building unit, the state library building unit is used to perform big data mining and build a definition state library, the definition state library is determined based on landscape pattern type-pattern characteristics-driving factors-factor driving relationship, wherein the landscape pattern type refers to the components of the landscape and their spatial configuration; the pattern characteristics refer to the spatial geometric characteristics and topological relationships of the landscape pattern; the driving factors refer to the internal reasons that lead to changes in the landscape pattern; and the factor driving relationship describes the interaction mechanism between the driving factors and the landscape pattern; A model building unit, the model building unit is used to build a long time series comparison model in combination with the defined state library, the long time series comparison model includes a qualitative comparison branch and a quantitative comparison branch in parallel, and a lateral interaction branch is established; A three-dimensional scene building unit, the three-dimensional scene building unit is used to connect to the visualization simulation platform, perform low-feature three-dimensional simulation based on the effective scene data, and build a landscape three-dimensional scene, wherein the three-dimensional simulation includes entity simulation and virtual simulation, wherein the landscape three-dimensional scene has seasonal and temporal characteristics; An evolution trend determination unit, the evolution trend determination unit is used to perform qualitative and quantitative spatiotemporal evolution comparison analysis and mutual verification on the three-dimensional landscape scene based on the long-term comparison model and the standardized landscape as a baseline, to determine the long-term evolution trend, the long-term evolution trend includes a real-time deviation trend and a predicted deviation trend, and the long-term evolution trend is synchronously updated with the passage of time; An analysis result acquisition unit is used to integrate the long-term evolution trend, add the long-term comparative analysis result, and perform deviation exceeding limit positioning and landscape maintenance management based on the real-time deviation trend and the predicted deviation trend.

Images